build/dashboard/builders.go

3229 строки
108 KiB
Go
Исходник Обычный вид История

// Copyright 2015 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package dashboard contains shared configuration and logic used by various
// pieces of the Go continuous build system.
package dashboard
import (
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
"fmt"
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
"os"
"sort"
"strconv"
"strings"
"time"
"golang.org/x/build/buildenv"
"golang.org/x/build/internal/gophers"
"golang.org/x/build/maintner/maintnerd/maintapi/version"
dashboard, cmd/coordinator, maintner/maintnerd: add support for BuildConfig.MinimumGoVersion field The new BuildConfig.MinimumGoVersion field specifies the minimum Go version the builder is allowed to use. It's useful when some of the builders are too new, and do not support all of the supported Go releases (e.g., openbsd-amd64-64 and freebsd-amd64-12_0 currently require Go 1.11 and don't work on Go 1.10). It only needs to be set when a builder doesn't support all supported Go releases, since we don't typically test unsupported Go releases. To allow cmd/coordinator to use this field and filter out work it receives from maintner/maintnerd's GoFindTryWork RPC call, we add a GoVersion slice to apipb.GerritTryWorkItem, and populate it in maintapi.apiService.GoFindTryWork method. For trybots on the Go repo, the GoVersion field is determined from the branch name. For "release-branch.goX.Y" branches, it parses out the major-minor Go version from the branch name. For master and other branches, it assumes the latest Go release. For trybots on subrepos, we already have the Go version information available, so use it directly. Afterwards, all that's left is to modify newTrySet in cmd/coordinator to make use of BuildConfig.MinimumGoVersion and work.GoVersion to skip builders that are too new for the Go version that needs to be tested. Fixes golang/go#29265 Change-Id: I50b01830647e33e37e9eb8b89e0f2518812fa44f Reviewed-on: https://go-review.googlesource.com/c/155463 Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2019-01-12 00:44:01 +03:00
"golang.org/x/build/types"
)
// slowBotAliases maps short names from TRY= comments to which builder to run.
//
// TODO: we'll likely expand this, or move it, or change the matching
// syntax entirely. This is a first draft.
var slowBotAliases = map[string]string{
// Known missing builders:
"ios-amd64": "", // There is no builder for the iOS Simulator. See issues 42100 and 42177.
"386": "linux-386",
"aix": "aix-ppc64",
"amd64": "linux-amd64",
"android": "android-amd64-emu",
"android-386": "android-386-emu",
"android-amd64": "android-amd64-emu",
"android-arm": "android-arm-corellium",
"android-arm64": "android-arm64-corellium",
"arm": "linux-arm-aws",
"arm64": "linux-arm64",
"darwin": "darwin-amd64-12_0",
"darwin-amd64": "darwin-amd64-12_0",
"darwin-arm64": "darwin-arm64-12",
"ios-arm64": "ios-arm64-corellium",
"dragonfly": "dragonfly-amd64-622",
"dragonfly-amd64": "dragonfly-amd64-622",
"freebsd": "freebsd-amd64-13_0",
"freebsd-386": "freebsd-386-13_0",
"freebsd-amd64": "freebsd-amd64-13_0",
"freebsd-arm": "freebsd-arm-paulzhol",
"freebsd-arm64": "freebsd-arm64-dmgk",
"freebsd-riscv64": "freebsd-riscv64-unmatched",
"illumos": "illumos-amd64",
"ios": "ios-arm64-corellium",
"js": "js-wasm",
"linux": "linux-amd64",
"linux-arm": "linux-arm-aws",
"linux-loong64": "linux-loong64-3a5000",
"linux-mips": "linux-mips-rtrk",
"linux-mips64": "linux-mips64-rtrk",
"linux-mips64le": "linux-mips64le-rtrk",
"linux-mipsle": "linux-mipsle-rtrk",
"linux-ppc64": "linux-ppc64-sid-buildlet",
"linux-ppc64le": "linux-ppc64le-buildlet",
"linux-ppc64le-power9": "linux-ppc64le-power9osu",
"linux-ppc64le-power10": "linux-ppc64le-power10osu",
"linux-riscv64": "linux-riscv64-unmatched",
"linux-s390x": "linux-s390x-ibm",
"longtest": "linux-amd64-longtest",
"loong64": "linux-loong64-3a5000",
"mips": "linux-mips-rtrk",
"mips64": "linux-mips64-rtrk",
"mips64le": "linux-mips64le-rtrk",
"mipsle": "linux-mipsle-rtrk",
"netbsd": "netbsd-amd64-9_3",
"netbsd-386": "netbsd-386-9_3",
"netbsd-amd64": "netbsd-amd64-9_3",
"netbsd-arm": "netbsd-arm-bsiegert",
"netbsd-arm64": "netbsd-arm64-bsiegert",
"nocgo": "linux-amd64-nocgo",
"openbsd": "openbsd-amd64-72",
"openbsd-386": "openbsd-386-72",
"openbsd-amd64": "openbsd-amd64-72",
"openbsd-arm": "openbsd-arm-jsing",
"openbsd-arm64": "openbsd-arm64-jsing",
"openbsd-mips64": "openbsd-mips64-jsing",
"plan9": "plan9-arm",
"plan9-386": "plan9-386-0intro",
"plan9-amd64": "plan9-amd64-0intro",
"ppc64": "linux-ppc64-sid-buildlet",
"ppc64le": "linux-ppc64le-buildlet",
"ppc64lep9": "linux-ppc64le-power9osu",
"ppc64lep10": "linux-ppc64le-power10osu",
"riscv64": "linux-riscv64-unmatched",
"s390x": "linux-s390x-ibm",
"solaris": "solaris-amd64-oraclerel",
"solaris-amd64": "solaris-amd64-oraclerel",
"wasm": "js-wasm",
"windows": "windows-amd64-2016",
"windows-386": "windows-386-2008",
"windows-amd64": "windows-amd64-2016",
"windows-arm": "windows-arm-zx2c4",
// TODO:(golang/go#47019) - uncomment when builder returns and
// update test.
// "windows-arm64": "windows-arm64-11",
}
// Builders are the different build configurations.
// The keys are like "darwin-amd64" or "linux-386-387".
// This map should not be modified by other packages.
// Initialization happens below, via calls to addBuilder.
var Builders = map[string]*BuildConfig{}
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
// GoBootstrap is the bootstrap Go version.
// Bootstrap Go builds with this name must be in the bucket,
// usually uploaded by 'genbootstrap -upload all'.
const GoBootstrap = "go1.17.13"
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
// Hosts contains the names and configs of all the types of
// buildlets. They can be VMs, containers, or dedicated machines.
//
// Please keep table sorted by map key.
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
var Hosts = map[string]*HostConfig{
"host-aix-ppc64-osuosl": {
Notes: "AIX 7.2 VM on OSU; run by Tony Reix",
Owners: []*gophers.Person{gh("trex58")},
IsReverse: true,
ExpectNum: 1,
},
"host-android-arm64-corellium-android": {
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
Notes: "Virtual Android devices hosted by Zenly on Corellium; see issues 31722 and 40523",
Owners: []*gophers.Person{gh("steeve"), gh("changkun")}, // See https://groups.google.com/g/golang-dev/c/oiuIE7qrWp0.
IsReverse: true,
ExpectNum: 3,
GoBootstrap: "none", // image has Go 1.15.3 (go.dev/issue/54246); cannot access storage.googleapis.com
env: []string{
"GOROOT_BOOTSTRAP=/data/data/com.termux/files/home/go-android-arm64-bootstrap",
// Only run one job at a time to avoid the OOM killer.
// Issue 50084.
"GOMAXPROCS=1",
},
},
"host-darwin-amd64-10_14-aws": {
IsReverse: true,
ExpectNum: 2,
Notes: "AWS macOS Mojave (10.14) VM under QEMU",
SSHUsername: "gopher",
HermeticReverse: true, // we destroy the VM when done & recreate
GoogleReverse: true,
},
"host-darwin-amd64-10_15-aws": {
IsReverse: true,
ExpectNum: 2,
Notes: "AWS macOS Catalina (10.15) VM under QEMU",
SSHUsername: "gopher",
HermeticReverse: true, // we destroy the VM when done & recreate
GoogleReverse: true,
},
"host-darwin-amd64-11-aws": {
IsReverse: true,
ExpectNum: 2,
Notes: "AWS macOS Big Sur (11) VM under QEMU",
SSHUsername: "gopher",
HermeticReverse: true, // we destroy the VM when done & recreate
GoogleReverse: true,
},
"host-darwin-amd64-12-aws": {
IsReverse: true,
ExpectNum: 6,
Notes: "AWS macOS Monterey (12) VM under QEMU",
SSHUsername: "gopher",
HermeticReverse: true, // we destroy the VM when done & recreate
GoogleReverse: true,
},
"host-darwin-amd64-13-aws": {
IsReverse: true,
ExpectNum: 2,
Notes: "AWS macOS Ventura (13) VM under QEMU",
SSHUsername: "gopher",
HermeticReverse: true, // we destroy the VM when done & recreate
GoogleReverse: true,
},
"host-darwin-arm64-11": {
IsReverse: true,
Notes: "macOS Big Sur (11) ARM64 (M1) on Mac minis in a Google office",
ExpectNum: 3,
SSHUsername: "gopher",
GoogleReverse: true,
},
"host-darwin-arm64-12": {
IsReverse: true,
ExpectNum: 3,
Notes: "macOS Monterey (12) ARM64 (M1) on Mac minis in a Google office",
SSHUsername: "gopher",
GoogleReverse: true,
},
"host-dragonfly-amd64-622": {
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
Notes: "DragonFly BSD 6.2.2 on GCE, built from build/env/dragonfly-amd64",
VMImage: "dragonfly-amd64-622",
SSHUsername: "root",
},
"host-freebsd-amd64-12_3": {
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
VMImage: "freebsd-amd64-123-stable-20211230",
Notes: "FreeBSD 12.3; GCE VM, built from build/env/freebsd-amd64",
SSHUsername: "gopher",
},
"host-freebsd-amd64-13_0": {
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
VMImage: "freebsd-amd64-130-stable-20211230",
Notes: "FreeBSD 13.0; GCE VM, built from build/env/freebsd-amd64",
SSHUsername: "gopher",
},
"host-freebsd-arm-paulzhol": {
IsReverse: true,
ExpectNum: 1,
Notes: "Raspberry Pi 3 Model B, FreeBSD 13.1-RELEASE with SCHED_4BSD",
Owners: []*gophers.Person{gh("paulzhol")},
},
"host-freebsd-arm64-dmgk": {
IsReverse: true,
ExpectNum: 1,
Notes: "AWS EC2 a1.large 2 vCPU 4GiB RAM, FreeBSD 12.1-STABLE",
Owners: []*gophers.Person{gh("dmgk")},
},
"host-freebsd-riscv64-unmatched": {
IsReverse: true,
ExpectNum: 1,
Notes: "SiFive HiFive Unmatched RISC-V board. 16 GB RAM. FreeBSD 13.1-RELEASE",
Owners: []*gophers.Person{gh("mengzhuo")},
GoBootstrap: "none",
env: []string{
"GOROOT_BOOTSTRAP=/home/gopher/go-freebsd-riscv64-bootstrap",
},
},
"host-illumos-amd64-jclulow": {
Notes: "SmartOS base64@19.1.0 zone",
Owners: []*gophers.Person{gh("jclulow")},
IsReverse: true,
ExpectNum: 1,
SSHUsername: "gobuild",
},
"host-ios-arm64-corellium-ios": {
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
Notes: "Virtual iOS devices hosted by Zenly on Corellium; see issues 31722 and 40523",
Owners: []*gophers.Person{gh("steeve"), gh("changkun")}, // See https://groups.google.com/g/golang-dev/c/oiuIE7qrWp0.
IsReverse: true,
ExpectNum: 3,
GoBootstrap: "none", // image has devel d6c4583ad4 (pre-Go 1.18) Dec 2021 (go.dev/issue/54246); cannot access storage.googleapis.com
env: []string{
"GOROOT_BOOTSTRAP=/var/root/go-ios-arm64-bootstrap",
},
},
"host-linux-amd64-alpine": {
Notes: "Alpine container",
ContainerImage: "linux-x86-alpine:latest",
SSHUsername: "root",
},
"host-linux-amd64-androidemu": {
Notes: "Debian Bullseye w/ Android SDK + emulator (use nested virt)",
ContainerImage: "android-amd64-emu:bff27c0c9263",
KonletVMImage: "android-amd64-emu-bullseye",
NestedVirt: true,
SSHUsername: "root",
},
"host-linux-amd64-bullseye": {
Notes: "Debian Bullseye",
ContainerImage: "linux-x86-bullseye:latest",
SSHUsername: "root",
},
"host-linux-amd64-bullseye-vmx": {
Notes: "Debian Bullseye w/ Nested Virtualization (VMX CPU bit) enabled",
ContainerImage: "linux-x86-bullseye:latest",
NestedVirt: true,
SSHUsername: "root",
},
"host-linux-amd64-buster": {
Notes: "Debian Buster",
ContainerImage: "linux-x86-buster:latest",
SSHUsername: "root",
},
"host-linux-amd64-clang": {
Notes: "Container with clang.",
ContainerImage: "linux-x86-clang:latest",
SSHUsername: "root",
},
"host-linux-amd64-fedora": {
Notes: "Fedora 30",
ContainerImage: "linux-x86-fedora:latest",
SSHUsername: "root",
},
"host-linux-amd64-js-wasm": {
Notes: "Container with Node.js 14 for testing js/wasm.",
ContainerImage: "js-wasm:latest",
SSHUsername: "root",
},
"host-linux-amd64-js-wasm-node18": {
Notes: "Container with Node.js 18 for testing js/wasm.",
ContainerImage: "js-wasm-node18:latest",
SSHUsername: "root",
},
"host-linux-amd64-localdev": {
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
IsReverse: true,
ExpectNum: 0,
Notes: "for localhost development of buildlets/gomote/coordinator only",
SSHUsername: os.Getenv("USER"),
},
"host-linux-amd64-perf": {
Notes: "Cascade Lake performance testing machines",
machineType: "c2-standard-8", // C2 has precisely defined, consistent server architecture.
ContainerImage: "linux-x86-bullseye:latest",
SSHUsername: "root",
CustomDeleteTimeout: 8 * time.Hour,
},
"host-linux-amd64-s390x-cross": {
Notes: "Container with s390x cross-compiler.",
ContainerImage: "linux-s390x-cross:latest",
},
"host-linux-amd64-sid": {
Notes: "Debian sid, updated occasionally.",
ContainerImage: "linux-x86-sid:latest",
SSHUsername: "root",
},
"host-linux-amd64-stretch": {
Notes: "Debian Stretch",
ContainerImage: "linux-x86-stretch:latest",
SSHUsername: "root",
},
"host-linux-amd64-wsl": {
Notes: "Windows 10 WSL2 Ubuntu",
Owners: []*gophers.Person{gh("mengzhuo")},
IsReverse: true,
ExpectNum: 2,
},
"host-linux-arm-aws": {
Notes: "Debian Buster, EC2 arm instance. See x/build/env/linux-arm/aws",
VMImage: "ami-07409163bccd5ac4d",
ContainerImage: "gobuilder-arm-aws:latest",
machineType: "m6g.xlarge",
isEC2: true,
SSHUsername: "root",
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
},
"host-linux-arm64-aws": {
Notes: "Debian Buster, EC2 arm64 instance. See x/build/env/linux-arm64/aws",
VMImage: "ami-03089323a1d38e652",
ContainerImage: "gobuilder-arm64-aws:latest",
machineType: "m6g.xlarge",
isEC2: true,
SSHUsername: "root",
},
"host-linux-arm64-bullseye": {
Notes: "Debian Bullseye",
ContainerImage: "linux-arm64-bullseye:latest",
machineType: "t2a",
SSHUsername: "root",
cosArchitecture: CosArchARM64,
},
"host-linux-arm64-bullseye-high-disk": {
Notes: "Debian Bullseye, larger boot disk size",
ContainerImage: "linux-arm64-bullseye:latest",
machineType: "t2a",
SSHUsername: "root",
cosArchitecture: CosArchARM64,
RootDriveSizeGB: 20,
},
"host-linux-loong64-3a5000": {
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
Notes: "Loongson 3A5000 Box hosted by Loongson; loong64 is the short name of LoongArch 64 bit version",
Owners: []*gophers.Person{gh("XiaodongLoong"), gh("abner-chenc")},
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
IsReverse: true,
ExpectNum: 5,
GoBootstrap: "none",
env: []string{
"GOROOT_BOOTSTRAP=/usr/lib/go-linux-loong64-bootstrap",
},
},
"host-linux-mips64-rtrk": {
Notes: "cavium,rhino_utm8 board hosted at RT-RK.com; quad-core cpu, 8GB of ram and 240GB ssd disks.",
Owners: []*gophers.Person{gh("draganmladjenovic")}, // See https://github.com/golang/go/issues/53574#issuecomment-1169891255.
IsReverse: true,
ExpectNum: 1,
},
"host-linux-mips64le-rtrk": {
Notes: "cavium,rhino_utm8 board hosted at RT-RK.com; quad-core cpu, 8GB of ram and 240GB ssd disks.",
Owners: []*gophers.Person{gh("draganmladjenovic")}, // See https://github.com/golang/go/issues/53574#issuecomment-1169891255.
IsReverse: true,
ExpectNum: 1,
},
"host-linux-ppc64-sid": {
Notes: "Debian sid; run by Go team on osuosl.org",
Owners: []*gophers.Person{gh("pmur")},
IsReverse: true,
ExpectNum: 5,
SSHUsername: "root",
HermeticReverse: true,
},
"host-linux-ppc64le-osu": {
Notes: "Ubuntu 20.04; run by Go team on osuosl.org; see x/build/env/linux-ppc64le/osuosl",
Owners: []*gophers.Person{gh("pmur")},
IsReverse: true,
ExpectNum: 5,
SSHUsername: "root",
HermeticReverse: true,
},
"host-linux-ppc64le-power10-osu": {
Notes: "Ubuntu 20.04; run by Go team on osuosl.org; see x/build/env/linux-ppc64le/osuosl",
Owners: []*gophers.Person{gh("pmur")},
IsReverse: true,
// GOPPC64=power10 is only supported in go1.20 and later. The container provides a patched boostrap compiler.
env: []string{
"GOPPC64=power10",
"GOROOT_BOOTSTRAP=/usr/local/go-bootstrap"},
SSHUsername: "root",
HermeticReverse: true,
},
"host-linux-ppc64le-power9-osu": {
Notes: "Ubuntu 20.04; run by Go team on osuosl.org; see x/build/env/linux-ppc64le/osuosl",
Owners: []*gophers.Person{gh("pmur")},
IsReverse: true,
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
env: []string{"GOPPC64=power9"},
SSHUsername: "root",
HermeticReverse: true,
},
"host-linux-riscv64-joelsing": {
Notes: "SiFive HiFive Unleashed RISC-V board. 8 GB RAM, 4 cores.",
IsReverse: true,
ExpectNum: 1,
Owners: []*gophers.Person{gh("4a6f656c")},
},
"host-linux-riscv64-unmatched": {
Notes: "SiFive HiFive Unmatched RISC-V board. 16 GB RAM, 4 cores.",
IsReverse: true,
ExpectNum: 2,
Owners: []*gophers.Person{gh("mengzhuo")},
},
"host-linux-s390x": {
Notes: "run by IBM",
Owners: []*gophers.Person{gh("Vishwanatha-HD"), gh("srinivas-pokala")},
IsReverse: true,
ExpectNum: 2, // See https://github.com/golang/go/issues/49557#issuecomment-969148789.
},
"host-netbsd-386-9_3": {
VMImage: "netbsd-i386-9-3-202211120320",
Notes: "NetBSD 9.3; GCE VM is built from script in build/env/netbsd-386",
machineType: "n2", // force Intel; see go.dev/issue/49209
SSHUsername: "root",
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-netbsd-amd64-9_3": {
VMImage: "netbsd-amd64-9-3-202211120320v2",
Notes: "NetBSD 9.3; GCE VM is built from script in build/env/netbsd-amd64",
machineType: "n2", // force Intel; see go.dev/issue/49209
SSHUsername: "root",
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-netbsd-arm-bsiegert": {
IsReverse: true,
ExpectNum: 1,
Owners: []*gophers.Person{gh("bsiegert")},
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-netbsd-arm64-bsiegert": {
IsReverse: true,
ExpectNum: 1,
Owners: []*gophers.Person{gh("bsiegert")},
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-openbsd-386-71": {
VMImage: "openbsd-386-71",
machineType: "n2", // force Intel; see go.dev/issue/49209
Notes: "OpenBSD 7.1; GCE VM, built from build/env/openbsd-386",
SSHUsername: "gopher",
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-openbsd-386-72": {
VMImage: "openbsd-386-72",
machineType: "n2", // force Intel; see go.dev/issue/49209
Notes: "OpenBSD 7.2; GCE VM, built from build/env/openbsd-386",
SSHUsername: "gopher",
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-openbsd-amd64-71": {
VMImage: "openbsd-amd64-71",
machineType: "n2", // force Intel; see go.dev/issue/49209
Notes: "OpenBSD 7.1; GCE VM, built from build/env/openbsd-amd64",
SSHUsername: "gopher",
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-openbsd-amd64-72": {
VMImage: "openbsd-amd64-72",
machineType: "n2", // force Intel; see go.dev/issue/49209
Notes: "OpenBSD 7.2; GCE VM, built from build/env/openbsd-amd64",
SSHUsername: "gopher",
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-openbsd-arm-joelsing": {
IsReverse: true,
ExpectNum: 1,
Owners: []*gophers.Person{gh("4a6f656c")},
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-openbsd-arm64-joelsing": {
IsReverse: true,
ExpectNum: 1,
Owners: []*gophers.Person{gh("4a6f656c")},
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-openbsd-mips64-joelsing": {
IsReverse: true,
ExpectNum: 1,
Owners: []*gophers.Person{gh("4a6f656c")},
GoBootstrap: "go1.19.2", // Go 1.17 is too old; see go.dev/issue/42422
},
"host-plan9-386-0intro": {
IsReverse: true,
ExpectNum: 1,
Notes: "QEMU VM, Plan 9 from Bell Labs",
Owners: []*gophers.Person{gh("0intro")},
},
"host-plan9-386-gce": {
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
VMImage: "plan9-386-v7",
Notes: "Plan 9 from 0intro; GCE VM, built from build/env/plan9-386",
env: []string{"GO_TEST_TIMEOUT_SCALE=3"},
},
"host-plan9-amd64-0intro": {
IsReverse: true,
ExpectNum: 1,
Notes: "QEMU VM, Plan 9 from Bell Labs, 9k kernel",
Owners: []*gophers.Person{gh("0intro")},
},
"host-plan9-arm-0intro": {
IsReverse: true,
ExpectNum: 1,
Notes: "Raspberry Pi 3 Model B, Plan 9 from Bell Labs",
Owners: []*gophers.Person{gh("0intro")},
},
"host-solaris-oracle-amd64-oraclerel": {
Notes: "Oracle Solaris amd64 Release System",
HostArch: "solaris-amd64",
Owners: []*gophers.Person{gh("rorth")}, // https://github.com/golang/go/issues/15581#issuecomment-550368581
IsReverse: true,
ExpectNum: 1,
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
},
"host-windows-amd64-2008": {
VMImage: "windows-amd64-server-2008r2-v8",
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
SSHUsername: "gopher",
},
"host-windows-amd64-2008-oldcc": {
VMImage: "windows-amd64-server-2008r2-v7",
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
SSHUsername: "gopher",
},
"host-windows-amd64-2012": {
VMImage: "windows-amd64-server-2012r2-v8",
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
SSHUsername: "gopher",
},
"host-windows-amd64-2012-oldcc": {
VMImage: "windows-amd64-server-2012r2-v7",
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
SSHUsername: "gopher",
},
"host-windows-amd64-2016": {
VMImage: "windows-amd64-server-2016-v8",
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
SSHUsername: "gopher",
},
"host-windows-amd64-2016-big": {
VMImage: "windows-amd64-server-2016-v8",
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
machineType: "e2-standard-16",
SSHUsername: "gopher",
},
"host-windows-amd64-2016-big-oldcc": {
VMImage: "windows-amd64-server-2016-v7",
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
machineType: "e2-standard-16",
SSHUsername: "gopher",
},
"host-windows-amd64-2016-oldcc": {
VMImage: "windows-amd64-server-2016-v7",
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
SSHUsername: "gopher",
},
"host-windows-arm64-zx2c4": {
IsReverse: true,
ExpectNum: 0,
Owners: []*gophers.Person{gh("zx2c4")},
},
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
}
func gh(githubUsername string) *gophers.Person {
p := gophers.GetPerson("@" + githubUsername)
if p == nil {
panic("person with GitHub username " + githubUsername + " does not exist in the golang.org/x/build/internal/gophers package")
}
return p
}
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
func init() {
for key, c := range Hosts {
if key == "" {
panic("empty string key in Hosts")
}
if c.HostType == "" {
c.HostType = key
}
if c.HostType != key {
panic(fmt.Sprintf("HostType %q != key %q", c.HostType, key))
}
if c.HostArch == "" {
f := strings.Split(c.HostType, "-")
if len(f) < 3 {
panic(fmt.Sprintf("invalid HostType %q", c.HostType))
}
c.HostArch = f[1] + "-" + f[2] // "linux-amd64"
if f[2] == "arm" {
c.HostArch += "-7" // assume newer ARM
}
}
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
if c.GoBootstrap == "" {
c.GoBootstrap = GoBootstrap
}
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
nSet := 0
if c.VMImage != "" {
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
nSet++
}
if c.ContainerImage != "" && !c.isEC2 {
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
nSet++
}
if c.IsReverse {
nSet++
}
if nSet != 1 {
panic(fmt.Sprintf("exactly one of VMImage, ContainerImage, IsReverse must be set for host %q; got %v", key, nSet))
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
}
}
}
// CosArch defines the diffrent COS images types used.
type CosArch string
const (
CosArchAMD64 CosArch = "cos-stable" // COS image for AMD64 architecture
CosArchARM64 = "cos-arm64-stable" // COS image for ARM64 architecture
)
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
// A HostConfig describes the available ways to obtain buildlets of
// different types. Some host configs can serve multiple
// builders. For example, a host config of "host-linux-amd64-bullseye" can
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
// serve linux-amd64, linux-amd64-race, linux-386, linux-386-387, etc.
type HostConfig struct {
// HostType is the unique name of this host config. It is also
// the key in the Hosts map.
HostType string
// HostArch is a string identifying the host architecture, to decide which binaries to run on it.
// If unset, it is derived in func init from the HostType
// (for example, host-linux-amd64-foo has HostArch "linux-amd64").
// For clarity, the HostArch for 32-bit arm always ends in -5 or -7
// to specify the GOARM value; implicit HostArch always use -7.
// (This specificity is necessary because there is no one value that
// works on all 32-bit ARM chips on non-Linux operating systems.)
//
// If set explicitly, HostArch should have the form "GOOS-GOARCH-suffix",
// where suffix is the GO$GOARCH setting (that is, GOARM for GOARCH=arm).
// For example "openbsd-arm-5" for GOOS=openbsd GOARCH=arm GOARM=5.
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
//
// (See the BuildletBinaryURL and GoBootstrapURL methods.)
HostArch string
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
// GoBootstrap is the version of Go to use for bootstrap.
// A bootstrap toolchain built with that version must be in the bucket.
// If unset, it is set in func init to GoBootstrap (the global constant).
//
// If GoBootstrap is set to "none", it means this buildlet is not given a new bootstrap
// toolchain for each build, typically because it cannot download from
// storage.googleapis.com.
//
// (See the GoBootstrapURL method.)
GoBootstrap string
// Exactly 1 of these must be set (with the exception of EC2 instances).
// An EC2 instance may run a container inside a VM. In that case, a VMImage
// and ContainerImage will both be set.
VMImage string // e.g. "openbsd-amd64-60"
ContainerImage string // e.g. "linux-buildlet-std:latest" (suffix after "gcr.io/<PROJ>/")
IsReverse bool // if true, only use the reverse buildlet pool
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
// GCE options, if VMImage != "" || ContainerImage != ""
machineType string // optional GCE instance type ("n2-standard-4") or instance class ("n2")
RegularDisk bool // if true, use spinning disk instead of SSD
MinCPUPlatform string // optional. e2 instances are not supported. see https://cloud.google.com/compute/docs/instances/specify-min-cpu-platform.
cosArchitecture CosArch // optional. GCE instances which use COS need the architecture set. Default: CosArchAMD64
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
// EC2 options
isEC2 bool // if true, the instance is configured to run on EC2
// GCE or EC2 options:
//
// CustomDeleteTimeout is an optional custom timeout after which the VM is forcibly destroyed and the build is retried.
// Zero duration defers decision to internal/coordinator/pool package, which is enough for longtest post-submit builds.
// (This is generally an internal implementation detail, currently left behind only for the -perf builder.)
CustomDeleteTimeout time.Duration
// Reverse options
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
ExpectNum int // expected number of reverse buildlets of this type
HermeticReverse bool // whether reverse buildlet has fresh env per conn
GoogleReverse bool // whether this reverse builder is owned by Google
// Container image options, if ContainerImage != "":
NestedVirt bool // container requires VMX nested virtualization. e2 and n2d instances are not supported.
KonletVMImage string // optional VM image (containing konlet) to use instead of default
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
// Optional base env.
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
env []string
Owners []*gophers.Person // owners; empty means golang-dev
Notes string // notes for humans
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
SSHUsername string // username to ssh as, empty means not supported
RootDriveSizeGB int64 // optional, GCE instance root size in base-2 GB. Default: default size for instance type.
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
}
// CosArchitecture returns the COS architecture to use with the host. The default is CosArchAMD64.
func (hc *HostConfig) CosArchitecture() CosArch {
if hc.cosArchitecture == CosArch("") {
return CosArchAMD64
}
return hc.cosArchitecture
}
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
// A BuildConfig describes how to run a builder.
type BuildConfig struct {
// Name is the unique name of the builder, in the form of
// "GOOS-GOARCH" or "GOOS-GOARCH-suffix". For example,
// "darwin-386", "linux-386-387", "linux-amd64-race". Some
// suffixes are well-known and carry special meaning, such as
// "-race".
Name string
// HostType is the required key into the Hosts map, describing
// the type of host this build will run on.
// For example, "host-linux-amd64-bullseye".
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
HostType string
// KnownIssues is a slice of non-zero go.dev/issue/nnn numbers for a
// builder that may fail due to a known issue, such as because it is a new
// builder still in development/testing, or because the feature
// or port that it's meant to test hasn't been added yet, etc.
//
// A non-zero value here means that failures on this builder should not
// be considered a serious regression and don't need investigation beyond
// what is already in scope of the listed issues.
KnownIssues []int
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
Notes string // notes for humans
// tryBot optionally specifies a policy func for whether trybots are enabled.
// nil means off. Even if tryBot returns true, BuildConfig.BuildsRepo must also
// return true. See the implementation of BuildConfig.BuildsRepoTryBot.
// The proj is "go", "net", etc. The branch is proj's branch.
// The goBranch is the same as branch for proj "go", else it's the go branch
// ("master, "release-branch.go1.12", etc).
tryBot func(proj, branch, goBranch string) bool
tryOnly bool // only used for trybots, and not regular builds
CompileOnly bool // if true, compile tests, but don't run them
FlakyNet bool // network tests are flaky (try anyway, but ignore some failures)
// buildsRepo optionally specifies whether this builder does
// builds (of any type) for the given repo ("go", "net", etc.)
// and its branch ("master", "release-branch.go1.12", "dev.link", etc.).
// goBranch is the branch of "go" to build against. If repo == "go",
// goBranch == branch.
//
// If nil, a default set of repos as reported by buildRepoByDefault
// is built. See buildsRepoAtAll method for details.
//
// To implement a minor change to the default policy, create a
// function that uses buildRepoByDefault. For example:
//
// buildsRepo: func(repo, branch, goBranch string) bool {
// b := buildRepoByDefault(repo)
// // ... modify b from the default value as needed ...
// return b
// }
//
buildsRepo func(repo, branch, goBranch string) bool
// RunBench enables benchmarking of the toolchain using x/benchmarks.
// This only applies when building x/benchmarks.
RunBench bool
dashboard, cmd/coordinator, maintner/maintnerd: add support for BuildConfig.MinimumGoVersion field The new BuildConfig.MinimumGoVersion field specifies the minimum Go version the builder is allowed to use. It's useful when some of the builders are too new, and do not support all of the supported Go releases (e.g., openbsd-amd64-64 and freebsd-amd64-12_0 currently require Go 1.11 and don't work on Go 1.10). It only needs to be set when a builder doesn't support all supported Go releases, since we don't typically test unsupported Go releases. To allow cmd/coordinator to use this field and filter out work it receives from maintner/maintnerd's GoFindTryWork RPC call, we add a GoVersion slice to apipb.GerritTryWorkItem, and populate it in maintapi.apiService.GoFindTryWork method. For trybots on the Go repo, the GoVersion field is determined from the branch name. For "release-branch.goX.Y" branches, it parses out the major-minor Go version from the branch name. For master and other branches, it assumes the latest Go release. For trybots on subrepos, we already have the Go version information available, so use it directly. Afterwards, all that's left is to modify newTrySet in cmd/coordinator to make use of BuildConfig.MinimumGoVersion and work.GoVersion to skip builders that are too new for the Go version that needs to be tested. Fixes golang/go#29265 Change-Id: I50b01830647e33e37e9eb8b89e0f2518812fa44f Reviewed-on: https://go-review.googlesource.com/c/155463 Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2019-01-12 00:44:01 +03:00
// MinimumGoVersion optionally specifies the minimum Go version
// this builder is allowed to use. It can be useful for skipping
// builders that are too new and no longer support some supported
// Go versions. It doesn't need to be set for builders that support
// all supported Go versions.
//
// Note: This field currently has effect on trybot runs only.
//
// TODO: unexport this and make buildsRepoAtAll return false on too-old
// of repos. The callers in coordinator will need updating.
dashboard, cmd/coordinator, maintner/maintnerd: add support for BuildConfig.MinimumGoVersion field The new BuildConfig.MinimumGoVersion field specifies the minimum Go version the builder is allowed to use. It's useful when some of the builders are too new, and do not support all of the supported Go releases (e.g., openbsd-amd64-64 and freebsd-amd64-12_0 currently require Go 1.11 and don't work on Go 1.10). It only needs to be set when a builder doesn't support all supported Go releases, since we don't typically test unsupported Go releases. To allow cmd/coordinator to use this field and filter out work it receives from maintner/maintnerd's GoFindTryWork RPC call, we add a GoVersion slice to apipb.GerritTryWorkItem, and populate it in maintapi.apiService.GoFindTryWork method. For trybots on the Go repo, the GoVersion field is determined from the branch name. For "release-branch.goX.Y" branches, it parses out the major-minor Go version from the branch name. For master and other branches, it assumes the latest Go release. For trybots on subrepos, we already have the Go version information available, so use it directly. Afterwards, all that's left is to modify newTrySet in cmd/coordinator to make use of BuildConfig.MinimumGoVersion and work.GoVersion to skip builders that are too new for the Go version that needs to be tested. Fixes golang/go#29265 Change-Id: I50b01830647e33e37e9eb8b89e0f2518812fa44f Reviewed-on: https://go-review.googlesource.com/c/155463 Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2019-01-12 00:44:01 +03:00
MinimumGoVersion types.MajorMinor
// SkipSnapshot, if true, means to not fetch a tarball
// snapshot of the world post-make.bash from the buildlet (and
// thus to not write it to Google Cloud Storage). This is
// incompatible with sharded tests, and should only be used
// for very slow builders or networks, unable to transfer
// the tarball in under ~5 minutes.
SkipSnapshot bool
// StopAfterMake causes the build to stop after the make
// script completes, returning its result as the result of the
// whole build. It does not run or compile any of the tests,
// nor does it write a snapshot of the world to cloud
// storage. This option is only supported for builders whose
// BuildConfig.SplitMakeRun returns true.
StopAfterMake bool
// needsGoProxy is whether this builder should have GOPROXY set.
// Currently this is only for the longtest builder, which needs
// to run cmd/go tests fetching from the network.
needsGoProxy bool
// privateGoProxy for builder has it's own Go proxy instead of
// proxy.golang.org, after setting this builder will respect
// GOPROXY environment value.
privateGoProxy bool
// InstallRacePackages controls which packages to "go install
// -race <pkgs>" after running make.bash (or equivalent). If
// the builder ends in "-race", the default if non-nil is just
// "std".
InstallRacePackages []string
// GoDeps is a list of of git sha1 commits that must be in the
// commit to be tested's history. If absent, this builder is
// not run for that commit.
GoDeps []string
// distTestAdjust optionally specifies a function that can
// adjust the cmd/dist test policy for this builder.
//
// The BuildConfig.ShouldRunDistTest method implements the
// default cmd/dist test policy, and then calls distTestAdjust
// to adjust that decision further, if distTestAdjust is not nil.
//
// The initial value of the run parameter is what the default
// policy said. The returned value from distTestAdjust is what
// BuildConfig.ShouldRunDistTest reports to the caller.
//
// For example:
//
// distTestAdjust: func(run bool, distTest string, isNormalTry bool) bool {
// // ... modify run from the initial value as needed ...
// return run
// }
//
distTestAdjust func(run bool, distTest string, isNormalTry bool) bool
cmd/coordinator, dashboard: remove some trybots, shard others wider I'm aiming to have trybot runs finish in under 5 minutes. This CL removes openbsd-386-gce58 and freebsd-386-gce101 from the trybot set. openbsd-386-gce58 is the slowest builder. It has an average speed of 722 seconds (and 95 percentile of 923 seconds) over the past week, and that's sharded over 4 machines. Too slow. It's not worth the resources to keep it as a trybot. It hasn't caught any interesting bugs. This builder will still run, but not as a pre-submit trybot. freebsd-386-gce101 is not slow, but we're removing it to shift its resources to shard other builders wider. The coordinator now supports varying the build sharding width based on whether a build is for a trybot or not. This CL defines separate numbers for each, sharding builds wider as needed for some trybots. freebsd-amd64-gce101 goes from 4 to 5 machines in try runs, and down to 3 when not in try runs. linux-amd64-race gets one more machine during try runs, and one fewer in regular runs. linux-arm goes from 7 machines always, to 3 or 8, depending on whether it's a try run. openbsd-amd64-58 goes from 4 to 3 or 6. windows-amd64-gce goes from 4 to 2 or 6. windows-amd64-race goes from 4 to 2 or 6. darwin-amd64-10_11 goes from 3 to 3 or 4. I'll see how these do over the next few days and readjust as needed. Also in this CL: fix the constants for the expected duration of make.bash, which impact when we schedule the creation of test sharding helper buildlets. We were creating them too early before, wasting resources. Change-Id: I38a9b24841e196f1eb668de058c49af8c1d1c64f Reviewed-on: https://go-review.googlesource.com/29116 Reviewed-by: Quentin Smith <quentin@golang.org>
2016-09-14 01:45:48 +03:00
// numTestHelpers is the number of _additional_ buildlets
// past the first one to help out with sharded tests.
// For TryBots and SlowBots, the numTryHelpers value is used,
// unless it's zero, in which case numTestHelpers is used.
cmd/coordinator, dashboard: remove some trybots, shard others wider I'm aiming to have trybot runs finish in under 5 minutes. This CL removes openbsd-386-gce58 and freebsd-386-gce101 from the trybot set. openbsd-386-gce58 is the slowest builder. It has an average speed of 722 seconds (and 95 percentile of 923 seconds) over the past week, and that's sharded over 4 machines. Too slow. It's not worth the resources to keep it as a trybot. It hasn't caught any interesting bugs. This builder will still run, but not as a pre-submit trybot. freebsd-386-gce101 is not slow, but we're removing it to shift its resources to shard other builders wider. The coordinator now supports varying the build sharding width based on whether a build is for a trybot or not. This CL defines separate numbers for each, sharding builds wider as needed for some trybots. freebsd-amd64-gce101 goes from 4 to 5 machines in try runs, and down to 3 when not in try runs. linux-amd64-race gets one more machine during try runs, and one fewer in regular runs. linux-arm goes from 7 machines always, to 3 or 8, depending on whether it's a try run. openbsd-amd64-58 goes from 4 to 3 or 6. windows-amd64-gce goes from 4 to 2 or 6. windows-amd64-race goes from 4 to 2 or 6. darwin-amd64-10_11 goes from 3 to 3 or 4. I'll see how these do over the next few days and readjust as needed. Also in this CL: fix the constants for the expected duration of make.bash, which impact when we schedule the creation of test sharding helper buildlets. We were creating them too early before, wasting resources. Change-Id: I38a9b24841e196f1eb668de058c49af8c1d1c64f Reviewed-on: https://go-review.googlesource.com/29116 Reviewed-by: Quentin Smith <quentin@golang.org>
2016-09-14 01:45:48 +03:00
numTestHelpers int
numTryTestHelpers int // For TryBots/SlowBots. If 0, numTestHelpers is used.
env []string // extra environment ("key=value") pairs
allScriptArgs []string
testHostConf *HostConfig // override HostConfig for testing, at least for now
// isRestricted marks if a builder should be restricted to a subset of users.
isRestricted bool
}
// Env returns the environment variables this builder should run with.
func (c *BuildConfig) Env() []string {
env := []string{"GO_BUILDER_NAME=" + c.Name}
if c.FlakyNet {
env = append(env, "GO_BUILDER_FLAKY_NET=1")
}
if c.IsLongTest() {
// Set a private hook in cmd/dist to run main Go repository tests
// without the default -short flag. See go.dev/issue/12508.
env = append(env, "GO_TEST_SHORT=0")
}
env = append(env, c.HostConfig().env...)
return append(env, c.env...)
}
// ModulesEnv returns the extra module-specific environment variables
// to append to this builder as a function of the repo being built
// ("go", "oauth2", "net", etc).
func (c *BuildConfig) ModulesEnv(repo string) (env []string) {
// EC2 and reverse builders should set the public module proxy
// address instead of the internal proxy.
if (c.HostConfig().isEC2 || c.IsReverse()) && repo != "go" && !c.PrivateGoProxy() {
env = append(env, "GOPROXY=https://proxy.golang.org")
}
switch repo {
case "go":
if !c.OutboundNetworkAllowed() {
env = append(env, "GOPROXY=off")
}
case "oauth2", "build", "perf", "website":
env = append(env, "GO111MODULE=on")
}
return env
}
func (c *BuildConfig) IsReverse() bool { return c.HostConfig().IsReverse }
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
func (c *BuildConfig) IsContainer() bool { return c.HostConfig().IsContainer() }
func (c *HostConfig) IsContainer() bool { return c.ContainerImage != "" }
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
func (c *BuildConfig) IsVM() bool { return c.HostConfig().IsVM() }
// IsVM reports whether the instance running the job is ultimately a VM. Hosts where
// a VM is used only to initiate a container are considered a container, not a VM.
// EC2 instances may be configured to run in containers that are running
// on custom AMIs.
func (c *HostConfig) IsVM() bool {
if c.isEC2 {
return c.VMImage != "" && c.ContainerImage == ""
}
return c.VMImage != ""
}
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
func (c *BuildConfig) GOOS() string { return c.Name[:strings.Index(c.Name, "-")] }
func (c *BuildConfig) GOARCH() string {
arch := c.Name[strings.Index(c.Name, "-")+1:]
i := strings.Index(arch, "-")
if i == -1 {
return arch
}
return arch[:i]
}
// MatchesSlowBotTerm reports whether some provided term from a
// TRY=... comment on a Run-TryBot+1 vote on Gerrit should match this
// build config.
func (c *BuildConfig) MatchesSlowBotTerm(term string) bool {
return term != "" && (term == c.Name || slowBotAliases[term] == c.Name)
}
// FilePathJoin is mostly like filepath.Join (without the cleaning) except
// it uses the path separator of c.GOOS instead of the host system's.
func (c *BuildConfig) FilePathJoin(x ...string) string {
if c.GOOS() == "windows" {
return strings.Join(x, "\\")
}
return strings.Join(x, "/")
}
func (c *BuildConfig) IsRestricted() bool {
return c.isRestricted
}
// DistTestsExecTimeout returns how long the coordinator should wait
// for a cmd/dist test execution to run the provided dist test names.
func (c *BuildConfig) DistTestsExecTimeout(distTests []string) time.Duration {
// TODO: consider using distTests? We never did before, but
// now we have the TestStats in the coordinator. Pass in a
// *buildstats.TestStats and use historical data times some
// fudge factor? For now just use the old 20 minute limit
// we've used since 2014, but scale it by the
// GO_TEST_TIMEOUT_SCALE for the super slow builders which
// struggle with, say, the cgo tests. (which should be broken
// up into separate dist tests or shards, like the test/ dir
// was)
d := 20 * time.Minute
d *= time.Duration(c.GoTestTimeoutScale())
return d
}
// GoTestTimeoutScale returns this builder's GO_TEST_TIMEOUT_SCALE value, or 1.
func (c *BuildConfig) GoTestTimeoutScale() int {
const pfx = "GO_TEST_TIMEOUT_SCALE="
for _, env := range [][]string{c.env, c.HostConfig().env} {
for _, kv := range env {
if strings.HasPrefix(kv, pfx) {
if n, err := strconv.Atoi(kv[len(pfx):]); err == nil && n > 0 {
return n
}
}
}
}
return 1
}
// HostConfig returns the host configuration of c.
func (c *BuildConfig) HostConfig() *HostConfig {
if c.testHostConf != nil {
return c.testHostConf
}
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
if c, ok := Hosts[c.HostType]; ok {
return c
}
panic(fmt.Sprintf("missing buildlet config for buildlet %q", c.Name))
}
// GoBootstrapURL returns the URL of a built Go 1.4+ tar.gz for the
// build configuration type c, or empty string if there isn't one.
func (c *BuildConfig) GoBootstrapURL(e *buildenv.Environment) string {
dashboard: update builders to Go 1.17.13 for bootstrap Compute GoBootstrapURL for all hosts. This removes another obviously computable field, assuming the bootstrap toolchains have all been uploaded with consistent names (as genbootstrap -upload now does). It also makes all builders work the same: they all start by downloading the bootstrap toolchain, instead of having some download and others use a toolchain installed into the boot image. The builders with pre-installed toolchains will continue to have those toolchains, but now nothing will use them, because the coordinator or gomote will see the GoBootstrapURL result and push an updated toolchain to $workdir/go1.4. As part of all this simplification, switch to Go 1.17.13 as the bootstrap toolchain, for golang/go#44505. (And also because we have no Go 1.4 for some of these platforms.) Updating the bootstrap toolchain again in a couple years will now be easy: run genbootstrap -upload '*' and then change GoBootstrap here. For golang/go#44505. (Won't be fixed until coordinator is redeployed.) Succeeded running all.bash/bat/rc using updated debugnewvm: android-386-emu android-amd64-emu dragonfly-amd64-622 freebsd-386-12_3 freebsd-386-13_0 freebsd-amd64-13_0 freebsd-amd64-race js-wasm linux-386 linux-386-buster linux-386-clang linux-386-longtest linux-386-sid linux-386-softfloat linux-386-stretch linux-amd64 linux-amd64-alpine linux-amd64-boringcrypto linux-amd64-bullseye linux-amd64-buster linux-amd64-clang linux-amd64-fedora linux-amd64-goamd64v3 linux-amd64-longtest linux-amd64-nocgo linux-amd64-noopt linux-amd64-nounified linux-amd64-perf linux-amd64-race linux-amd64-racecompile linux-amd64-sid linux-amd64-ssacheck linux-amd64-staticlockranking linux-amd64-stretch linux-amd64-unified linux-amd64-vmx netbsd-386-9_0 netbsd-amd64-9_0 openbsd-386-68 openbsd-386-70 openbsd-amd64-68 openbsd-amd64-70 windows-386-2008 windows-386-2012 windows-amd64-2008 windows-amd64-2012 windows-amd64-2016 windows-amd64-longtest windows-amd64-newcc-race windows-amd64-race Succeeded running make.bash using updated debugnewvm: linux-s390x-crosscompile Succeeded running make.bash using updated gomote (real time for put14 shown): aix-ppc64 (5.2s) darwin-amd64-10_14 (3.4s) darwin-amd64-10_15 (3.5s) darwin-amd64-11_0 (4.5s) darwin-amd64-12_0 (4.5s) darwin-amd64-nocgo (4.4s) darwin-amd64-race (4.2s) darwin-arm64-11 (1.7s) darwin-arm64-12 (1.8s) freebsd-arm64-dmgk (4.0s) freebsd-arm-paulzhol (35.1s) illumos-amd64 (5.9s) linux-amd64-wsl (126.7s) linux-arm64-aws (2.8s) linux-arm-aws (3.8s) linux-mips64le-rtrk (20.4s) linux-mips64-rtrk (16.0s) linux-mipsle-rtrk (16.3s) linux-mips-rtrk (16.3s) linux-ppc64-buildlet (4.7s) linux-ppc64le-buildlet (3.6s) linux-ppc64le-power9osu (5.6s) linux-riscv64-unmatched (38.2s) linux-s390x-ibm (4.4s) netbsd-arm64-bsiegert (35.0s) openbsd-arm64-jsing (64.9s) openbsd-mips64-jsing (154.2s) plan9-386-0intro (8.7s) plan9-arm (18.2s) windows-arm64-10 (7.1s) windows-arm64-11 (4.1s) Put14 failed using gomote: android-arm64-corellium (403 Forbidden in fetch) android-arm-corellium (403 Forbidden in fetch) ios-arm64-corellium (403 Forbidden in fetch) - fails build, kept old GOROOT_BOOTSTRAP linux-loong64-3a5000 (proxyconnect tcp: dial tcp 117.50.175.110:7890: i/o timeout) - falls back to preinstalled Go 1.18, kept old GOROOT_BOOTSTRAP Builds fine, failing tests but not on dashboard anyway: freebsd-386-11_4 freebsd-amd64-11_4 linux-amd64-androidemu plan9-386 Failed in same way as current builders: openbsd-386-70-n2d windows-386-2008-newcc windows-386-2012-newcc windows-amd64-2008-newcc windows-amd64-2012-newcc windows-amd64-2016-newcc windows-amd64-longtest-newcc Can't connect with gomote, so untested: linux-riscv64-jsing (but linux-riscv64-unmatched works) netbsd-arm-bsiegert openbsd-arm-jsing plan9-amd64-0intro solaris-amd64-oraclerel (but illumos-amd64 works) Change-Id: If4354a8681dbac28e1d34897b6356b881592c241 Reviewed-on: https://go-review.googlesource.com/c/build/+/419993 TryBot-Result: Gopher Robot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Jenny Rakoczy <jenny@golang.org> Run-TryBot: Russ Cox <rsc@golang.org>
2022-07-29 03:55:40 +03:00
hc := c.HostConfig()
if hc.GoBootstrap == "none" {
return ""
}
return "https://storage.googleapis.com/" + e.BuildletBucket +
"/gobootstrap-" + hc.HostArch + "-" + hc.GoBootstrap + ".tar.gz"
}
// BuildletBinaryURL returns the public URL of this builder's buildlet.
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
func (c *HostConfig) BuildletBinaryURL(e *buildenv.Environment) string {
return "https://storage.googleapis.com/" + e.BuildletBucket + "/buildlet." + c.HostArch
}
// IsRace reports whether this is a race builder.
// A race builder runs tests without the -short flag.
//
// A builder is considered to be a race builder
// if and only if it one of the components of the builder
// name is "race" (components are delimited by "-").
func (c *BuildConfig) IsRace() bool {
for _, s := range strings.Split(c.Name, "-") {
if s == "race" {
return true
}
}
return false
}
// IsLongTest reports whether this is a longtest builder.
// A longtest builder runs tests without the -short flag.
//
// A builder is considered to be a longtest builder
// if and only if it one of the components of the builder
// name is "longtest" (components are delimited by "-").
func (c *BuildConfig) IsLongTest() bool {
for _, s := range strings.Split(c.Name, "-") {
if s == "longtest" {
return true
}
}
return false
}
// OutboundNetworkAllowed reports whether this builder should be
// allowed to make outbound network requests. This is only enforced
// on some builders. (Currently most Linux ones)
func (c *BuildConfig) OutboundNetworkAllowed() bool {
return c.IsLongTest()
}
func (c *BuildConfig) GoInstallRacePackages() []string {
if c.InstallRacePackages != nil {
return append([]string(nil), c.InstallRacePackages...)
}
if c.IsRace() {
return []string{"std"}
}
return nil
}
// AllScript returns the relative path to the operating system's script to
// do the build and run its standard set of tests.
// Example values are "src/all.bash", "src/all.bat", "src/all.rc".
func (c *BuildConfig) AllScript() string {
if c.Name == "" {
panic("bogus BuildConfig")
}
if c.IsRace() {
if strings.HasPrefix(c.Name, "windows-") {
return "src/race.bat"
}
return "src/race.bash"
}
if strings.HasPrefix(c.Name, "windows-") {
return "src/all.bat"
}
if strings.HasPrefix(c.Name, "plan9-") {
return "src/all.rc"
}
if strings.HasPrefix(c.Name, "misc-compile") {
return "src/buildall.bash"
}
return "src/all.bash"
}
// SplitMakeRun reports whether the coordinator should first compile
// (using c.MakeScript), then snapshot, then run the tests (ideally
// sharded) using cmd/dist test.
// Eventually this function should always return true (and then be deleted)
// but for now we've only set up the scripts and verified that the main
// configurations work.
func (c *BuildConfig) SplitMakeRun() bool {
switch c.AllScript() {
case "src/all.bash", "src/all.bat",
"src/race.bash", "src/race.bat",
"src/all.rc":
// These we've verified to work.
return true
}
// TODO(bradfitz): buildall.bash should really just be N small container
// jobs instead of a "buildall.bash". Then we can delete this whole method.
return false
}
func (c *BuildConfig) IsTryOnly() bool { return c.tryOnly }
func (c *BuildConfig) NeedsGoProxy() bool { return c.needsGoProxy }
// PrivateGoProxy for builder has it's own Go proxy instead of proxy.golang.org
func (c *BuildConfig) PrivateGoProxy() bool { return c.privateGoProxy }
// BuildsRepoPostSubmit reports whether the build configuration type c
// should build the given repo ("go", "net", etc) and branch
// ("master", "release-branch.go1.12") as a post-submit build
// that shows up on https://build.golang.org/.
func (c *BuildConfig) BuildsRepoPostSubmit(repo, branch, goBranch string) bool {
if c.tryOnly {
return false
}
return c.buildsRepoAtAll(repo, branch, goBranch)
}
// BuildsRepoTryBot reports whether the build configuration type c
// should build the given repo ("go", "net", etc) and branch
// ("master", "release-branch.go1.12") as a trybot.
func (c *BuildConfig) BuildsRepoTryBot(repo, branch, goBranch string) bool {
return c.tryBot != nil && c.tryBot(repo, branch, goBranch) && c.buildsRepoAtAll(repo, branch, goBranch)
}
// ShouldRunDistTest reports whether the named cmd/dist test should be
// run for this build config. The isNormalTry parameter is whether this
// is for a normal TryBot (non-SlowBot) run.
//
// In general, this returns true. When in normal trybot mode,
// some slow portable tests are only run on the fastest builder.
//
// Individual builders can adjust this policy to fit their needs.
func (c *BuildConfig) ShouldRunDistTest(distTest string, isNormalTry bool) bool {
run := true
// This section implements the default cmd/dist test policy.
// Any changes here will affect test coverage on all builders.
if isNormalTry {
slowPortableTest := distTest == "api" // Whether a test is slow and has the same behavior everywhere.
fastestBuilder := c.Name == "linux-amd64"
if slowPortableTest && !fastestBuilder {
// Don't run the test on this builder.
run = false
}
}
// Let individual builders adjust the cmd/dist test policy.
if c.distTestAdjust != nil {
run = c.distTestAdjust(run, distTest, isNormalTry)
}
return run
}
// buildsRepoAtAll reports whether we should do builds of the provided
// repo ("go", "sys", "net", etc). This applies to both post-submit
// and trybot builds. Use BuildsRepoPostSubmit for only post-submit
// or BuildsRepoTryBot for trybots.
//
// The branch is the branch of repo ("master",
// "release-branch.go1.12", etc); it is required. The goBranch is the
// branch of Go itself. It's required if repo != "go". When repo ==
// "go", the goBranch defaults to the value of branch.
func (c *BuildConfig) buildsRepoAtAll(repo, branch, goBranch string) bool {
if goBranch == "" {
if repo == "go" {
goBranch = branch
} else {
panic("missing goBranch")
}
}
if branch == "" {
panic("missing branch")
}
if repo == "" {
panic("missing repo")
}
// Don't build old branches.
const minGo1x = 11
for _, b := range []string{branch, goBranch} {
if bmaj, bmin, ok := version.ParseReleaseBranch(b); ok {
if bmaj != 1 || bmin < minGo1x {
return false
}
bmm := types.MajorMinor{Major: bmaj, Minor: bmin}
if bmm.Less(c.MinimumGoVersion) {
return false
}
if repo == "exp" {
// Don't test exp against release branches; it's experimental.
return false
}
}
}
// Build dev.boringcrypto branches only on linux/amd64 and windows/386 (see go.dev/issue/26791).
if repo == "go" && (branch == "dev.boringcrypto" || strings.HasPrefix(branch, "dev.boringcrypto.")) {
if c.Name != "linux-amd64" && !strings.HasPrefix(c.Name, "windows-386") {
return false
}
}
if repo != "go" && !c.SplitMakeRun() {
return false
}
if p := c.buildsRepo; p != nil {
return p(repo, branch, goBranch)
}
return buildRepoByDefault(repo)
}
// buildRepoByDefault reports whether builders should do builds
// for the given repo ("go", "net", etc.) by default.
//
// It's used directly by BuildConfig.buildsRepoAtAll method for
// builders with a nil BuildConfig.buildsRepo value.
// It's also used by many builders that provide a custom build
// repo policy (a non-nil BuildConfig.buildsRepo value) as part
// of making the decision of whether to build a given repo.
//
// As a result, it effectively implements the default build repo policy.
// Any changes here will affect repo coverage of many builders.
func buildRepoByDefault(repo string) bool {
switch repo {
case "go":
// Build the main Go repository by default.
return true
case "mobile", "exp", "build", "vulndb":
// Don't build x/mobile, x/exp, x/build, x/vulndb by default.
//
// Builders need to explicitly opt-in to build these repos.
return false
default:
// Build all other golang.org/x repositories by default.
return true
}
}
var (
defaultPlusExp = defaultPlus("exp")
defaultPlusExpBuild = defaultPlus("exp", "build")
defaultPlusExpBuildVulnDB = defaultPlus("exp", "build", "vulndb")
)
// defaultPlus returns a buildsRepo policy function that returns true for all
// all the repos listed, plus the default repos.
func defaultPlus(repos ...string) func(repo, branch, goBranch string) bool {
return func(repo, _, _ string) bool {
for _, r := range repos {
if r == repo {
return true
}
}
return buildRepoByDefault(repo)
}
}
// AllScriptArgs returns the set of arguments that should be passed to the
// all.bash-equivalent script. Usually empty.
func (c *BuildConfig) AllScriptArgs() []string {
return append([]string(nil), c.allScriptArgs...)
}
// MakeScript returns the relative path to the operating system's script to
// do the build.
// Example values are "src/make.bash", "src/make.bat", "src/make.rc".
func (c *BuildConfig) MakeScript() string {
if strings.HasPrefix(c.Name, "windows-") {
return "src/make.bat"
}
if strings.HasPrefix(c.Name, "plan9-") {
return "src/make.rc"
}
return "src/make.bash"
}
// MakeScriptArgs returns the set of arguments that should be passed to the
// make.bash-equivalent script. Usually empty.
func (c *BuildConfig) MakeScriptArgs() []string {
return c.AllScriptArgs()
}
// GorootFinal returns the default install location for
// releases for this platform.
func (c *BuildConfig) GorootFinal() string {
if strings.HasPrefix(c.Name, "windows-") {
return "c:\\go"
}
return "/usr/local/go"
}
// MachineType returns the AWS or GCE machine type to use for this builder.
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
func (c *HostConfig) MachineType() string {
if c.IsEC2() {
return c.machineType
}
typ := c.machineType
if typ == "" {
if c.NestedVirt || c.MinCPUPlatform != "" {
// e2 instances do not support nested virtualization, but n2 instances do.
// Same for MinCPUPlatform: https://cloud.google.com/compute/docs/instances/specify-min-cpu-platform#limitations
typ = "n2"
} else {
typ = "e2"
}
}
if strings.Contains(typ, "-") { // full type like "n2-standard-8"
return typ
}
if c.IsContainer() {
// Set a higher default machine size for containers,
// so their /workdir tmpfs can be larger. The COS
// image has no swap, so we want to make sure the
// /workdir fits completely in memory.
return typ + "-standard-16" // 16 vCPUs, 64 GB mem
}
return typ + "-standard-8" // 8 vCPUs, 32 GB mem
}
// IsEC2 returns true if the machine type is an EC2 arm64 type.
func (c *HostConfig) IsEC2() bool {
return c.isEC2
}
// PoolName returns a short summary of the builder's host type for the
// https://farmer.golang.org/builders page.
func (c *HostConfig) PoolName() string {
switch {
case c.IsReverse:
return "Reverse (dedicated machine/VM)"
case c.IsEC2():
return "EC2 VM Container"
case c.IsVM():
return "GCE VM"
case c.IsContainer():
return "Container"
}
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
panic("unknown builder type")
}
// ContainerVMImage returns the base VM name (not the fully qualified
// URL resource name of the VM) that starts the konlet program that
// pulls & runs a container.
// The empty string means that no particular VM image is required
// and the caller can run this container in any host.
//
// This method is only applicable when c.IsContainer() is true.
func (c *HostConfig) ContainerVMImage() string {
if c.KonletVMImage != "" {
return c.KonletVMImage
}
if c.NestedVirt {
return "debian-bullseye-vmx"
}
if c.isEC2 && c.ContainerImage != "" {
return fmt.Sprintf("gcr.io/%s/%s", buildenv.Production.ProjectName, c.ContainerImage)
}
return ""
}
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
// IsHermetic reports whether this host config gets a fresh
// environment (including /usr, /var, etc) for each execution. This is
// true for VMs, GKE, and reverse buildlets running their containers
// running in Docker, but false on some reverse buildlets.
func (c *HostConfig) IsHermetic() bool {
switch {
case c.IsReverse:
return c.HermeticReverse
case c.IsEC2():
return true
case c.IsVM():
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
return true
case c.IsContainer():
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
return true
}
panic("unknown builder type")
}
// IsGoogle reports whether this host is operated by Google,
// implying that we trust it to be secure and available.
func (c *HostConfig) IsGoogle() bool {
if c.IsContainer() || c.IsVM() {
return true
}
if c.IsReverse && c.GoogleReverse {
return true
}
return false
}
// NumTestHelpers reports how many additional buildlets
// past the first one to help out with sharded tests.
//
// isTry specifies whether it's for a pre-submit test
// run (a TryBot or SlowBot) where speed matters more.
cmd/coordinator, dashboard: remove some trybots, shard others wider I'm aiming to have trybot runs finish in under 5 minutes. This CL removes openbsd-386-gce58 and freebsd-386-gce101 from the trybot set. openbsd-386-gce58 is the slowest builder. It has an average speed of 722 seconds (and 95 percentile of 923 seconds) over the past week, and that's sharded over 4 machines. Too slow. It's not worth the resources to keep it as a trybot. It hasn't caught any interesting bugs. This builder will still run, but not as a pre-submit trybot. freebsd-386-gce101 is not slow, but we're removing it to shift its resources to shard other builders wider. The coordinator now supports varying the build sharding width based on whether a build is for a trybot or not. This CL defines separate numbers for each, sharding builds wider as needed for some trybots. freebsd-amd64-gce101 goes from 4 to 5 machines in try runs, and down to 3 when not in try runs. linux-amd64-race gets one more machine during try runs, and one fewer in regular runs. linux-arm goes from 7 machines always, to 3 or 8, depending on whether it's a try run. openbsd-amd64-58 goes from 4 to 3 or 6. windows-amd64-gce goes from 4 to 2 or 6. windows-amd64-race goes from 4 to 2 or 6. darwin-amd64-10_11 goes from 3 to 3 or 4. I'll see how these do over the next few days and readjust as needed. Also in this CL: fix the constants for the expected duration of make.bash, which impact when we schedule the creation of test sharding helper buildlets. We were creating them too early before, wasting resources. Change-Id: I38a9b24841e196f1eb668de058c49af8c1d1c64f Reviewed-on: https://go-review.googlesource.com/29116 Reviewed-by: Quentin Smith <quentin@golang.org>
2016-09-14 01:45:48 +03:00
func (c *BuildConfig) NumTestHelpers(isTry bool) int {
if isTry && c.numTryTestHelpers != 0 {
return c.numTryTestHelpers
}
return c.numTestHelpers
}
// defaultTrySet returns a trybot policy function that reports whether
// a project should use trybots. All the default projects are included,
// plus any given in extraProj.
func defaultTrySet(extraProj ...string) func(proj, branch, goBranch string) bool {
return func(proj, branch, goBranch string) bool {
if proj == "go" {
return true
}
for _, p := range extraProj {
if proj == p {
return true
}
}
switch proj {
case "grpc-review":
return false
}
return true
}
}
// explicitTrySet returns a trybot policy function that reports
// whether a project should use trybots. Only the provided projects in
// projs are enabled.
func explicitTrySet(projs ...string) func(proj, branch, goBranch string) bool {
return func(proj, branch, goBranch string) bool {
for _, p := range projs {
if proj == p {
return true
}
}
return false
}
}
func init() {
addBuilder(BuildConfig{
Name: "freebsd-amd64-12_3",
HostType: "host-freebsd-amd64-12_3",
tryBot: defaultTrySet("sys"),
distTestAdjust: fasterTrybots, // If changing this policy, update TestShouldRunDistTest accordingly.
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "freebsd-386-12_3",
HostType: "host-freebsd-amd64-12_3",
env: []string{"GOARCH=386", "GOHOSTARCH=386"},
distTestAdjust: fasterTrybots,
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "freebsd-amd64-race",
HostType: "host-freebsd-amd64-13_0",
})
addBuilder(BuildConfig{
Name: "freebsd-amd64-13_0",
HostType: "host-freebsd-amd64-13_0",
tryBot: explicitTrySet("sys"),
distTestAdjust: fasterTrybots, // If changing this policy, update TestShouldRunDistTest accordingly.
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "freebsd-386-13_0",
HostType: "host-freebsd-amd64-13_0",
tryBot: explicitTrySet("sys"),
env: []string{"GOARCH=386", "GOHOSTARCH=386"},
distTestAdjust: fasterTrybots,
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "linux-386",
HostType: "host-linux-amd64-bullseye",
distTestAdjust: fasterTrybots,
tryBot: defaultTrySet(),
Notes: "Debian stable (currently Debian bullseye).",
env: []string{
"GOARCH=386",
"GOHOSTARCH=386",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
numTestHelpers: 1,
numTryTestHelpers: 3,
})
addBuilder(BuildConfig{
Name: "linux-386-softfloat",
Notes: "GO386=softfloat",
buildsRepo: func(repo, branch, goBranch string) bool {
return repo == "go" || repo == "crypto"
},
HostType: "host-linux-amd64-bullseye",
env: []string{"GOARCH=386", "GOHOSTARCH=386", "GO386=softfloat"},
})
addBuilder(BuildConfig{
Name: "linux-amd64",
HostType: "host-linux-amd64-bullseye",
tryBot: defaultTrySet(),
buildsRepo: defaultPlusExpBuildVulnDB,
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
numTestHelpers: 1,
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "linux-amd64-boringcrypto",
HostType: "host-linux-amd64-bullseye",
Notes: "GOEXPERIMENT=boringcrypto",
tryBot: defaultTrySet(),
env: []string{
"GOEXPERIMENT=boringcrypto",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
numTestHelpers: 1,
numTryTestHelpers: 4,
buildsRepo: func(repo, branch, goBranch string) bool {
return atLeastGo1(goBranch, 19) && buildRepoByDefault(repo)
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-vmx",
HostType: "host-linux-amd64-bullseye-vmx",
buildsRepo: disabledBuilder,
})
addBuilder(BuildConfig{
Name: "linux-amd64-alpine",
HostType: "host-linux-amd64-alpine",
buildsRepo: func(repo, branch, goBranch string) bool {
return atLeastGo1(goBranch, 20) && buildRepoByDefault(repo)
},
})
dashboard: include at most 3 ports per misc-compile TryBot Between Go 1.15, 1.16 and tip, the following 29 ports don't have a real TryBot and instead rely on misc-compile TryBots for their pre-submit coverage: • aix/ppc64 • darwin/amd64 • darwin/arm64 • dragonfly/amd64 • freebsd/386 • freebsd/arm • freebsd/arm64 • illumos/amd64 • linux/mips • linux/mips64 • linux/mipsle • linux/mips64le • linux/ppc64 • linux/ppc64le • linux/riscv64 • linux/s390x • netbsd/386 • netbsd/amd64 • netbsd/arm • netbsd/arm64 • openbsd/386 • openbsd/arm • openbsd/arm64 • openbsd/mips64 • plan9/386 • plan9/amd64 • plan9/arm • solaris/amd64 • windows/arm The previous approach for misc-compile target selection was to break them up primarily by GOOS value. However, as new architectures were added over time, some misc-compile TryBots got to a point where they were testing upwards of 5 ports (for example, misc-compile-openbsd was testing 386, amd64, arm, arm64, and mips64 architectures). Since each port is tested sequentially, allocating too many to one misc-compile TryBot can cause it to become the bottleneck of an entire TryBot run, exceeding the 10 minute completion time goal. Arrange it so misc-compile TryBot target selection is done explicitly in x/build, and pick 3 as max number of targets per TryBot for now. Based on recent timing observations, that should strike a decent balance between resource use (spinning up a builder) vs chance of a misc-compile TryBot becoming a bottleneck. It will also give us an opportunity to compare timing of 1, 2 and 3 targets per misc-compile in the future. (When we start tracking timing for TryBot completion time holistically, we'll be in a better position to refine this strategy further.) Making misc-compile target selection explicit in x/build also enables removing unnecessary duplicate misc-compile coverage from ports that already have a real TryBot (for example, openbsd/amd64 was previously tested via both the openbsd-amd64-68 TryBot and misc-compile-openbsd). This shouldn't be needed, so it's no longer done. For golang/go#17104. Fixes golang/go#32632. Change-Id: Iac918377b91af3e48780b38ffdf3153e213eeba2 Reviewed-on: https://go-review.googlesource.com/c/build/+/313210 Trust: Dmitri Shuralyov <dmitshur@golang.org> Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
2021-04-23 23:56:37 +03:00
// addMiscCompileGo1 adds a misc-compile TryBot that
// runs buildall.bash on the specified target(s), up to 3 max.
// The targets are matched against the "go tool dist list" name,
// but with hyphens instead of forward slashes ("linux-amd64", etc).
// If min is non-zero, it specifies the minimum Go 1.x version.
dashboard: include at most 3 ports per misc-compile TryBot Between Go 1.15, 1.16 and tip, the following 29 ports don't have a real TryBot and instead rely on misc-compile TryBots for their pre-submit coverage: • aix/ppc64 • darwin/amd64 • darwin/arm64 • dragonfly/amd64 • freebsd/386 • freebsd/arm • freebsd/arm64 • illumos/amd64 • linux/mips • linux/mips64 • linux/mipsle • linux/mips64le • linux/ppc64 • linux/ppc64le • linux/riscv64 • linux/s390x • netbsd/386 • netbsd/amd64 • netbsd/arm • netbsd/arm64 • openbsd/386 • openbsd/arm • openbsd/arm64 • openbsd/mips64 • plan9/386 • plan9/amd64 • plan9/arm • solaris/amd64 • windows/arm The previous approach for misc-compile target selection was to break them up primarily by GOOS value. However, as new architectures were added over time, some misc-compile TryBots got to a point where they were testing upwards of 5 ports (for example, misc-compile-openbsd was testing 386, amd64, arm, arm64, and mips64 architectures). Since each port is tested sequentially, allocating too many to one misc-compile TryBot can cause it to become the bottleneck of an entire TryBot run, exceeding the 10 minute completion time goal. Arrange it so misc-compile TryBot target selection is done explicitly in x/build, and pick 3 as max number of targets per TryBot for now. Based on recent timing observations, that should strike a decent balance between resource use (spinning up a builder) vs chance of a misc-compile TryBot becoming a bottleneck. It will also give us an opportunity to compare timing of 1, 2 and 3 targets per misc-compile in the future. (When we start tracking timing for TryBot completion time holistically, we'll be in a better position to refine this strategy further.) Making misc-compile target selection explicit in x/build also enables removing unnecessary duplicate misc-compile coverage from ports that already have a real TryBot (for example, openbsd/amd64 was previously tested via both the openbsd-amd64-68 TryBot and misc-compile-openbsd). This shouldn't be needed, so it's no longer done. For golang/go#17104. Fixes golang/go#32632. Change-Id: Iac918377b91af3e48780b38ffdf3153e213eeba2 Reviewed-on: https://go-review.googlesource.com/c/build/+/313210 Trust: Dmitri Shuralyov <dmitshur@golang.org> Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
2021-04-23 23:56:37 +03:00
addMiscCompileGo1 := func(min int, suffix string, targets ...string) {
if len(targets) > 3 {
// This limit will do until we have better visibility
// into holistic TryBot completion times via metrics.
panic("at most 3 targets may be specified to avoid making TryBots slow; see issues 32632 and 17104")
}
var v types.MajorMinor
var alsoNote string
if min != 0 {
v = types.MajorMinor{Major: 1, Minor: min}
alsoNote = fmt.Sprintf(" Applies to Go 1.%d and newer.", min)
}
addBuilder(BuildConfig{
Name: "misc-compile" + suffix,
HostType: "host-linux-amd64-bullseye",
tryBot: defaultTrySet(),
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
tryOnly: true,
MinimumGoVersion: v,
CompileOnly: true,
dashboard: include at most 3 ports per misc-compile TryBot Between Go 1.15, 1.16 and tip, the following 29 ports don't have a real TryBot and instead rely on misc-compile TryBots for their pre-submit coverage: • aix/ppc64 • darwin/amd64 • darwin/arm64 • dragonfly/amd64 • freebsd/386 • freebsd/arm • freebsd/arm64 • illumos/amd64 • linux/mips • linux/mips64 • linux/mipsle • linux/mips64le • linux/ppc64 • linux/ppc64le • linux/riscv64 • linux/s390x • netbsd/386 • netbsd/amd64 • netbsd/arm • netbsd/arm64 • openbsd/386 • openbsd/arm • openbsd/arm64 • openbsd/mips64 • plan9/386 • plan9/amd64 • plan9/arm • solaris/amd64 • windows/arm The previous approach for misc-compile target selection was to break them up primarily by GOOS value. However, as new architectures were added over time, some misc-compile TryBots got to a point where they were testing upwards of 5 ports (for example, misc-compile-openbsd was testing 386, amd64, arm, arm64, and mips64 architectures). Since each port is tested sequentially, allocating too many to one misc-compile TryBot can cause it to become the bottleneck of an entire TryBot run, exceeding the 10 minute completion time goal. Arrange it so misc-compile TryBot target selection is done explicitly in x/build, and pick 3 as max number of targets per TryBot for now. Based on recent timing observations, that should strike a decent balance between resource use (spinning up a builder) vs chance of a misc-compile TryBot becoming a bottleneck. It will also give us an opportunity to compare timing of 1, 2 and 3 targets per misc-compile in the future. (When we start tracking timing for TryBot completion time holistically, we'll be in a better position to refine this strategy further.) Making misc-compile target selection explicit in x/build also enables removing unnecessary duplicate misc-compile coverage from ports that already have a real TryBot (for example, openbsd/amd64 was previously tested via both the openbsd-amd64-68 TryBot and misc-compile-openbsd). This shouldn't be needed, so it's no longer done. For golang/go#17104. Fixes golang/go#32632. Change-Id: Iac918377b91af3e48780b38ffdf3153e213eeba2 Reviewed-on: https://go-review.googlesource.com/c/build/+/313210 Trust: Dmitri Shuralyov <dmitshur@golang.org> Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
2021-04-23 23:56:37 +03:00
Notes: "Runs buildall.bash to cross-compile & vet std+cmd packages for " + strings.Join(targets, " & ") + ", but doesn't run any tests." + alsoNote,
allScriptArgs: []string{
// Filtering pattern to buildall.bash:
dashboard: include at most 3 ports per misc-compile TryBot Between Go 1.15, 1.16 and tip, the following 29 ports don't have a real TryBot and instead rely on misc-compile TryBots for their pre-submit coverage: • aix/ppc64 • darwin/amd64 • darwin/arm64 • dragonfly/amd64 • freebsd/386 • freebsd/arm • freebsd/arm64 • illumos/amd64 • linux/mips • linux/mips64 • linux/mipsle • linux/mips64le • linux/ppc64 • linux/ppc64le • linux/riscv64 • linux/s390x • netbsd/386 • netbsd/amd64 • netbsd/arm • netbsd/arm64 • openbsd/386 • openbsd/arm • openbsd/arm64 • openbsd/mips64 • plan9/386 • plan9/amd64 • plan9/arm • solaris/amd64 • windows/arm The previous approach for misc-compile target selection was to break them up primarily by GOOS value. However, as new architectures were added over time, some misc-compile TryBots got to a point where they were testing upwards of 5 ports (for example, misc-compile-openbsd was testing 386, amd64, arm, arm64, and mips64 architectures). Since each port is tested sequentially, allocating too many to one misc-compile TryBot can cause it to become the bottleneck of an entire TryBot run, exceeding the 10 minute completion time goal. Arrange it so misc-compile TryBot target selection is done explicitly in x/build, and pick 3 as max number of targets per TryBot for now. Based on recent timing observations, that should strike a decent balance between resource use (spinning up a builder) vs chance of a misc-compile TryBot becoming a bottleneck. It will also give us an opportunity to compare timing of 1, 2 and 3 targets per misc-compile in the future. (When we start tracking timing for TryBot completion time holistically, we'll be in a better position to refine this strategy further.) Making misc-compile target selection explicit in x/build also enables removing unnecessary duplicate misc-compile coverage from ports that already have a real TryBot (for example, openbsd/amd64 was previously tested via both the openbsd-amd64-68 TryBot and misc-compile-openbsd). This shouldn't be needed, so it's no longer done. For golang/go#17104. Fixes golang/go#32632. Change-Id: Iac918377b91af3e48780b38ffdf3153e213eeba2 Reviewed-on: https://go-review.googlesource.com/c/build/+/313210 Trust: Dmitri Shuralyov <dmitshur@golang.org> Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
2021-04-23 23:56:37 +03:00
"^(" + strings.Join(targets, "|") + ")$",
},
})
}
// addMiscCompile adds a misc-compile TryBot
// for all supported Go versions.
dashboard: include at most 3 ports per misc-compile TryBot Between Go 1.15, 1.16 and tip, the following 29 ports don't have a real TryBot and instead rely on misc-compile TryBots for their pre-submit coverage: • aix/ppc64 • darwin/amd64 • darwin/arm64 • dragonfly/amd64 • freebsd/386 • freebsd/arm • freebsd/arm64 • illumos/amd64 • linux/mips • linux/mips64 • linux/mipsle • linux/mips64le • linux/ppc64 • linux/ppc64le • linux/riscv64 • linux/s390x • netbsd/386 • netbsd/amd64 • netbsd/arm • netbsd/arm64 • openbsd/386 • openbsd/arm • openbsd/arm64 • openbsd/mips64 • plan9/386 • plan9/amd64 • plan9/arm • solaris/amd64 • windows/arm The previous approach for misc-compile target selection was to break them up primarily by GOOS value. However, as new architectures were added over time, some misc-compile TryBots got to a point where they were testing upwards of 5 ports (for example, misc-compile-openbsd was testing 386, amd64, arm, arm64, and mips64 architectures). Since each port is tested sequentially, allocating too many to one misc-compile TryBot can cause it to become the bottleneck of an entire TryBot run, exceeding the 10 minute completion time goal. Arrange it so misc-compile TryBot target selection is done explicitly in x/build, and pick 3 as max number of targets per TryBot for now. Based on recent timing observations, that should strike a decent balance between resource use (spinning up a builder) vs chance of a misc-compile TryBot becoming a bottleneck. It will also give us an opportunity to compare timing of 1, 2 and 3 targets per misc-compile in the future. (When we start tracking timing for TryBot completion time holistically, we'll be in a better position to refine this strategy further.) Making misc-compile target selection explicit in x/build also enables removing unnecessary duplicate misc-compile coverage from ports that already have a real TryBot (for example, openbsd/amd64 was previously tested via both the openbsd-amd64-68 TryBot and misc-compile-openbsd). This shouldn't be needed, so it's no longer done. For golang/go#17104. Fixes golang/go#32632. Change-Id: Iac918377b91af3e48780b38ffdf3153e213eeba2 Reviewed-on: https://go-review.googlesource.com/c/build/+/313210 Trust: Dmitri Shuralyov <dmitshur@golang.org> Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
2021-04-23 23:56:37 +03:00
addMiscCompile := func(suffix string, targets ...string) { addMiscCompileGo1(0, suffix, targets...) }
// Arrange so that no more than 3 ports are tested sequentially in each misc-compile
// TryBot to avoid any individual misc-compile TryBot from becoming a bottleneck for
// overall TryBot completion time (currently 10 minutes; see go.dev/issue/17104).
dashboard: include at most 3 ports per misc-compile TryBot Between Go 1.15, 1.16 and tip, the following 29 ports don't have a real TryBot and instead rely on misc-compile TryBots for their pre-submit coverage: • aix/ppc64 • darwin/amd64 • darwin/arm64 • dragonfly/amd64 • freebsd/386 • freebsd/arm • freebsd/arm64 • illumos/amd64 • linux/mips • linux/mips64 • linux/mipsle • linux/mips64le • linux/ppc64 • linux/ppc64le • linux/riscv64 • linux/s390x • netbsd/386 • netbsd/amd64 • netbsd/arm • netbsd/arm64 • openbsd/386 • openbsd/arm • openbsd/arm64 • openbsd/mips64 • plan9/386 • plan9/amd64 • plan9/arm • solaris/amd64 • windows/arm The previous approach for misc-compile target selection was to break them up primarily by GOOS value. However, as new architectures were added over time, some misc-compile TryBots got to a point where they were testing upwards of 5 ports (for example, misc-compile-openbsd was testing 386, amd64, arm, arm64, and mips64 architectures). Since each port is tested sequentially, allocating too many to one misc-compile TryBot can cause it to become the bottleneck of an entire TryBot run, exceeding the 10 minute completion time goal. Arrange it so misc-compile TryBot target selection is done explicitly in x/build, and pick 3 as max number of targets per TryBot for now. Based on recent timing observations, that should strike a decent balance between resource use (spinning up a builder) vs chance of a misc-compile TryBot becoming a bottleneck. It will also give us an opportunity to compare timing of 1, 2 and 3 targets per misc-compile in the future. (When we start tracking timing for TryBot completion time holistically, we'll be in a better position to refine this strategy further.) Making misc-compile target selection explicit in x/build also enables removing unnecessary duplicate misc-compile coverage from ports that already have a real TryBot (for example, openbsd/amd64 was previously tested via both the openbsd-amd64-68 TryBot and misc-compile-openbsd). This shouldn't be needed, so it's no longer done. For golang/go#17104. Fixes golang/go#32632. Change-Id: Iac918377b91af3e48780b38ffdf3153e213eeba2 Reviewed-on: https://go-review.googlesource.com/c/build/+/313210 Trust: Dmitri Shuralyov <dmitshur@golang.org> Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
2021-04-23 23:56:37 +03:00
//
// The TestTryBotsCompileAllPorts test is used to detect any gaps in TryBot coverage
// when new ports are added, and the misc-compile pairs below can be re-arranged.
//
// (In the past, we used flexible regexp patterns that matched all architectures
// for a given GOOS value. However, over time as new architectures were added,
// some misc-compile TryBot could become much slower than others.)
//
// See go.dev/issue/32632.
addMiscCompile("-windows-arm", "windows-arm", "windows-arm64")
addMiscCompile("-darwin", "darwin-amd64", "darwin-arm64")
dashboard: include at most 3 ports per misc-compile TryBot Between Go 1.15, 1.16 and tip, the following 29 ports don't have a real TryBot and instead rely on misc-compile TryBots for their pre-submit coverage: • aix/ppc64 • darwin/amd64 • darwin/arm64 • dragonfly/amd64 • freebsd/386 • freebsd/arm • freebsd/arm64 • illumos/amd64 • linux/mips • linux/mips64 • linux/mipsle • linux/mips64le • linux/ppc64 • linux/ppc64le • linux/riscv64 • linux/s390x • netbsd/386 • netbsd/amd64 • netbsd/arm • netbsd/arm64 • openbsd/386 • openbsd/arm • openbsd/arm64 • openbsd/mips64 • plan9/386 • plan9/amd64 • plan9/arm • solaris/amd64 • windows/arm The previous approach for misc-compile target selection was to break them up primarily by GOOS value. However, as new architectures were added over time, some misc-compile TryBots got to a point where they were testing upwards of 5 ports (for example, misc-compile-openbsd was testing 386, amd64, arm, arm64, and mips64 architectures). Since each port is tested sequentially, allocating too many to one misc-compile TryBot can cause it to become the bottleneck of an entire TryBot run, exceeding the 10 minute completion time goal. Arrange it so misc-compile TryBot target selection is done explicitly in x/build, and pick 3 as max number of targets per TryBot for now. Based on recent timing observations, that should strike a decent balance between resource use (spinning up a builder) vs chance of a misc-compile TryBot becoming a bottleneck. It will also give us an opportunity to compare timing of 1, 2 and 3 targets per misc-compile in the future. (When we start tracking timing for TryBot completion time holistically, we'll be in a better position to refine this strategy further.) Making misc-compile target selection explicit in x/build also enables removing unnecessary duplicate misc-compile coverage from ports that already have a real TryBot (for example, openbsd/amd64 was previously tested via both the openbsd-amd64-68 TryBot and misc-compile-openbsd). This shouldn't be needed, so it's no longer done. For golang/go#17104. Fixes golang/go#32632. Change-Id: Iac918377b91af3e48780b38ffdf3153e213eeba2 Reviewed-on: https://go-review.googlesource.com/c/build/+/313210 Trust: Dmitri Shuralyov <dmitshur@golang.org> Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
2021-04-23 23:56:37 +03:00
addMiscCompile("-mips", "linux-mips", "linux-mips64")
addMiscCompile("-mipsle", "linux-mipsle", "linux-mips64le")
addMiscCompile("-ppc", "linux-ppc64", "linux-ppc64le", "aix-ppc64")
addMiscCompile("-freebsd", "freebsd-386", "freebsd-arm", "freebsd-arm64")
addMiscCompile("-netbsd", "netbsd-386", "netbsd-amd64")
addMiscCompile("-netbsd-arm", "netbsd-arm", "netbsd-arm64")
addMiscCompile("-openbsd", "openbsd-386", "openbsd-mips64")
addMiscCompile("-openbsd-arm", "openbsd-arm", "openbsd-arm64")
addMiscCompile("-plan9", "plan9-386", "plan9-amd64", "plan9-arm")
addMiscCompile("-solaris", "solaris-amd64", "illumos-amd64")
addMiscCompile("-other-1", "dragonfly-amd64", "linux-loong64")
dashboard: include at most 3 ports per misc-compile TryBot Between Go 1.15, 1.16 and tip, the following 29 ports don't have a real TryBot and instead rely on misc-compile TryBots for their pre-submit coverage: • aix/ppc64 • darwin/amd64 • darwin/arm64 • dragonfly/amd64 • freebsd/386 • freebsd/arm • freebsd/arm64 • illumos/amd64 • linux/mips • linux/mips64 • linux/mipsle • linux/mips64le • linux/ppc64 • linux/ppc64le • linux/riscv64 • linux/s390x • netbsd/386 • netbsd/amd64 • netbsd/arm • netbsd/arm64 • openbsd/386 • openbsd/arm • openbsd/arm64 • openbsd/mips64 • plan9/386 • plan9/amd64 • plan9/arm • solaris/amd64 • windows/arm The previous approach for misc-compile target selection was to break them up primarily by GOOS value. However, as new architectures were added over time, some misc-compile TryBots got to a point where they were testing upwards of 5 ports (for example, misc-compile-openbsd was testing 386, amd64, arm, arm64, and mips64 architectures). Since each port is tested sequentially, allocating too many to one misc-compile TryBot can cause it to become the bottleneck of an entire TryBot run, exceeding the 10 minute completion time goal. Arrange it so misc-compile TryBot target selection is done explicitly in x/build, and pick 3 as max number of targets per TryBot for now. Based on recent timing observations, that should strike a decent balance between resource use (spinning up a builder) vs chance of a misc-compile TryBot becoming a bottleneck. It will also give us an opportunity to compare timing of 1, 2 and 3 targets per misc-compile in the future. (When we start tracking timing for TryBot completion time holistically, we'll be in a better position to refine this strategy further.) Making misc-compile target selection explicit in x/build also enables removing unnecessary duplicate misc-compile coverage from ports that already have a real TryBot (for example, openbsd/amd64 was previously tested via both the openbsd-amd64-68 TryBot and misc-compile-openbsd). This shouldn't be needed, so it's no longer done. For golang/go#17104. Fixes golang/go#32632. Change-Id: Iac918377b91af3e48780b38ffdf3153e213eeba2 Reviewed-on: https://go-review.googlesource.com/c/build/+/313210 Trust: Dmitri Shuralyov <dmitshur@golang.org> Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Heschi Kreinick <heschi@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
2021-04-23 23:56:37 +03:00
addMiscCompile("-other-2", "linux-riscv64", "linux-s390x", "linux-arm-arm5") // 'linux-arm-arm5' is linux/arm with GOARM=5.
addMiscCompileGo1(20, "-go1.20", "freebsd-riscv64")
// TODO: Issue 25963, get the misc-compile trybots for Android/iOS.
// Then consider subrepos too, so "mobile" can at least be included
// as a misc-compile for ^android- and ^ios-.
addBuilder(BuildConfig{
Name: "linux-amd64-nocgo",
HostType: "host-linux-amd64-bullseye",
Notes: "cgo disabled",
buildsRepo: func(repo, branch, goBranch string) bool {
b := buildRepoByDefault(repo)
switch repo {
case "perf":
// Requires sqlite, which requires cgo.
b = false
case "exp":
b = true
}
return b
},
env: []string{
"CGO_ENABLED=0",
"GO_DISABLE_OUTBOUND_NETWORK=1",
// This USER=root was required for Docker-based builds but probably isn't required
// in the VM anymore, since the buildlet probably already has this in its environment.
// (It was required because without cgo, it couldn't find the username)
"USER=root",
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-noopt",
Notes: "optimizations and inlining disabled",
HostType: "host-linux-amd64-bullseye",
buildsRepo: onlyGo,
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
"GO_GCFLAGS=-N -l",
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-ssacheck",
HostType: "host-linux-amd64-bullseye",
buildsRepo: onlyGo,
tryBot: nil, // TODO: add a func to conditionally run this trybot if compiler dirs are touched
CompileOnly: true,
Notes: "SSA internal checks enabled",
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
"GO_GCFLAGS=-d=ssa/check/on,dclstack",
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-staticlockranking",
HostType: "host-linux-amd64-bullseye",
Notes: "builder with GOEXPERIMENT=staticlockranking, see go.dev/issue/37937",
buildsRepo: func(repo, branch, goBranch string) bool {
return repo == "go"
},
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
"GOEXPERIMENT=staticlockranking",
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-unified",
HostType: "host-linux-amd64-buster",
Notes: "builder with GOEXPERIMENT=unified, see go.dev/issue/46786",
tryBot: func(repo, branch, goBranch string) bool {
// TODO(go.dev/issue/52150): Restore testing against tools repo.
return (repo == "go" /*|| repo == "tools"*/) && (goBranch == "master" || goBranch == "dev.unified")
},
buildsRepo: func(repo, branch, goBranch string) bool {
return (repo == "go" || repo == "tools") && (goBranch == "master" || goBranch == "dev.unified")
},
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
"GOEXPERIMENT=unified",
},
numTestHelpers: 1,
numTryTestHelpers: 4,
KnownIssues: []int{52150},
})
addBuilder(BuildConfig{
Name: "linux-amd64-nounified",
HostType: "host-linux-amd64-buster",
Notes: "builder with GOEXPERIMENT=nounified, see go.dev/issue/51397",
tryBot: func(repo, branch, goBranch string) bool {
return (repo == "go" || repo == "tools") && (goBranch == "master" || goBranch == "dev.unified")
},
buildsRepo: func(repo, branch, goBranch string) bool {
return (repo == "go" || repo == "tools") && (goBranch == "master" || goBranch == "dev.unified")
},
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
"GOEXPERIMENT=nounified",
},
numTestHelpers: 1,
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "linux-amd64-goamd64v3",
HostType: "host-linux-amd64-bullseye",
Notes: "builder with GOAMD64=v3, see proposal 45453 and issue 48505",
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
"GOAMD64=v3",
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-racecompile",
HostType: "host-linux-amd64-bullseye",
tryBot: nil, // TODO: add a func to conditionally run this trybot if compiler dirs are touched
CompileOnly: true,
SkipSnapshot: true,
StopAfterMake: true,
InstallRacePackages: []string{"cmd/compile", "cmd/link"},
Notes: "race-enabled cmd/compile and cmd/link",
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
})
addBuilder(BuildConfig{
cmd/coordinator, dashboard: remove some trybots, shard others wider I'm aiming to have trybot runs finish in under 5 minutes. This CL removes openbsd-386-gce58 and freebsd-386-gce101 from the trybot set. openbsd-386-gce58 is the slowest builder. It has an average speed of 722 seconds (and 95 percentile of 923 seconds) over the past week, and that's sharded over 4 machines. Too slow. It's not worth the resources to keep it as a trybot. It hasn't caught any interesting bugs. This builder will still run, but not as a pre-submit trybot. freebsd-386-gce101 is not slow, but we're removing it to shift its resources to shard other builders wider. The coordinator now supports varying the build sharding width based on whether a build is for a trybot or not. This CL defines separate numbers for each, sharding builds wider as needed for some trybots. freebsd-amd64-gce101 goes from 4 to 5 machines in try runs, and down to 3 when not in try runs. linux-amd64-race gets one more machine during try runs, and one fewer in regular runs. linux-arm goes from 7 machines always, to 3 or 8, depending on whether it's a try run. openbsd-amd64-58 goes from 4 to 3 or 6. windows-amd64-gce goes from 4 to 2 or 6. windows-amd64-race goes from 4 to 2 or 6. darwin-amd64-10_11 goes from 3 to 3 or 4. I'll see how these do over the next few days and readjust as needed. Also in this CL: fix the constants for the expected duration of make.bash, which impact when we schedule the creation of test sharding helper buildlets. We were creating them too early before, wasting resources. Change-Id: I38a9b24841e196f1eb668de058c49af8c1d1c64f Reviewed-on: https://go-review.googlesource.com/29116 Reviewed-by: Quentin Smith <quentin@golang.org>
2016-09-14 01:45:48 +03:00
Name: "linux-amd64-race",
HostType: "host-linux-amd64-bullseye",
tryBot: defaultTrySet(),
buildsRepo: defaultPlusExpBuild,
distTestAdjust: fasterTrybots,
numTestHelpers: 1,
cmd/coordinator, dashboard: remove some trybots, shard others wider I'm aiming to have trybot runs finish in under 5 minutes. This CL removes openbsd-386-gce58 and freebsd-386-gce101 from the trybot set. openbsd-386-gce58 is the slowest builder. It has an average speed of 722 seconds (and 95 percentile of 923 seconds) over the past week, and that's sharded over 4 machines. Too slow. It's not worth the resources to keep it as a trybot. It hasn't caught any interesting bugs. This builder will still run, but not as a pre-submit trybot. freebsd-386-gce101 is not slow, but we're removing it to shift its resources to shard other builders wider. The coordinator now supports varying the build sharding width based on whether a build is for a trybot or not. This CL defines separate numbers for each, sharding builds wider as needed for some trybots. freebsd-amd64-gce101 goes from 4 to 5 machines in try runs, and down to 3 when not in try runs. linux-amd64-race gets one more machine during try runs, and one fewer in regular runs. linux-arm goes from 7 machines always, to 3 or 8, depending on whether it's a try run. openbsd-amd64-58 goes from 4 to 3 or 6. windows-amd64-gce goes from 4 to 2 or 6. windows-amd64-race goes from 4 to 2 or 6. darwin-amd64-10_11 goes from 3 to 3 or 4. I'll see how these do over the next few days and readjust as needed. Also in this CL: fix the constants for the expected duration of make.bash, which impact when we schedule the creation of test sharding helper buildlets. We were creating them too early before, wasting resources. Change-Id: I38a9b24841e196f1eb668de058c49af8c1d1c64f Reviewed-on: https://go-review.googlesource.com/29116 Reviewed-by: Quentin Smith <quentin@golang.org>
2016-09-14 01:45:48 +03:00
numTryTestHelpers: 5,
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
})
addBuilder(BuildConfig{
Name: "linux-386-clang",
HostType: "host-linux-amd64-clang",
Notes: "Debian Buster + clang 7.0 instead of gcc",
env: []string{"CC=/usr/bin/clang", "GOHOSTARCH=386"},
})
addBuilder(BuildConfig{
Name: "linux-amd64-clang",
HostType: "host-linux-amd64-clang",
Notes: "Debian Buster + clang 7.0 instead of gcc",
env: []string{"CC=/usr/bin/clang"},
})
addBuilder(BuildConfig{
Name: "linux-386-sid",
HostType: "host-linux-amd64-sid",
Notes: "Debian sid (unstable)",
env: []string{"GOHOSTARCH=386"},
})
addBuilder(BuildConfig{
Name: "linux-amd64-sid",
HostType: "host-linux-amd64-sid",
Notes: "Debian sid (unstable)",
})
addBuilder(BuildConfig{
Name: "linux-amd64-fedora",
HostType: "host-linux-amd64-fedora",
Notes: "Fedora",
})
addBuilder(BuildConfig{
Name: "linux-amd64-androidemu",
HostType: "host-linux-amd64-androidemu",
env: []string{
"GOARCH=amd64",
"GOOS=linux",
"CGO_ENABLED=1",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
tryBot: func(repo, branch, goBranch string) bool {
// Only for mobile repo for now, not "go":
return repo == "mobile" && branch == "master" && goBranch == "master"
},
buildsRepo: func(repo, branch, goBranch string) bool {
return repo == "mobile" && branch == "master" && goBranch == "master"
},
Notes: "Runs GOOS=linux but with the Android emulator attached, for running x/mobile host tests.",
})
addBuilder(BuildConfig{
Name: "linux-amd64-stretch",
HostType: "host-linux-amd64-stretch",
Notes: "Debian Stretch.",
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
buildsRepo: func(repo, branch, goBranch string) bool {
return atMostGo1(goBranch, 19) && buildRepoByDefault(repo) // Stretch was EOL at the start of the 1.20 cycle.
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-bullseye",
HostType: "host-linux-amd64-bullseye",
Notes: "Debian Bullseye.",
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-buster",
HostType: "host-linux-amd64-buster",
Notes: "Debian Buster.",
env: []string{
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
})
addBuilder(BuildConfig{
Name: "linux-386-stretch",
HostType: "host-linux-amd64-stretch",
Notes: "Debian Stretch, 32-bit builder.",
env: []string{
"GOARCH=386",
"GOHOSTARCH=386",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
buildsRepo: func(repo, branch, goBranch string) bool {
return atMostGo1(goBranch, 19) && buildRepoByDefault(repo) // Stretch was EOL at the start of the 1.20 cycle.
},
})
addBuilder(BuildConfig{
Name: "linux-386-buster",
HostType: "host-linux-amd64-buster",
Notes: "Debian Buster, 32-bit builder.",
env: []string{
"GOARCH=386",
"GOHOSTARCH=386",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
})
addBuilder(BuildConfig{
Name: "linux-386-bullseye",
HostType: "host-linux-amd64-bullseye",
Notes: "Debian Bullseye, 32-bit builder.",
env: []string{
"GOARCH=386",
"GOHOSTARCH=386",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-longtest",
HostType: "host-linux-amd64-bullseye",
Notes: "Debian Bullseye with go test -short=false",
tryBot: func(repo, branch, goBranch string) bool {
onReleaseBranch := strings.HasPrefix(branch, "release-branch.")
return repo == "go" && onReleaseBranch // See issue 37827.
},
buildsRepo: func(repo, branch, goBranch string) bool {
// Test all repos, ignoring buildRepoByDefault.
// For golang.org/x repos, don't test non-latest versions.
return repo == "go" || (branch == "master" && goBranch == "master")
},
needsGoProxy: true, // for cmd/go module tests
env: []string{
"GO_TEST_TIMEOUT_SCALE=5", // give them lots of time
},
numTryTestHelpers: 4, // Target time is < 15 min for go.dev/issue/42661.
})
addBuilder(BuildConfig{
Name: "linux-amd64-longtest-race",
HostType: "host-linux-amd64-bullseye",
Notes: "Debian Bullseye with the race detector enabled and go test -short=false",
buildsRepo: func(repo, branch, goBranch string) bool {
b := buildRepoByDefault(repo)
if repo != "go" && !(branch == "master" && goBranch == "master") {
// For golang.org/x repos, don't test non-latest versions.
b = false
}
return b
},
needsGoProxy: true, // for cmd/go module tests
env: []string{
"GO_TEST_TIMEOUT_SCALE=5", // Inherited from the longtest builder.
},
numTryTestHelpers: 4, // Target time is < 15 min for go.dev/issue/56907.
})
addBuilder(BuildConfig{
Name: "linux-386-longtest",
HostType: "host-linux-amd64-bullseye",
Notes: "Debian Bullseye with go test -short=false; to get 32-bit coverage",
tryBot: func(repo, branch, goBranch string) bool {
onReleaseBranch := strings.HasPrefix(branch, "release-branch.")
return repo == "go" && onReleaseBranch // See issue 37827.
},
buildsRepo: func(repo, branch, goBranch string) bool {
b := buildRepoByDefault(repo)
if repo != "go" && !(branch == "master" && goBranch == "master") {
// For golang.org/x repos, don't test non-latest versions.
b = false
}
return b
},
needsGoProxy: true, // for cmd/go module tests
env: []string{
"GOARCH=386",
"GOHOSTARCH=386",
"GO_TEST_TIMEOUT_SCALE=5", // give them lots of time
},
numTryTestHelpers: 4, // Target time is < 15 min for go.dev/issue/42661.
})
addBuilder(BuildConfig{
Name: "js-wasm",
HostType: "host-linux-amd64-js-wasm",
tryBot: explicitTrySet("go"),
buildsRepo: func(repo, branch, goBranch string) bool {
b := buildRepoByDefault(repo)
switch repo {
case "benchmarks", "debug", "perf", "talks", "tools", "tour", "website":
// Don't test these golang.org/x repos.
b = false
}
if repo != "go" && !(branch == "master" && goBranch == "master") {
// For golang.org/x repos, don't test non-latest versions.
b = false
}
return b
},
distTestAdjust: func(run bool, distTest string, isNormalTry bool) bool {
if isNormalTry {
if strings.Contains(distTest, "/internal/") ||
strings.Contains(distTest, "vendor/golang.org/x/arch") {
return false
}
switch distTest {
case "nolibgcc:crypto/x509", "reboot":
return false
}
}
return run
},
numTryTestHelpers: 5,
env: []string{
"GOOS=js", "GOARCH=wasm", "GOHOSTOS=linux", "GOHOSTARCH=amd64",
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/workdir/go/misc/wasm",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
})
addBuilder(BuildConfig{
Name: "js-wasm-node18",
HostType: "host-linux-amd64-js-wasm-node18",
KnownIssues: []int{57017},
buildsRepo: func(repo, branch, goBranch string) bool {
b := buildRepoByDefault(repo)
switch repo {
case "benchmarks", "debug", "perf", "talks", "tools", "tour", "website":
// Don't test these golang.org/x repos.
b = false
}
if repo != "go" && !(branch == "master" && goBranch == "master") {
// For golang.org/x repos, don't test non-latest versions.
b = false
}
return b
},
distTestAdjust: func(run bool, distTest string, isNormalTry bool) bool {
if isNormalTry && (strings.Contains(distTest, "/internal/") || distTest == "reboot") {
// Skip some tests in an attempt to speed up normal trybots, inherited from CL 121938.
run = false
}
return run
},
numTryTestHelpers: 3,
env: []string{
"GOOS=js", "GOARCH=wasm", "GOHOSTOS=linux", "GOHOSTARCH=amd64",
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/workdir/go/misc/wasm",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
})
addBuilder(BuildConfig{
Name: "openbsd-amd64-71",
HostType: "host-openbsd-amd64-71",
distTestAdjust: noTestDirAndNoReboot,
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "openbsd-amd64-72",
HostType: "host-openbsd-amd64-72",
tryBot: defaultTrySet(),
distTestAdjust: noTestDirAndNoReboot,
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "openbsd-386-71",
HostType: "host-openbsd-386-71",
buildsRepo: func(repo, branch, goBranch string) bool {
// https://go.dev/issue/49529: git seems to be too slow on this
// platform.
return repo != "review" && buildRepoByDefault(repo)
},
distTestAdjust: noTestDirAndNoReboot,
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "openbsd-386-72",
HostType: "host-openbsd-386-72",
tryBot: explicitTrySet("sys"),
buildsRepo: func(repo, branch, goBranch string) bool {
// https://go.dev/issue/49529: git seems to be too slow on this
// platform.
return repo != "review" && buildRepoByDefault(repo)
},
distTestAdjust: noTestDirAndNoReboot,
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "openbsd-arm-jsing",
HostType: "host-openbsd-arm-joelsing",
SkipSnapshot: true,
FlakyNet: true,
buildsRepo: func(repo, branch, goBranch string) bool {
switch repo {
case "go", "net", "sys":
return branch == "master" && goBranch == "master"
default:
return false
}
},
distTestAdjust: noTestDirAndNoReboot,
tryBot: nil,
env: []string{
// The machine is slow.
"GO_TEST_TIMEOUT_SCALE=5",
},
})
addBuilder(BuildConfig{
Name: "openbsd-arm64-jsing",
HostType: "host-openbsd-arm64-joelsing",
SkipSnapshot: true,
FlakyNet: true,
buildsRepo: func(repo, branch, goBranch string) bool {
switch repo {
case "go", "net", "sys":
return branch == "master" && goBranch == "master"
default:
return false
}
},
distTestAdjust: noTestDirAndNoReboot,
tryBot: nil,
env: []string{
// The machine is slow.
"GO_TEST_TIMEOUT_SCALE=5",
},
})
addBuilder(BuildConfig{
Name: "openbsd-mips64-jsing",
HostType: "host-openbsd-mips64-joelsing",
SkipSnapshot: true,
FlakyNet: true,
buildsRepo: func(repo, branch, goBranch string) bool {
switch repo {
case "go", "net", "sys":
return branch == "master" && goBranch == "master"
default:
return false
}
},
distTestAdjust: noTestDirAndNoReboot,
tryBot: nil,
env: []string{
// The machine is slow.
"GO_TEST_TIMEOUT_SCALE=5",
},
})
addBuilder(BuildConfig{
Name: "netbsd-386-9_3",
HostType: "host-netbsd-386-9_3",
distTestAdjust: noTestDirAndNoReboot,
})
addBuilder(BuildConfig{
Name: "netbsd-amd64-9_3",
HostType: "host-netbsd-amd64-9_3",
distTestAdjust: noTestDirAndNoReboot,
tryBot: explicitTrySet("sys"),
})
addBuilder(BuildConfig{
Name: "netbsd-arm-bsiegert",
HostType: "host-netbsd-arm-bsiegert",
buildsRepo: func(repo, branch, goBranch string) bool {
if repo == "review" {
// https://go.dev/issue/49530: This test seems to be too slow even
// with a long scale factor.
return false
}
return buildRepoByDefault(repo)
},
distTestAdjust: noTestDirAndNoReboot,
tryBot: nil,
env: []string{
// The machine is slow.
"GO_TEST_TIMEOUT_SCALE=10",
},
FlakyNet: true,
})
addBuilder(BuildConfig{
Name: "netbsd-arm64-bsiegert",
HostType: "host-netbsd-arm64-bsiegert",
distTestAdjust: noTestDirAndNoReboot,
tryBot: nil,
env: []string{
// The machine is slow.
"GO_TEST_TIMEOUT_SCALE=10",
},
FlakyNet: true,
})
addBuilder(BuildConfig{
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
Name: "plan9-386",
HostType: "host-plan9-386-gce",
numTestHelpers: 1,
tryOnly: true, // disable it for now; Issue 31261, Issue 29801
distTestAdjust: func(run bool, distTest string, isNormalTry bool) bool {
switch distTest {
case "api",
"go_test:cmd/go": // takes over 20 minutes without working SMP
return false
}
return run
},
buildsRepo: plan9Default,
KnownIssues: []int{29801},
})
addBuilder(BuildConfig{
Name: "windows-amd64-2008",
HostType: "host-windows-amd64-2008",
distTestAdjust: noTestDirAndNoReboot,
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has modern/recent C compilers installed,
// meaning that we only want to use it with 1.20+ versions
// of Go, hence the atLeastGo1 call below; versions of Go
// prior to 1.20 will use the *-oldcc variant instead. See
// issue 35006 for more details.
return onlyGo(repo, branch, goBranch) &&
atLeastGo1(goBranch, 20)
},
env: []string{
"GOARCH=amd64",
"GOHOSTARCH=amd64",
// cmd/go takes ~188 seconds on windows-amd64
// now, which is over the 180 second default
// dist test timeout. So, bump this builder
// up:
"GO_TEST_TIMEOUT_SCALE=2",
},
})
addBuilder(BuildConfig{
Name: "windows-amd64-2008-oldcc",
HostType: "host-windows-amd64-2008-oldcc",
distTestAdjust: noTestDirAndNoReboot,
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has legacy C compilers installed, suitable
// for versions of Go prior to 1.20, hence the atMostGo1
// call below. Newer (1.20 and later) will use the
// non-oldcc variant instead. See issue 35006 for more
// details.
return onlyGo(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
},
env: []string{
"GOARCH=amd64",
"GOHOSTARCH=amd64",
// cmd/go takes ~188 seconds on windows-amd64
// now, which is over the 180 second default
// dist test timeout. So, bump this builder
// up:
"GO_TEST_TIMEOUT_SCALE=2",
},
})
addBuilder(BuildConfig{
Name: "windows-386-2008",
HostType: "host-windows-amd64-2008",
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has modern/recent C compilers installed,
// meaning that we only want to use it with 1.20+ versions
// of Go, hence the atLeastGo1 call below; versions of Go
// prior to 1.20 will use the *-oldcc variant instead. See
// issue 35006 for more details.
return defaultPlusExpBuild(repo, branch, goBranch) &&
atLeastGo1(goBranch, 20)
},
env: []string{"GOARCH=386", "GOHOSTARCH=386"},
tryBot: func(repo, branch, goBranch string) bool {
// See comment above about the atLeastGo1 call below.
dft := defaultTrySet()
return dft(repo, branch, goBranch) &&
atLeastGo1(goBranch, 20)
},
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "windows-386-2008-oldcc",
HostType: "host-windows-amd64-2008-oldcc",
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has legacy C compilers installed, suitable
// for versions of Go prior to 1.20, hence the atMostGo1
// call below. Newer (1.20 and later) will use the
// non-oldcc variant instead. See issue 35006 for more
// details.
return defaultPlusExpBuild(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
},
distTestAdjust: fasterTrybots,
env: []string{"GOARCH=386", "GOHOSTARCH=386"},
tryBot: func(repo, branch, goBranch string) bool {
// See comment above about the atMostGo1 call below.
dft := defaultTrySet()
return dft(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
},
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "windows-386-2012",
HostType: "host-windows-amd64-2012",
distTestAdjust: fasterTrybots,
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has modern/recent C compilers installed,
// meaning that we only want to use it with 1.20+ versions
// of Go, hence the atLeastGo1 call below; versions of Go
// prior to 1.20 will use the *-oldcc variant instead. See
// issue 35006 for more details.
return onlyGo(repo, branch, goBranch) &&
atLeastGo1(goBranch, 20)
},
env: []string{"GOARCH=386", "GOHOSTARCH=386"},
tryBot: func(repo, branch, goBranch string) bool {
// See comment above about the atLeastGo1 call below.
dft := defaultTrySet()
return dft(repo, branch, goBranch) &&
atLeastGo1(goBranch, 20)
},
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "windows-386-2012-oldcc",
HostType: "host-windows-amd64-2012-oldcc",
distTestAdjust: fasterTrybots,
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has legacy C compilers installed, suitable
// for versions of Go prior to 1.20, hence the atMostGo1
// call below. Newer (1.20 and later) will use the
// non-oldcc variant instead. See issue 35006 for more
// details.
return onlyGo(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
},
env: []string{"GOARCH=386", "GOHOSTARCH=386"},
tryBot: func(repo, branch, goBranch string) bool {
// See comment above about the atMostGo1 call below.
dft := defaultTrySet()
return dft(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
},
numTryTestHelpers: 4,
})
addBuilder(BuildConfig{
Name: "windows-amd64-2012-oldcc",
HostType: "host-windows-amd64-2012-oldcc",
distTestAdjust: noTestDirAndNoReboot,
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has legacy C compilers installed, suitable
// for versions of Go prior to 1.20, hence the atMostGo1
// call below. Newer (1.20 and later) will use the
// non-oldcc variant instead. See issue 35006 for more
// details.
return onlyGo(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
},
env: []string{
"GOARCH=amd64",
"GOHOSTARCH=amd64",
// cmd/go takes ~188 seconds on windows-amd64
// now, which is over the 180 second default
// dist test timeout. So, bump this builder
// up:
"GO_TEST_TIMEOUT_SCALE=2",
},
})
addBuilder(BuildConfig{
Name: "windows-amd64-2012",
HostType: "host-windows-amd64-2012",
distTestAdjust: noTestDirAndNoReboot,
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has modern/recent C compilers installed,
// meaning that we only want to use it with 1.20+ versions
// of Go, hence the atLeastGo1 call below; versions of Go
// prior to 1.20 will use the *-oldcc variant instead. See
// issue 35006 for more details.
return onlyGo(repo, branch, goBranch) &&
atLeastGo1(goBranch, 20)
},
env: []string{
"GOARCH=amd64",
"GOHOSTARCH=amd64",
// cmd/go takes ~188 seconds on windows-amd64
// now, which is over the 180 second default
// dist test timeout. So, bump this builder
// up:
"GO_TEST_TIMEOUT_SCALE=2",
},
})
addBuilder(BuildConfig{
Name: "windows-amd64-2016",
HostType: "host-windows-amd64-2016",
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has modern/recent C compilers installed,
// meaning that we only want to use it with 1.20+ versions
// of Go, hence the atLeastGo1 call below; versions of Go
// prior to 1.20 will use the *-oldcc variant instead. See
// issue 35006 for more details.
return defaultPlusExpBuild(repo, branch, goBranch) &&
atLeastGo1(goBranch, 20)
},
distTestAdjust: fasterTrybots,
env: []string{
"GOARCH=amd64",
"GOHOSTARCH=amd64",
// cmd/go takes ~188 seconds on windows-amd64
// now, which is over the 180 second default
// dist test timeout. So, bump this builder
// up:
"GO_TEST_TIMEOUT_SCALE=2",
},
tryBot: func(repo, branch, goBranch string) bool {
// See comment above about the atLeastGo1 call below.
dft := defaultTrySet()
return dft(repo, branch, goBranch) &&
atLeastGo1(goBranch, 20)
},
numTryTestHelpers: 5,
})
addBuilder(BuildConfig{
Name: "windows-amd64-2016-oldcc",
HostType: "host-windows-amd64-2016-oldcc",
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has legacy C compilers installed, suitable
// for versions of Go prior to 1.20, hence the atMostGo1
// call below. Newer (1.20 and later) will use the
// non-oldcc variant instead. See issue 35006 for more
// details.
return defaultPlusExpBuild(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
},
distTestAdjust: fasterTrybots,
env: []string{
"GOARCH=amd64",
"GOHOSTARCH=amd64",
// cmd/go takes ~188 seconds on windows-amd64
// now, which is over the 180 second default
// dist test timeout. So, bump this builder
// up:
"GO_TEST_TIMEOUT_SCALE=2",
},
tryBot: func(repo, branch, goBranch string) bool {
// See comment above about the atMostGo1 call below.
dft := defaultTrySet()
return dft(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
},
numTryTestHelpers: 5,
})
addBuilder(BuildConfig{
Name: "windows-amd64-longtest-oldcc",
HostType: "host-windows-amd64-2016-big-oldcc",
Notes: "Windows Server 2016 with go test -short=false",
tryBot: func(repo, branch, goBranch string) bool {
onReleaseBranch := strings.HasPrefix(branch, "release-branch.")
// This builder has legacy C compilers installed, suitable
// for versions of Go prior to 1.20, hence the atMostGo1
// call below. Newer (1.20 and later) will use the
// non-oldcc variant instead. See issue 35006 for more
// details.
return atMostGo1(goBranch, 19) &&
repo == "go" && onReleaseBranch // See issue 37827.
},
buildsRepo: func(repo, branch, goBranch string) bool {
// See comment above about the atMostGo1 call below.
b := defaultPlusExpBuild(repo, branch, goBranch) &&
atMostGo1(goBranch, 19)
if repo != "go" && !(branch == "master" && goBranch == "master") {
// For golang.org/x repos, don't test non-latest versions.
b = false
}
return b
},
needsGoProxy: true, // for cmd/go module tests
env: []string{
"GO_TEST_TIMEOUT_SCALE=5", // give them lots of time
},
numTryTestHelpers: 4, // Target time is < 15 min for go.dev/issue/42661.
})
addBuilder(BuildConfig{
Name: "windows-amd64-longtest",
HostType: "host-windows-amd64-2016-big",
Notes: "Windows Server 2016 with go test -short=false",
tryBot: func(repo, branch, goBranch string) bool {
onReleaseBranch := strings.HasPrefix(branch, "release-branch.")
// This builder has modern/recent C compilers installed,
// meaning that we only want to use it with 1.20+ versions
// of Go, hence the atLeastGo1 call below; versions of Go
// prior to 1.20 will use the *-oldcc variant instead. See
// issue 35006 for more details.
return atLeastGo1(goBranch, 20) &&
repo == "go" && onReleaseBranch // See issue 37827.
},
buildsRepo: func(repo, branch, goBranch string) bool {
// See comment above about the atMostGo1 call below.
b := atLeastGo1(goBranch, 20) &&
defaultPlusExpBuild(repo, branch, goBranch)
if repo != "go" && !(branch == "master" && goBranch == "master") {
// For golang.org/x repos, don't test non-latest versions.
b = false
}
return b
},
needsGoProxy: true, // for cmd/go module tests
env: []string{
"GO_TEST_TIMEOUT_SCALE=5", // give them lots of time
},
numTryTestHelpers: 4, // Target time is < 15 min for go.dev/issue/42661.
})
addBuilder(BuildConfig{
Name: "windows-amd64-oldcc-race",
HostType: "host-windows-amd64-2016-oldcc",
Notes: "Only runs -race tests (./race.bat)",
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has legacy C compilers installed, suitable
// for versions of Go prior to 1.20, hence the atMostGo1
// call below. Newer (1.20 and later) will use the
// non-oldcc variant instead. See issue 35006 for more
// details.
return atMostGo1(goBranch, 19) &&
buildRepoByDefault(repo)
},
env: []string{
"GOARCH=amd64",
"GOHOSTARCH=amd64",
// cmd/go takes ~188 seconds on windows-amd64
// now, which is over the 180 second default
// dist test timeout. So, bump this builder
// up:
"GO_TEST_TIMEOUT_SCALE=2"},
})
addBuilder(BuildConfig{
Name: "windows-amd64-race",
HostType: "host-windows-amd64-2016",
Notes: "Only runs -race tests (./race.bat)",
buildsRepo: func(repo, branch, goBranch string) bool {
// This builder has modern/recent C compilers installed,
// meaning that we only want to use it with 1.20+ versions
// of Go, hence the atLeastGo1 call below; versions of Go
// prior to 1.20 will use the *-oldcc variant instead. See
// issue 35006 for more details.
return atLeastGo1(goBranch, 20) &&
buildRepoByDefault(repo)
},
env: []string{
"GOARCH=amd64",
"GOHOSTARCH=amd64",
// cmd/go takes ~188 seconds on windows-amd64
// now, which is over the 180 second default
// dist test timeout. So, bump this builder
// up:
"GO_TEST_TIMEOUT_SCALE=2"},
})
addBuilder(BuildConfig{
Name: "windows-arm-zx2c4",
HostType: "host-windows-arm64-zx2c4",
env: []string{
"GOARM=7",
"GO_TEST_TIMEOUT_SCALE=3"},
})
addBuilder(BuildConfig{
Name: "darwin-amd64-10_14",
HostType: "host-darwin-amd64-10_14-aws",
distTestAdjust: macTestPolicy,
buildsRepo: defaultPlusExp,
})
addBuilder(BuildConfig{
Name: "darwin-amd64-10_15",
HostType: "host-darwin-amd64-10_15-aws",
distTestAdjust: macTestPolicy,
buildsRepo: defaultPlusExpBuild,
})
addBuilder(BuildConfig{
Name: "darwin-amd64-11_0",
HostType: "host-darwin-amd64-11-aws",
distTestAdjust: macTestPolicy,
buildsRepo: defaultPlusExpBuild,
})
addBuilder(BuildConfig{
Name: "darwin-amd64-12_0",
HostType: "host-darwin-amd64-12-aws",
distTestAdjust: macTestPolicy,
buildsRepo: defaultPlusExpBuild,
})
addBuilder(BuildConfig{
Name: "darwin-amd64-13",
HostType: "host-darwin-amd64-13-aws",
distTestAdjust: macTestPolicy,
buildsRepo: defaultPlusExpBuild,
})
addBuilder(BuildConfig{
Name: "darwin-amd64-nocgo",
HostType: "host-darwin-amd64-12-aws",
distTestAdjust: noTestDirAndNoReboot,
env: []string{"CGO_ENABLED=0"},
})
addBuilder(BuildConfig{
Name: "darwin-arm64-11",
HostType: "host-darwin-arm64-11",
distTestAdjust: macTestPolicy,
buildsRepo: defaultPlusExpBuild,
})
addBuilder(BuildConfig{
Name: "darwin-arm64-12",
HostType: "host-darwin-arm64-12",
distTestAdjust: macTestPolicy,
buildsRepo: defaultPlusExpBuild,
})
addBuilder(BuildConfig{
Name: "darwin-amd64-race",
HostType: "host-darwin-amd64-12-aws",
distTestAdjust: macTestPolicy,
buildsRepo: onlyGo,
})
addBuilder(BuildConfig{
Name: "ios-arm64-corellium",
HostType: "host-ios-arm64-corellium-ios",
Notes: "Virtual iPhone SE running on Corellium; owned by zenly (github.com/znly)",
buildsRepo: func(repo, branch, goBranch string) bool {
return repo == "go" && branch == "master" && goBranch == "master"
},
})
addBuilder(BuildConfig{
Name: "android-arm64-corellium",
HostType: "host-android-arm64-corellium-android",
Notes: "Virtual Android running on Corellium; owned by zenly (github.com/znly)",
buildsRepo: func(repo, branch, goBranch string) bool {
return repo == "go" && branch == "master" && goBranch == "master"
},
})
addBuilder(BuildConfig{
Name: "android-arm-corellium",
HostType: "host-android-arm64-corellium-android",
Notes: "Virtual Android running on Corellium; owned by zenly (github.com/znly)",
buildsRepo: func(repo, branch, goBranch string) bool {
return repo == "go" && branch == "master" && goBranch == "master"
},
env: []string{
"CGO_ENABLED=1",
"GOARCH=arm",
"GO_TEST_TIMEOUT_SCALE=2", // inherited from cmd/dist's default for GOARCH=arm
},
})
addBuilder(BuildConfig{
Name: "android-386-emu",
HostType: "host-linux-amd64-androidemu", // same amd64 host is used for 386 builder
Notes: "Android emulator on GCE (GOOS=android GOARCH=386)",
buildsRepo: func(repo, branch, goBranch string) bool {
b := buildRepoByDefault(repo)
switch repo {
case "mobile":
b = true
case "build", "blog", "talks", "review", "tour", "website":
b = false
}
return b
},
env: []string{
"GOARCH=386",
"GOOS=android",
"GOHOSTARCH=amd64",
"GOHOSTOS=linux",
"CGO_ENABLED=1",
},
KnownIssues: []int{42212, 51001, 52724},
})
addBuilder(BuildConfig{
Name: "android-amd64-emu",
HostType: "host-linux-amd64-androidemu",
Notes: "Android emulator on GCE (GOOS=android GOARCH=amd64)",
numTryTestHelpers: 3,
tryBot: func(repo, branch, goBranch string) bool {
// See discussion in go.dev/issue/53377.
switch repo {
case "mobile":
return true
}
return false
},
buildsRepo: func(repo, branch, goBranch string) bool {
b := buildRepoByDefault(repo)
switch repo {
case "mobile":
b = true
case "build", "blog", "talks", "review", "tour", "website":
b = false
}
return b
},
env: []string{
"GOARCH=amd64",
"GOOS=android",
"GOHOSTARCH=amd64",
"GOHOSTOS=linux",
"CGO_ENABLED=1",
},
KnownIssues: []int{42212, 51001, 52724},
})
addBuilder(BuildConfig{
Name: "illumos-amd64",
HostType: "host-illumos-amd64-jclulow",
})
addBuilder(BuildConfig{
Name: "solaris-amd64-oraclerel",
HostType: "host-solaris-oracle-amd64-oraclerel",
Notes: "Oracle Solaris release version",
FlakyNet: true,
KnownIssues: []int{51443},
})
addBuilder(BuildConfig{
Name: "linux-ppc64-sid-buildlet",
HostType: "host-linux-ppc64-sid",
FlakyNet: true,
distTestAdjust: ppc64DistTestPolicy,
env: []string{"GO_TEST_TIMEOUT_SCALE=2"}, // see go.dev/issues/44422
})
addBuilder(BuildConfig{
Name: "linux-ppc64le-buildlet",
HostType: "host-linux-ppc64le-osu",
FlakyNet: true,
distTestAdjust: ppc64DistTestPolicy,
env: []string{"GO_TEST_TIMEOUT_SCALE=2"}, // see go.dev/issues/44422
})
addBuilder(BuildConfig{
Name: "linux-ppc64le-power9osu",
HostType: "host-linux-ppc64le-power9-osu",
FlakyNet: true,
distTestAdjust: ppc64DistTestPolicy,
env: []string{"GO_TEST_TIMEOUT_SCALE=2"}, // see go.dev/issues/44422
})
addBuilder(BuildConfig{
Name: "linux-ppc64le-power10osu",
HostType: "host-linux-ppc64le-power10-osu",
FlakyNet: true,
distTestAdjust: ppc64DistTestPolicy,
env: []string{"GO_TEST_TIMEOUT_SCALE=2"}, // see go.dev/issues/44422
buildsRepo: func(repo, branch, goBranch string) bool {
return atLeastGo1(goBranch, 20) && buildRepoByDefault(repo)
},
})
addBuilder(BuildConfig{
Name: "linux-arm64-aws",
HostType: "host-linux-arm64-aws",
})
addBuilder(BuildConfig{
Name: "linux-arm64",
HostType: "host-linux-arm64-bullseye",
tryBot: defaultTrySet(),
numTryTestHelpers: 1,
})
addBuilder(BuildConfig{
Name: "linux-arm64-boringcrypto",
HostType: "host-linux-arm64-aws",
env: []string{
"GOEXPERIMENT=boringcrypto",
"GO_DISABLE_OUTBOUND_NETWORK=1",
},
buildsRepo: func(repo, branch, goBranch string) bool {
return atLeastGo1(goBranch, 19) && buildRepoByDefault(repo)
},
})
addBuilder(BuildConfig{
Name: "linux-arm64-longtest",
HostType: "host-linux-arm64-bullseye-high-disk",
Notes: "Debian Bullseye with go test -short=false",
tryBot: func(repo, branch, goBranch string) bool {
onReleaseBranch := strings.HasPrefix(branch, "release-branch.")
return repo == "go" && onReleaseBranch // See issue 37827.
},
buildsRepo: func(repo, branch, goBranch string) bool {
b := atLeastGo1(goBranch, 20) && buildRepoByDefault(repo)
if repo != "go" && !(branch == "master" && goBranch == "master") {
// For golang.org/x repos, don't test non-latest versions.
b = false
}
return b
},
needsGoProxy: true, // for cmd/go module tests
env: []string{
"GO_TEST_TIMEOUT_SCALE=5", // give them lots of time
},
numTryTestHelpers: 4, // Target time is < 15 min for go.dev/issue/42661.
})
addBuilder(BuildConfig{
Name: "linux-arm-aws",
HostType: "host-linux-arm-aws",
tryBot: defaultTrySet(),
numTryTestHelpers: 1,
env: []string{
"GOARCH=arm",
"GOARM=6",
"GOHOSTARCH=arm",
"CGO_CFLAGS=-march=armv6",
"CGO_LDFLAGS=-march=armv6",
"GO_TEST_TIMEOUT_SCALE=2", // inherited from cmd/dist's default for GOARCH=arm
},
})
addBuilder(BuildConfig{
FlakyNet: true,
HostType: "host-linux-loong64-3a5000",
Name: "linux-loong64-3a5000",
SkipSnapshot: true,
distTestAdjust: func(run bool, distTest string, isNormalTry bool) bool {
switch distTest {
case "api", "reboot":
return false
}
return run
},
buildsRepo: func(repo, branch, goBranch string) bool {
switch repo {
case "go":
return atLeastGo1(goBranch, 19)
case "net", "sys":
return branch == "master" && atLeastGo1(goBranch, 19)
default:
return false
}
},
privateGoProxy: true, // this builder is behind firewall
env: []string{
"GOARCH=loong64",
"GOHOSTARCH=loong64",
},
KnownIssues: []int{53116, 53093},
})
addBuilder(BuildConfig{
FlakyNet: true,
HostType: "host-linux-mips64le-rtrk",
Name: "linux-mips64le-rtrk",
SkipSnapshot: true,
distTestAdjust: mipsDistTestPolicy,
buildsRepo: mipsBuildsRepoPolicy,
env: []string{
"GOARCH=mips64le",
"GOHOSTARCH=mips64le",
"GO_TEST_TIMEOUT_SCALE=4", // inherited from cmd/dist's default for GOARCH=mips{,le,64,64le}
},
})
addBuilder(BuildConfig{
FlakyNet: true,
HostType: "host-linux-mips64le-rtrk",
Name: "linux-mipsle-rtrk",
SkipSnapshot: true,
distTestAdjust: mipsDistTestPolicy,
buildsRepo: mipsBuildsRepoPolicy,
env: []string{
"GOARCH=mipsle",
"GOHOSTARCH=mipsle",
"GO_TEST_TIMEOUT_SCALE=4", // inherited from cmd/dist's default for GOARCH=mips{,le,64,64le}
},
})
addBuilder(BuildConfig{
FlakyNet: true,
HostType: "host-linux-mips64-rtrk",
Name: "linux-mips64-rtrk",
SkipSnapshot: true,
distTestAdjust: mipsDistTestPolicy,
buildsRepo: mipsBuildsRepoPolicy,
env: []string{
"GOARCH=mips64",
"GOHOSTARCH=mips64",
"GO_TEST_TIMEOUT_SCALE=4", // inherited from cmd/dist's default for GOARCH=mips{,le,64,64le}
},
})
addBuilder(BuildConfig{
FlakyNet: true,
HostType: "host-linux-mips64-rtrk",
Name: "linux-mips-rtrk",
SkipSnapshot: true,
distTestAdjust: mipsDistTestPolicy,
buildsRepo: mipsBuildsRepoPolicy,
env: []string{
"GOARCH=mips",
"GOHOSTARCH=mips",
"GO_TEST_TIMEOUT_SCALE=4", // inherited from cmd/dist's default for GOARCH=mips{,le,64,64le}
},
})
addBuilder(BuildConfig{
HostType: "host-linux-riscv64-joelsing",
Name: "linux-riscv64-jsing",
SkipSnapshot: true,
FlakyNet: true,
env: []string{"GO_TEST_TIMEOUT_SCALE=4"},
distTestAdjust: riscvDistTestPolicy,
buildsRepo: func(repo, branch, goBranch string) bool {
switch repo {
case "go", "net", "sys":
return branch == "master" && goBranch == "master"
default:
return false
}
},
})
addBuilder(BuildConfig{
HostType: "host-linux-riscv64-unmatched",
Name: "linux-riscv64-unmatched",
env: []string{"GO_TEST_TIMEOUT_SCALE=4"},
FlakyNet: true,
distTestAdjust: riscvDistTestPolicy,
privateGoProxy: true, // this builder is behind firewall
buildsRepo: func(repo, branch, goBranch string) bool {
// see https://go.dev/issue/53745
if repo == "perf" {
return false
}
return onlyMasterDefault(repo, branch, goBranch)
},
})
addBuilder(BuildConfig{
Name: "linux-s390x-ibm",
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
HostType: "host-linux-s390x",
cmd/coordinator, dashboard: remove some trybots, shard others wider I'm aiming to have trybot runs finish in under 5 minutes. This CL removes openbsd-386-gce58 and freebsd-386-gce101 from the trybot set. openbsd-386-gce58 is the slowest builder. It has an average speed of 722 seconds (and 95 percentile of 923 seconds) over the past week, and that's sharded over 4 machines. Too slow. It's not worth the resources to keep it as a trybot. It hasn't caught any interesting bugs. This builder will still run, but not as a pre-submit trybot. freebsd-386-gce101 is not slow, but we're removing it to shift its resources to shard other builders wider. The coordinator now supports varying the build sharding width based on whether a build is for a trybot or not. This CL defines separate numbers for each, sharding builds wider as needed for some trybots. freebsd-amd64-gce101 goes from 4 to 5 machines in try runs, and down to 3 when not in try runs. linux-amd64-race gets one more machine during try runs, and one fewer in regular runs. linux-arm goes from 7 machines always, to 3 or 8, depending on whether it's a try run. openbsd-amd64-58 goes from 4 to 3 or 6. windows-amd64-gce goes from 4 to 2 or 6. windows-amd64-race goes from 4 to 2 or 6. darwin-amd64-10_11 goes from 3 to 3 or 4. I'll see how these do over the next few days and readjust as needed. Also in this CL: fix the constants for the expected duration of make.bash, which impact when we schedule the creation of test sharding helper buildlets. We were creating them too early before, wasting resources. Change-Id: I38a9b24841e196f1eb668de058c49af8c1d1c64f Reviewed-on: https://go-review.googlesource.com/29116 Reviewed-by: Quentin Smith <quentin@golang.org>
2016-09-14 01:45:48 +03:00
numTestHelpers: 0,
FlakyNet: true,
})
addBuilder(BuildConfig{
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
Name: "linux-s390x-crosscompile",
HostType: "host-linux-amd64-s390x-cross",
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
Notes: "s390x cross-compile builder for releases; doesn't run tests",
CompileOnly: true,
tryOnly: true, // but not in trybot set for now
env: []string{
"CGO_ENABLED=1",
"GOARCH=s390x",
"GOHOSTARCH=amd64",
"CC_FOR_TARGET=s390x-linux-gnu-gcc",
},
})
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
addBuilder(BuildConfig{
Name: "linux-amd64-localdev",
HostType: "host-linux-amd64-localdev",
Notes: "for localhost development only",
tryOnly: true,
cmd/coordinator, cmd/buildlet, cmd/gomote: add SSH support This adds an SSH server to farmer.golang.org on port 2222 that proxies SSH connections to users' gomote-created buildlet instances. For example: $ gomote create openbsd-amd64-60 user-bradfitz-openbsd-amd64-60-1 $ gomote ssh user-bradfitz-openbsd-amd64-60-1 Warning: Permanently added '[localhost]:33351' (ECDSA) to the list of known hosts. OpenBSD 6.0 (GENERIC.MP) golang/go#2319: Tue Jul 26 13:00:43 MDT 2016 Welcome to OpenBSD: The proactively secure Unix-like operating system. Please use the sendbug(1) utility to report bugs in the system. Before reporting a bug, please try to reproduce it with the latest version of the code. With bug reports, please try to ensure that enough information to reproduce the problem is enclosed, and if a known fix for it exists, include that as well. $ As before, if the coordinator process is restarted (or crashes, is evicted, etc), all gomote instances die. Not yet supported: * scp (help wanted) * not all host types are configured. most are. some will need slight config tweaks to the Docker image (e.g. adding openssh-server) Supports currently: * linux-amd64 (host type shared by 386, nacl) * linux-arm * linux-arm64 * darwin * freebsd * openbsd * plan9-386 * windows Implementation details: * the ssh server process listens on port 2222 in the coordinator (farmer.golang.org), which is behind a GKE TCP load balancer. * the ssh server library is github.com/gliderlabs/ssh * authentication is done via Github users' public keys. It's assumed that gomote user == github user. But there's a mapping in the code for known exceptions. * we can't give out access to this too widely. too many things are accessible from within the host environment if you look in the right places. Details omitted. But the Go team and other trusted gomote users can use this. * the buildlet binary has a new /connect-ssh handler that acts like a CONNECT request but instead of taking an explicit host:port, just says "give me your machine's SSH connection". The buildlet can also start sshd if needed for the environment. The /connect-ssh handler also installs the coordinator's public key. * a new buildlet client library method "ConnectSSH" hits the /connect-ssh handler and returns a net.Conn. * the coordinator's ssh.Handler is just running the OpenSSH ssh client. * because the OpenSSH ssh child process can't connect to a net.Conn, an emphemeral localhost port is created on the coordinator to proxy between the ssh client and the net.Conn returned by ConnectSSH. * The /connect-ssh handler requires http.Hijacker, which requires fully compliant net.Conn implementations as of Go 1.8. So I needed to flesh out revdial too, testing it with the golang.org/x/net/nettest package. * plan9 doesn't have an ssh server, so we use 0intro's new conterm program (drawterm without GUI support) to connect to plan9 from the coordinator ssh proxy instead of using the OpenSSH ssh client binary. * windows doesn't have an ssh server, so we enable the telnet service and the coordinator ssh proxy uses telnet instead on the backend on the private network. (There is a Windows ssh server but only in new versions.) Happy debugging over ssh! Fixes golang/go#19956 Change-Id: I80a62064c5f85af1f195f980c862ba29af4015f0 Reviewed-on: https://go-review.googlesource.com/50750 Reviewed-by: Herbie Ong <herbie@google.com> Reviewed-by: Jessie Frazelle <me@jessfraz.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2017-07-22 22:15:56 +03:00
})
addBuilder(BuildConfig{
Name: "dragonfly-amd64-622",
HostType: "host-dragonfly-amd64-622",
Notes: "DragonFly BSD 6.2.2, running on GCE",
SkipSnapshot: true,
buildsRepo: func(repo, branch, goBranch string) bool {
return atLeastGo1(goBranch, 20) && buildRepoByDefault(repo)
},
})
addBuilder(BuildConfig{
Name: "freebsd-arm-paulzhol",
HostType: "host-freebsd-arm-paulzhol",
distTestAdjust: noTestDirAndNoReboot,
SkipSnapshot: true,
FlakyNet: true,
buildsRepo: func(repo, branch, goBranch string) bool {
// This was a fragile little machine with limited memory.
// Only run a few of the core subrepos for now while
// we figure out what's killing it.
switch repo {
case "go", "sys", "net":
return true
}
return false
},
env: []string{
"GOARM=7",
"CGO_ENABLED=1",
"GO_TEST_TIMEOUT_SCALE=8", // from builder's local environment as of 2022-12-06
},
})
addBuilder(BuildConfig{
Name: "freebsd-arm64-dmgk",
HostType: "host-freebsd-arm64-dmgk",
})
addBuilder(BuildConfig{
Name: "freebsd-riscv64-unmatched",
HostType: "host-freebsd-riscv64-unmatched",
env: []string{"GO_TEST_TIMEOUT_SCALE=4"},
FlakyNet: true,
distTestAdjust: riscvDistTestPolicy,
privateGoProxy: true, // this builder is behind firewall
SkipSnapshot: true, // The builder has a slow uplink bandwidth.
buildsRepo: func(repo, branch, goBranch string) bool {
// see https://go.dev/issue/53745
if repo == "perf" {
return false
}
return onlyMasterDefault(repo, branch, goBranch)
},
})
addBuilder(BuildConfig{
Name: "plan9-arm",
HostType: "host-plan9-arm-0intro",
distTestAdjust: noTestDirAndNoReboot,
buildsRepo: plan9Default,
KnownIssues: []int{49338},
env: []string{
"GO_TEST_TIMEOUT_SCALE=3", // from builder's local environment as of 2022-12-06
},
})
addBuilder(BuildConfig{
Name: "plan9-amd64-0intro",
HostType: "host-plan9-amd64-0intro",
distTestAdjust: func(run bool, distTest string, isNormalTry bool) bool {
run = noTestDirAndNoReboot(run, distTest, isNormalTry)
switch distTest {
case "api",
"go_test:cmd/go": // takes over 20 minutes without working SMP
return false
}
return run
},
buildsRepo: plan9Default,
KnownIssues: []int{49756, 49327},
})
addBuilder(BuildConfig{
Name: "plan9-386-0intro",
HostType: "host-plan9-386-0intro",
distTestAdjust: func(run bool, distTest string, isNormalTry bool) bool {
run = noTestDirAndNoReboot(run, distTest, isNormalTry)
switch distTest {
case "api",
"go_test:cmd/go": // takes over 20 minutes without working SMP
return false
}
return run
},
buildsRepo: plan9Default,
KnownIssues: []int{50137, 50878},
})
addBuilder(BuildConfig{
Name: "aix-ppc64",
HostType: "host-aix-ppc64-osuosl",
env: []string{
"PATH=/opt/freeware/bin:/usr/bin:/etc:/usr/sbin:/usr/ucb:/usr/bin/X11:/sbin:/usr/java7_64/jre/bin:/usr/java7_64/bin",
},
buildsRepo: func(repo, branch, goBranch string) bool {
switch repo {
case "vulndb", "vuln":
// vulndb currently uses a dependency which does not build cleanly
// on aix-ppc64. Until that issue is resolved, skip vulndb on
// this builder.
// (https://go.dev/issue/49218).
return false
}
return buildRepoByDefault(repo)
},
})
addBuilder(BuildConfig{
Name: "linux-amd64-wsl",
HostType: "host-linux-amd64-wsl",
Notes: "Windows 10 WSL2 Ubuntu",
FlakyNet: true,
SkipSnapshot: true, // The builder has a slow uplink bandwidth.
privateGoProxy: true, // this builder is behind firewall
})
addBuilder(BuildConfig{
Name: "linux-amd64-perf",
HostType: "host-linux-amd64-perf",
Notes: "Performance testing for linux-amd64",
buildsRepo: func(repo, branch, goBranch string) bool {
if repo == "benchmarks" {
// Benchmark the main Go repo.
return true
}
if repo == "tools" {
// Benchmark x/tools.
//
// When benchmarking subrepos, we ignore the Go
// commit and always use the most recent Go
// release, meaning we get identical duplicate
// runs for each Go commit that runs at the
// same subrepo commit.
//
// Limit to running on release branches since
// they have far fewer Go commits than tip,
// thus reducing the the number of duplicate
// runs.
return strings.HasPrefix(goBranch, "release-branch.")
}
return false
},
RunBench: true,
SkipSnapshot: true,
KnownIssues: []int{53538},
})
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
}
// addBuilder adds c to the Builders map after doing some checks.
func addBuilder(c BuildConfig) {
if c.Name == "" {
panic("empty name")
}
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
if c.HostType == "" {
panic(fmt.Sprintf("missing HostType for builder %q", c.Name))
}
if _, dup := Builders[c.Name]; dup {
panic("dup name " + c.Name)
}
if c.HostConfig().GoogleReverse && !c.IsReverse() {
panic("GoogleReverse is set but the builder isn't reverse")
}
all: split builder config into builder & host configs Our builders are named of the form "GOOS-GOARCH" or "GOOS-GOARCH-suffix". Over time we've grown many builders. This CL doesn't change that. Builders continue to be named and operate as before. Previously the build configuration file (dashboard/builders.go) made each builder type ("linux-amd64-race", etc) define how to create a host running a buildlet of that type, even though many builders had identical host configs. For example, these builders all share the same host type (a Kubernetes container): linux-amd64 linux-amd64-race linux-386 linux-386-387 And these are the same host type (a GCE VM): windows-amd64-gce windows-amd64-race windows-386-gce This CL creates a new concept of a "hostType" which defines how the buildlet is created (Kube, GCE, Reverse, and how), and then each builder itself references a host type. Users never see the hostType. (except perhaps in gomote list output) But they at least never need to care about them. Reverse buildlets now can only be one hostType at a time, which simplifies things. We were no longer using multiple roles per machine once moving to VMs for OS X. gomote continues to operate as it did previously but its underlying protocol changed and clients will need to be updated. As a new feature, gomote now has a new flag to let you reuse a buildlet host connection for different builder rules if they share the same underlying host type. But users can ignore that. This CL is a long-standing TODO (previously attempted and aborted) and will make many things easier and faster, including the linux-arm cross-compilation effort, and keeping pre-warmed buildlets of VM types ready to go. Updates golang/go#17104 Change-Id: Iad8387f48680424a8441e878a2f4762bf79ea4d2 Reviewed-on: https://go-review.googlesource.com/29551 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
2016-09-22 00:27:37 +03:00
if _, ok := Hosts[c.HostType]; !ok {
panic(fmt.Sprintf("undefined HostType %q for builder %q", c.HostType, c.Name))
}
if c.SkipSnapshot && (c.numTestHelpers > 0 || c.numTryTestHelpers > 0) {
panic(fmt.Sprintf("config %q's SkipSnapshot is not compatible with sharded test helpers", c.Name))
}
for i, issue := range c.KnownIssues {
if issue == 0 {
panic(fmt.Errorf("config %q's KnownIssues slice has a zero issue at index %d", c.Name, i))
}
}
types := 0
for _, fn := range []func() bool{c.IsReverse, c.IsContainer, c.IsVM} {
if fn() {
types++
}
}
if types != 1 {
panic(fmt.Sprintf("build config %q host type inconsistent (must be Reverse, Image, or VM)", c.Name))
}
Builders[c.Name] = &c
}
// tryNewMiscCompile is an intermediate step towards adding a real addMiscCompile TryBot.
// It adds a post-submit-only builder with KnownIssue, GoDeps set to the provided values,
// and runs on a limited set of branches to get test results without potential disruption
// for contributors. It can be modified as needed when onboarding a misc-compile builder.
func tryNewMiscCompile(suffix, rx string, knownIssue int, goDeps []string) {
if knownIssue == 0 {
panic("tryNewMiscCompile: knownIssue parameter must be non-zero")
}
addBuilder(BuildConfig{
Name: "misc-compile" + suffix,
HostType: "host-linux-amd64-bullseye",
buildsRepo: func(repo, branch, goBranch string) bool { return repo == "go" && branch == "master" },
KnownIssues: []int{knownIssue},
GoDeps: goDeps,
env: []string{"GO_DISABLE_OUTBOUND_NETWORK=1"},
CompileOnly: true,
Notes: fmt.Sprintf("Tries buildall.bash to cross-compile & vet std+cmd packages for "+rx+", but doesn't run any tests. See go.dev/issue/%d.", knownIssue),
allScriptArgs: []string{
// Filtering pattern to buildall.bash:
rx,
},
})
}
// fasterTrybots is a distTestAdjust policy function.
// It skips (returns false) the test/ directory and reboot tests for trybots.
func fasterTrybots(run bool, distTest string, isNormalTry bool) bool {
if isNormalTry {
if strings.HasPrefix(distTest, "test:") || distTest == "reboot" {
return false // skip test
}
}
return run
}
// noTestDirAndNoReboot is a distTestAdjust policy function.
// It skips (returns false) the test/ directory and reboot tests for all builds.
func noTestDirAndNoReboot(run bool, distTest string, isNormalTry bool) bool {
if strings.HasPrefix(distTest, "test:") || distTest == "reboot" {
return false // skip test
}
return run
}
// ppc64DistTestPolicy is a distTestAdjust policy function
// that's shared by linux-ppc64le, -ppc64le-power{9,10}-osu, and -ppc64.
func ppc64DistTestPolicy(run bool, distTest string, isNormalTry bool) bool {
if distTest == "reboot" {
// Skip test. It seems to use a lot of memory?
// See https://go.dev/issue/35233.
return false
}
return run
}
// mipsDistTestPolicy is a distTestAdjust policy function
// that's shared by the slow mips builders.
func mipsDistTestPolicy(run bool, distTest string, isNormalTry bool) bool {
switch distTest {
case "api", "reboot":
return false
}
return run
}
// mipsBuildsRepoPolicy is a buildsRepo policy function
// that's shared by the slow mips builders.
func mipsBuildsRepoPolicy(repo, branch, goBranch string) bool {
switch repo {
case "go", "net", "sys":
return branch == "master" && goBranch == "master"
default:
return false
}
}
// riscvDistTestPolicy is same as mipsDistTestPolicy for now.
var riscvDistTestPolicy = mipsDistTestPolicy
// TryBuildersForProject returns the builders that should run as part of
// a TryBot set for the given project.
// The project argument is of the form "go", "net", "sys", etc.
// The branch is the branch of that project ("master", "release-branch.go1.12", etc)
// The goBranch is the branch of Go to use. If proj == "go", then branch == goBranch.
func TryBuildersForProject(proj, branch, goBranch string) []*BuildConfig {
return buildersForProject(proj, branch, goBranch, (*BuildConfig).BuildsRepoTryBot)
}
// isBuilderFunc is the type of functions that report whether a builder
// should be run given a project, branch and goBranch.
type isBuilderFunc func(conf *BuildConfig, proj, branch, goBranch string) bool
// buildersForProject returns the builders that should be run for the given project,
// using isBuilder to test each builder.
// See TryBuildersForProject for the valid forms of proj, branch and goBranch.
func buildersForProject(proj, branch, goBranch string, isBuilder isBuilderFunc) []*BuildConfig {
var confs []*BuildConfig
for _, conf := range Builders {
if isBuilder(conf, proj, branch, goBranch) {
confs = append(confs, conf)
}
}
sort.Slice(confs, func(i, j int) bool {
return confs[i].Name < confs[j].Name
})
return confs
}
// atLeastGo1 reports whether branch is "release-branch.go1.N" where N >= min.
// It assumes "master" and "dev.*" branches are already greater than min, and
// always includes them.
func atLeastGo1(branch string, min int) bool {
if branch == "master" {
return true
}
if strings.HasPrefix(branch, "dev.") {
// Treat dev branches current.
// If a dev branch is active, it will be current.
// If it is not active, it doesn't matter anyway.
// TODO: dev.boringcrypto.go1.N branches may be the
// exception. Currently we only build boringcrypto
// on linux/amd64 and windows/386, which support all
// versions of Go, so it doesn't actually matter.
return true
}
major, minor, ok := version.ParseReleaseBranch(branch)
return ok && major == 1 && minor >= min
}
// atMostGo1 reports whether branch is "release-branch.go1.N" where N <= max.
// It assumes "master" branch is already greater than max, and doesn't include it.
func atMostGo1(branch string, max int) bool {
major, minor, ok := version.ParseReleaseBranch(branch)
return ok && major == 1 && minor <= max
}
// onlyGo is a common buildsRepo policy value that only builds the main "go" repo.
func onlyGo(repo, branch, goBranch string) bool { return repo == "go" }
// onlyMasterDefault is a common buildsRepo policy value that only builds
// default repos on the master branch.
func onlyMasterDefault(repo, branch, goBranch string) bool {
return branch == "master" && goBranch == "master" && buildRepoByDefault(repo)
}
// plan9Default is like onlyMasterDefault, but also omits repos that are
// both filesystem-intensive and unlikely to be relevant to plan9 users.
func plan9Default(repo, branch, goBranch string) bool {
switch repo {
case "review":
// The x/review repo tests a Git hook, but the plan9 "git" doesn't have the
// same command-line API as "git" everywhere else.
return false
case "website":
// The x/website tests read and check the website code snippets,
// which require many filesystem walk and read operations.
return false
case "vulndb", "vuln":
// vulncheck can't read plan9 binaries.
return false
default:
return onlyMasterDefault(repo, branch, goBranch)
}
}
// disabledBuilder is a buildsRepo policy function that always return false.
func disabledBuilder(repo, branch, goBranch string) bool { return false }
// macTestPolicy is the test policy for Macs.
//
// We have limited Mac resources. It's not worth wasting time testing
// portable things on them. That is, if there's a slow test that will
// still fail slowly on another builder where we have more resources
// (like linux-amd64), then there's no point testing it redundantly on
// the Macs.
func macTestPolicy(run bool, distTest string, isNormalTry bool) bool {
if strings.HasPrefix(distTest, "test:") {
return false
}
switch distTest {
case "reboot", "api", "doc_progs",
"wiki", "bench_go1", "codewalk":
return false
}
if isNormalTry {
switch distTest {
case "runtime:cpu124", "race", "moved_goroot":
return false
}
// TODO: more. Look at bigquery results once we have more data.
}
return run
}