This change amends the initial README section to make it clear that
- the Go project has officially adopted modules as the way to manage dependencies in Go;
- the Module System is no longer a prototype, and it's ready for production use.
Fixes golang/go#35894
Previously, the "-x" in
git submodule foreach --recursive git clean -x
was being applied to the "git submodule" command not the "git clean"
command. git submodule foreach expects the inner command to be wrapped
in a single string, so pass git clean as a single string.
Verified this works as expected on the command line.
Originally submitted as #2164 by Daniele Rodina
<geaaru@sabayonlinux.org>, re-submitted by me to amend the commit and
merge immediately.
Fixes#2164.
Fixes#2168.
The most recent version of staticcheck drops the "--ignore" flag in
favor of inline ignore directives, so we need to use those or fix on
the older version.
Good news is two of our ignore directives are now no longer needed.
Updates dominikh/go-tools#514.
Updates dominikh/go-tools#515.
Fixes#2178.
This enables support for private gitlab files and other places where
the URL requires basic authentication.
Initial proposal and implementation came from Johnny
(github.com/wesgur).
Fixes#2061.
Fixes#1898.
Apply the same change as in golang.org/cl/175219 to this copy of the
parseMetaGoImports function, helping keep them in sync.
The "mod" type is not a real version control system (VCS), it applies
only when in module mode. Skip it and continue to consider only real
VCS types.
This resolves parseMetaGoImports returning a "multiple meta tags match
import path" error on packages that offer go-import meta tags with both
a true VCS and the "mod" type.
Reference: https://golang.org/cmd/go/#hdr-Remote_import_pathsFixes#2151
In January, #2102 was closed with 5ae9d8b to build ARMv6/7 binaries.
This seems to be working, but there are no release files uploaded to
Github for this architecture, because the wrong filenames are used in
.travis-ci. This PR fixes that.
It must be noted that the dep install script is already modified to
expect a ARMv6/7 binary in the Github releases, so it errors out on
404 instead of emitting a clear error message which it did before:
```
Fetching https://github.com/golang/dep/releases/download/v0.5.1/dep-linux-arm
Request failed with code 404
```
Originally submitted by Github user confiks as #2135.
Originally submitted by Iskander Sharipov, who has signed the CLA, as
https://github.com/golang/dep/pull/2029; submitting under my name to
get past the CLA bot.
Signed-off-by: Iskander Sharipov <quasilyte@gmail.com>
Simplify `x += 1` further to `x++`.
This was originally submitted by Iskander Sharipov, who's signed the
CLA, but the CLA bot is not letting me submit on his behalf, so I'm
submitting it.
Signed-off-by: Iskander Sharipov <quasilyte@gmail.com>
On path to close#1807 on [oauth2_proxy arm support](https://github.com/pusher/oauth2_proxy/issues/16)
Tested:
- Building armv6 and arm64 binaries on amd64
- Running arm binaries
- Using armv6 binary on RPi 3B to build another project [oauth2_proxy](https://github.com/kskewes/oauth2_proxy/tree/build)
(`dep ensure`)
Not verified:
- Install script completes as binaries not published, looks okay
though.
- `hack/test.bash` - must have missed something.
armv6 (RPi 3B):
```
karl@pi1:~/go/src/github.com/pusher/oauth2_proxy $ dep version
dep:
version : v0.5.0-31-g73b3afef-dirty
build date : 2019-02-01
git hash : 73b3afef
go version : go1.11.5
go compiler : gc
platform : linux/arm
features : ImportDuringSolve=false
karl@pi1:~/go/src/github.com/pusher/oauth2_proxy $ make dep
dep ensure --vendor-only
karl@pi1:~/go/src/github.com/pusher/oauth2_proxy $ uname -m
armv7l
karl@pi1:~ $ ./install.sh
ARCH = arm
OS = linux
Will install into /home/karl/go/bin
Fetching https://github.com/golang/dep/releases/latest..
Release Tag = v0.5.0
Fetching https://github.com/golang/dep/releases/tag/v0.5.0..
Fetching https://github.com/golang/dep/releases/download/v0.5.0/dep-linux-arm..
Request failed with code 404
karl@pi1:~ $ ./test.bash
go test: -race is only supported on linux/amd64, linux/ppc64le, freebsd/amd64, netbsd/amd64, darwin/amd64 and windows/amd64
```
arm64 (Rock64):
```
rock64@k8s-w-01:~$ ./dep-linux-arm64 version
dep:
version : v0.5.0-31-g73b3afef-dirty
build date : 2019-02-01
git hash : 73b3afef
go version : go1.11.5
go compiler : gc
platform : linux/arm64
features : ImportDuringSolve=false
rock64@k8s-w-01:~$ uname -m
aarch64
rock64@k8s-w-01:~$ ./install.sh
ARCH = arm64
OS = linux
Will install into /home/rock64/go/bin
Fetching https://github.com/golang/dep/releases/latest..
Release Tag = v0.5.0
Fetching https://github.com/golang/dep/releases/tag/v0.5.0..
Fetching https://github.com/golang/dep/releases/download/v0.5.0/dep-linux-arm64..
Request failed with code 404
```
Previously, if dep tried to clone a repository over ssh that contained
zero padded file modes, dep would fail to clone and then hang. You can
reproduce this failure with the following .gitconfig:
[url "git@github.com:"]
insteadOf = https://github.com/
[transfer]
fsckobjects = true
[fetch]
fsckobjects = true
[receive]
fsckobjects = true
and the following Gopkg.toml. (It is not my intention to single out
this project - I searched Github for Go projects with zero padded file
mode errors, and found this one.)
[[constraint]]
name = "github.com/remogatto/gospeccy"
branch = "master"
`dep ensure` hangs because the git clone operation spins up an `ssh`
process as a child process. `cmd.Process.Kill` kills the parent `git`
operation, but not the child `ssh` operation, so it's orphaned and we
don't return properly from the function. (It's unclear to me at this
point why the ssh operation did not return when it completed.)
By sending the negative Pgrp value to the Kill() function we can kill
the entire child process group, not just the "parent child." See
https://medium.com/@felixge/killing-a-child-process-and-all-of-its-children-in-go-54079af94773
for more information.
Updates #1257.
The tip tests were failing due to the lack of a GOCACHE environment
variable, and bad defaults in tests where home directories did not
exist. Fix these by setting the GOCACHE to a place we know exists.
This and the Xenial fix combined allow us to re-enable the test suite,
e.g. to permit failures in the test suite to actually fail the build.
(We may still want to address the GOCACHE tests locally for users that
do not have a GOCACHE environment variable set; as it stands, they
still fail.)
Gofmt in Go 1.11 beta 2 handles struct alignment differently, as
demonstrated in this diff. Running "make test" makes these same
whitespace changes to the files in question.
Previously it would error with the following message:
gps/pkgtree/pkgtree.go:258:30: +build comment must appear before package clause and be followed by a blank line
The Go builder has no problem ignoring this declaration, but surround
it in quotes to be perfectly safe.
When running the test runner or the test coverage runner, ignore
failures in a CI environment.
We need to fix the tests, too, but first we need to get the build back
to green so we can merge changes.
Updates #2089.
Daniel Martí
sam boyer
Alberto Donizetti
Sunny
Anthony Fok
Kevin Burke
Dmitri Shuralyov
jmank88
Kevin Burke
Cristhian Amaya
Oleg Schwann
Xiang Dai
Tim Heckman
Iskander Sharipov
tariqibrahim
Karl Skewes
Matthew Cale
Benjamin Wuethrich
Iskander (Alex) Sharipov