Android requires us to use -fprofile-generate / -fprofile-use instead of
-fprofile-instr-generate / -fprofile-instr-use. Additionally, we have
to pass in the path to write out the profile data as /data/local/tmp,
since the cwd from where the Fennec process is invoked is not writeable.
Differential Revision: https://phabricator.services.mozilla.com/D20697
--HG--
extra : moz-landing-system : lando
In bug 1522354, we changed host and target detection to not invoke
config.sub, assuming the output from config.guess would satisfy our
needs in split_target.
It turns out that on some plaforms, that doesn't work out, so, while we
still skip config.sub, we now catch errors from split_target when doing
so, and try again after running config.sub when split_target fails.
Differential Revision: https://phabricator.services.mozilla.com/D19937
Consequently, this removes:
- MOZ_LIBPRIO, which is now always enabled.
- non_msvc_compiler, which is now always true.
- The cl.py wrapper, since it's not used anymore.
- CL_INCLUDES_PREFIX, which was only used for the cl.py wrapper.
- NONASCII, which was only there to ensure CL_INCLUDES_PREFIX still
worked in non-ASCII cases.
This however keeps a large part of detecting and configuring for MSVC,
because we still do need it for at least headers, libraries, and midl.
Depends on D19614
Differential Revision: https://phabricator.services.mozilla.com/D19615
--HG--
extra : moz-landing-system : lando
When the build system compiler is clang, and bindgen autodetection
actually finds a different clang via llvm-config, or a different clang
was given with --with-clang-path, we do want the proper flags to be
used for that clang, so we always get the right flags for that clang
instead of trying to shortcut.
Bug 1526857 will take care of making things more proper, while this is a
quick fix to unbreak builds in some unfortunately common kind of local
setups.
Differential Revision: https://phabricator.services.mozilla.com/D19328
--HG--
extra : moz-landing-system : lando
libclang 3.9 has a bug that makes bindgen unable to distinguish some typedefs
from the underlying type, which matters for bug 1523071.
We have had quite a few workarounds for this bug and I don't really want to add
more, since in this case it is non-trivial. I think requiring libclang 4.0+ is
reasonable at this point.
Of the distros that can't build Firefox out of the box with clang, dropping support
for clang 3.9 would only break Ubuntu 14.04 LTS, which support ends April 2019,
right before we release 67.
Differential Revision: https://phabricator.services.mozilla.com/D18889
--HG--
rename : build/build-clang/clang-3.9-linux64.json => build/build-clang/clang-4.0-linux64.json
rename : taskcluster/scripts/misc/build-clang-3.9-linux.sh => taskcluster/scripts/misc/build-clang-4.0-linux.sh
extra : moz-landing-system : lando
These are copies of the corresponding gcc-based base toolchain tasks,
with FORCE_GCC and the gcc dependency removed.
We also tweak things a little for those builds to actually end up green.
Differential Revision: https://phabricator.services.mozilla.com/D18903
--HG--
extra : moz-landing-system : lando
This apparently was necessary back when Android support was added 8.5
years ago, but is not desirable as a flag set globally. But it turns out
everything builds just fine without it now.
Differential Revision: https://phabricator.services.mozilla.com/D19128
--HG--
extra : moz-landing-system : lando
These are copies of the corresponding gcc-based base toolchain tasks,
with FORCE_GCC and the gcc dependency removed.
We also tweak things a little for those builds to actually end up green.
Differential Revision: https://phabricator.services.mozilla.com/D18903
--HG--
extra : moz-landing-system : lando
basic_bindgen_cflags's function is to set some flags for use with
rust-bindgen through clang/libclang for C++ code. Part of the flags it
sets are for the C++ standard, and the target.
Unfortunately, some of the logic wrt target-specific flags is currently
broken. It wants to apply per-compiler flags on Windows, but fails to do
so. First, because the condition test is wrong, and second, because it
only cares about msvc and not clang-cl.
OTOH, we already have those flags when the compiler is clang or
clang-cl. And we already have code to get the right flags for a given
compiler. So when the compiler is not clang or clang-cl, we can use that
to get the right flags for the clang we're going to use for bindgen.
Depends on D18316
Differential Revision: https://phabricator.services.mozilla.com/D18317
--HG--
extra : moz-landing-system : lando
In upcoming changes, we're going to use the normal compiler flags for
bindgen instead of having a separate logic for essentially the same
flags (and there's not much reason not to use the same flags after all).
One hiccup, though is that for some reason, --target=i686-linux-gnu
doesn't work with bindgen while it works with clang in the same setup.
But -m32 does.
Considering -m32 and -m64 are standard flags and that we're using them
in many cases, it doesn't hurt to use them instead of --target with
clang.
While we're doing that, we might as well refactor a little to avoid the
multiple branches handling the use of -m32/-m64, and fix the theoretical
compile-x86_64-with-x86-clang-cl case, as well as the weird check for
aarch64.
And because only two cases actually require -Xclang, only one of which
requires a condition, and the control flow is not too complex to avoid
flag duplication, we just remove the append_flag function; it was too
confusing, with all the cases we skipped it when -Xclang was not wanted.
Differential Revision: https://phabricator.services.mozilla.com/D18316
--HG--
extra : moz-landing-system : lando
This removes all .flake8 files except for the one at the root of the repo.
Instead we use the new 'per-file-ignores' config introduced in 3.7. To ignore
specific errors in a subdirectory, add a line like this to the root .flake8:
[per-file-ignores]
path/to/subdir/*: E100, F200, ...
The reasons for this change are:
1. Unblock flake8 blacklist (bug 1367092).
2. Simplify configuration and code.
3. Encourage more consistent styling.
4. Improve performance.
5. Greater editor consistency.
Differential Revision: https://phabricator.services.mozilla.com/D18354
--HG--
extra : moz-landing-system : lando
Since ./mach bootstrap installs Android SDK into ~/.mozbuild, we should detect
this location as default SDK install path.
Also, --with-android-max-sdk and --with-android-min-sdk are still in android.m4
because confvars.sh sets MOZ_ANDROID_MIN_SDK_VERSION.
Differential Revision: https://phabricator.services.mozilla.com/D15463
--HG--
extra : moz-landing-system : lando
Sometimes we want to store state that only applies to a particular srcdir, but
there isn't a standard directory where this lives. Let's add an argument to
'get_state_dir()' to provide an "official" place.
The new API to get the local state dir is 'get_state_dir(srcdir=True)'. Like
the global state dir, this directory is not guaranteed to exist. A reference to
this value can also be obtained via 'self._mach_context.local_state_dir' from
within a mach command (in this case it will be created automatically if it
doesn't exist).
Note: we should probably just make sure both exist at mach startup, but it felt
outside the scope of this change.
Differential Revision: https://phabricator.services.mozilla.com/D15724
--HG--
extra : moz-landing-system : lando
mozboot.util.get_state_dir() returns a tuple of (<path>, <bool). The bool
denotes whether or not the state dir came from an environment variable.
But this value is only used in a single place, and is very easy to test for
anyway. It's not worth the added complexity it imposes on all other consumers
of this function. Let's just make this function return the path.
Differential Revision: https://phabricator.services.mozilla.com/D15723
--HG--
extra : moz-landing-system : lando
We default to -Zi for Windows debug flags, which is for generating debug info
into a PDB file, but clang-cl doesn't actually implement that so -Zi is
an alias for -Z7 in clang-cl:
http://clang.llvm.org/docs/UsersManual.html#id9
sccache has special handling for -Zi because multiple compiles writing
to the same PDB file is not cacheable, so we've always overridden -Zi with -Z7
in CI when using sccache.
Given that everyone should be using clang-cl nowadays and MSVC will no longer
be supported soon, this patch changes the default and removes some PDB
file name flag setting from rules.mk, as well as the no-longer-necessary
overrides from mozconfig.cache.
Differential Revision: https://phabricator.services.mozilla.com/D18264
--HG--
extra : moz-landing-system : lando
Also use armv7a as the default when no target is given at all.
Also change bootstrap to create a simpler mozconfig. The downside is
that the resulting mozconfig would not work when building older
revisions.
Add unit tests for this as well as the simplications added in bug
1523341.
Differential Revision: https://phabricator.services.mozilla.com/D17906
--HG--
extra : moz-landing-system : lando
While not related, it turns out this test was opted in by ourselves,
while it's practically useless. Recent autoconf documentation says it's
obsolescent, and even autoconf 2.13 documentation said with "recent
versions of the GNU C compiler", the issue it's detecting "is becoming
a less prevalent problem". Recent as of 20 years ago.
Differential Revision: https://phabricator.services.mozilla.com/D18054
--HG--
extra : moz-landing-system : lando
Now that everything is lined up nicely, we don't need those anymore.
We can fill cargo-linker.bat with HOST_LINKER and HOST_LINKER_LIBPATHS,
which, unfortunately, can't be used as is, so we add a variant for
substitution in .bat files.
Differential Revision: https://phabricator.services.mozilla.com/D18034
--HG--
extra : moz-landing-system : lando
Now that it's automatically derived and that in-tree mozconfigs are
setting DIA_SDK_PATH, we don't need to set WIN_DIA_SDK_BIN_DIR on
automation, or elsewhere, for that matter.
Differential Revision: https://phabricator.services.mozilla.com/D17915
--HG--
extra : moz-landing-system : lando
Clang 8 seems to generate destructor decls from different source
locations which breaks an assertion in the code. This patch updates the
code to remove the assertion and more robustly handle the new
declarations.
Differential Revision: https://phabricator.services.mozilla.com/D17902
--HG--
extra : moz-landing-system : lando
We reject MSVC compilers < 2017 already, there's no point checking for
smaller versions after that.
Differential Revision: https://phabricator.services.mozilla.com/D17770
--HG--
extra : moz-landing-system : lando
And remove all the variables that configure will figure out for us as a
consequence. This has the side effect of making the automation builds a
little more like local builds, in that they don't rely on preset PATH,
LIB, etc. for the build to work.
Differential Revision: https://phabricator.services.mozilla.com/D17789
--HG--
extra : moz-landing-system : lando
We currently rely on WIN_DIA_SDK_BIN_DIR being passed, but we can
actually derive it from the DIA SDK directory. So we now do that, except
when it's given explicitly.
While in the vicinity, move the dia2.h check to python configure.
With WIN_DIA_SDK_BIN_DIR being derived and not set when dia2.h is not
found, we don't really need MSVC_HAS_DIA_SDK anymore, so we just check
for WIN_DIA_SDK_BIN_DIR to determine whether to build dump_syms or not.
One exception to the above is when WIN_DIA_SDK_BIN_DIR is passed in,
which we only keep for the in-tree mozconfigs for now. We'll remove that
possibility after bug 1523201.
Depends on D17892
Differential Revision: https://phabricator.services.mozilla.com/D17893
--HG--
extra : moz-landing-system : lando
It turns out, we don't need to `mk_add_options export` the variables
from the in-tree mozconfigs. If anything, that causes problems when
trying to simplify the mozconfigs, because it makes the variables
exported from .mozconfig.mk, overriding what configure may change and
store in autoconf.mk.
All the variables are handled by configure in a way that makes them
available in autoconf.mk, so there's no loss there, and with the
python/shell-based mozconfig loader, it turns out we don't need to go
through extra normalization via cmd.
autospider.py, being its own pseudo-mozconfig parser, still does need
it, though, but it was hooking into it already, so just inline that.
Differential Revision: https://phabricator.services.mozilla.com/D17769
--HG--
extra : moz-landing-system : lando
Previously, we hardcoded HostX64 because configure would autodetect a
x86 host on x64 machines, but the x86 MSVC compiler wouldn't be
suitable.
While the x86 MSVC compiler might still not be suitable, now that
configure detects x64 hosts properly, when the configure host is
detected as x86, we can't even execute the x64 compiler, so we can at
least try with the x86 one correctly.
Differential Revision: https://phabricator.services.mozilla.com/D17788
--HG--
extra : moz-landing-system : lando
The combination of bug 1515579 and bug 1520394 made things harder for
old-configure, because it doesn't necessarily have a complete view of
the search PATH that has been used. This doesn't actually cause problems
on non-Windows builds because things work out fine, but on Windows,
some of the executions of clang-cl in old-configure insist on being able
to find a MSVC install. That, again, doesn't currently cause problems in
general on local builds because clang-cl finds it through the registry
(presumably), and on automation, because it's in the `mk_add_options
export`'ed PATH, but the latter is due to change.
Depends on D17772
Differential Revision: https://phabricator.services.mozilla.com/D17783
--HG--
extra : moz-landing-system : lando
Also, while here, replace subprocess.check_output with check_cmd_output.
Depends on D17771
Differential Revision: https://phabricator.services.mozilla.com/D17772
--HG--
extra : moz-landing-system : lando
Bug 1520394 changed things such that the configure sandbox is using a
copy of os.environ. So when mozconfig injects environment changes, they
only affect the sandbox. Which means when the which module uses
os.environ to get PATH, it gets the original unmodified environment.
So instead of relying on the which module getting PATH itself, we feed
it with it. It's worth noting that the which module adds `.` on Windows,
but we don't copy this behavior, because in the context of configure,
it's actually not important (`.` would be the topobjdir).
Differential Revision: https://phabricator.services.mozilla.com/D17771
--HG--
extra : moz-landing-system : lando
And use it as an alternative to vswhere, instead of the current late
check. This allows to use VC_PATH when using a MSVC archive that is not
installed through the VS installer, and not have to care about PATH.
Depends on D17786
Differential Revision: https://phabricator.services.mozilla.com/D17787
--HG--
extra : moz-landing-system : lando
- Only expose it as well as --with-visual-studio-version when the host
system is Windows.
- Don't run vswhere twice (once for host and once for target).
Depends on D17785
Differential Revision: https://phabricator.services.mozilla.com/D17786
--HG--
extra : moz-landing-system : lando
config.guess doesn't actually figure out the real host system on
Windows. All it does is end up finding the CPU for which the msys used
for the shell is built for. Which has the unfortunate effect that if you
build on a 64-bits host, configure pretends the host is 32-bits.
So instead, rely on some environment variables that should be set
mostly everywhere, and derive the corresponding triplet.
This has two effects:
- configure now knows that the host is 64-bits when it is.
- consequently, configure will build a 64-bits Firefox on 64-bits hosts
by default (since by default, the target is derived from the host), and
that's actually a desired side effect.
Depends on D17620
Differential Revision: https://phabricator.services.mozilla.com/D17621
--HG--
extra : moz-landing-system : lando
- We assume the output from config.guess is sufficiently normalized for
our needs.
- If we partially derive the target from the host (with only the cpu
given for --target), we assume the derived value doesn't need
canonicalization.
Differential Revision: https://phabricator.services.mozilla.com/D17620
--HG--
extra : moz-landing-system : lando
In fact, "simply" use whatever python configure does to find a shell to
execute config.guess and config.sub, and get both the mozconfig content
and the real, canonicalized target alias. This has the side effect of
making builds with --target=$cpu use a complete obj-$cpu-$os default
objdir instead of obj-$cpu. This will also allow to change the
host-guessing logic without having to duplicate code.
Differential Revision: https://phabricator.services.mozilla.com/D17618
--HG--
extra : moz-landing-system : lando
which, at this point, is only the AC_SUBST, so we don't even need to
give it to old-configure anymore.
Depends on D17207
Differential Revision: https://phabricator.services.mozilla.com/D17208
--HG--
extra : moz-landing-system : lando
which, at this point, is only the AC_SUBST, so we don't even need to
give it to old-configure anymore.
Depends on D17207
Differential Revision: https://phabricator.services.mozilla.com/D17208
--HG--
extra : moz-landing-system : lando
Because old-configure is only refreshed when, essentially,
old-configure.in changes, hardcoded (absolute) paths don't necessarily
match the build environment of the current build.
So instead, use an environment variable that we pass from python
configure when invoking old-configure.
Also do dummy changes to old-configure.in so that old-configure is
refreshed at least once to get the environment-based value.
Differential Revision: https://phabricator.services.mozilla.com/D17077
--HG--
extra : moz-landing-system : lando
Back when those were added, option defaults could not indirectly depend
on `target` or `host`, but that changed with bug 1322025.
As a side effect, this allows to turn them into straight js_options
without resorting to the manual add_old_configure_arg.
Differential Revision: https://phabricator.services.mozilla.com/D16664
--HG--
extra : moz-landing-system : lando
Rust ships with some natvis files that enable nicer display of Rust standard
library types in Microsoft's debuggers. rustc will add the right linker
options to include them when it invokes the linker but since we don't link
libxul with rustc we need to explicitly pass them to the linker ourselves.
This change locates all natvis files in the Rust sysroot and adds them
to the libxul link line.
Differential Revision: https://phabricator.services.mozilla.com/D16544
--HG--
extra : moz-landing-system : lando
Add a dummy change to old-configure.in so that old-configure is
force-refreshed.
Differential Revision: https://phabricator.services.mozilla.com/D16797
--HG--
extra : moz-landing-system : lando
Back when those were added, option defaults could not indirectly depend
on `target` or `host`, but that changed with bug 1322025.
As a side effect, this allows to turn them into straight js_options
without resorting to the manual add_old_configure_arg.
Differential Revision: https://phabricator.services.mozilla.com/D16664
--HG--
extra : moz-landing-system : lando
Before bug 1322025, running configure --help along --host and/or
--target would show the options that are specific to the
host/target/cross-compilation (e.g. those with when=target_is_windows)
which was convenient. Bug 1322025 unfortunately broke that, and this
restores the feature.
Differential Revision: https://phabricator.services.mozilla.com/D16673
--HG--
extra : moz-landing-system : lando
Instead, use the actual log file name as configured through python
logging. This doesn't make a practical difference now, but will when
subconfigure logs to the toplevel log.
Depends on D16666
Differential Revision: https://phabricator.services.mozilla.com/D16667
--HG--
extra : moz-landing-system : lando
We're going to want to log some more. At the moment, this just adds the
output of subconfigure to top-level config.log, but this will become
more useful down the road.
Depends on D16665
Differential Revision: https://phabricator.services.mozilla.com/D16666
--HG--
extra : moz-landing-system : lando
Back when those were added, option defaults could not indirectly depend
on `target` or `host`, but that changed with bug 1322025.
As a side effect, this allows to turn them into straight js_options
without resorting to the manual add_old_configure_arg.
Differential Revision: https://phabricator.services.mozilla.com/D16664
--HG--
extra : moz-landing-system : lando
Use an I/O wrapper on the configure output handler to add the "js/src>"
prefix.
Depends on D16643
Differential Revision: https://phabricator.services.mozilla.com/D16644
--HG--
extra : moz-landing-system : lando
It was manually removed because autoconf's AC_OUTPUT_SUBDIRS passes it,
but we've not been using AC_OUTPUT_SUBDIRS for a while.
Depends on D16637
Differential Revision: https://phabricator.services.mozilla.com/D16638
--HG--
extra : moz-landing-system : lando
The code only actually handles autoconf 2.5 cache, and was useful when
we have such subconfigures, that would die when things changed in ways
that would make the subconfigure abort on its own. But we have no such
things left, and while we may want to clear the cache for old-configure
(for e.g. bug 1520232), this is not helping to get there.
Differential Revision: https://phabricator.services.mozilla.com/D16636
--HG--
extra : moz-landing-system : lando
Instead, include the module and inline its main function.
Depends on D16622
Differential Revision: https://phabricator.services.mozilla.com/D16623
--HG--
extra : moz-landing-system : lando
This happens to remove the last use of perl from configure.
Depends on D16621
Differential Revision: https://phabricator.services.mozilla.com/D16622
--HG--
extra : moz-landing-system : lando
The File class is now used only in one place, only for its mtime
property, which can be replaced with os.path.getmtime.
Depends on D16402
Differential Revision: https://phabricator.services.mozilla.com/D16403
--HG--
extra : moz-landing-system : lando
Because subconfigure.py is only used for js/src and we're not going to
add any more subconfigures, we can hardcode js/src into the script, and
remove the branches that aren't taken as a consequence.
Depends on D16382
Differential Revision: https://phabricator.services.mozilla.com/D16402
--HG--
extra : moz-landing-system : lando
It was useful to run the prepare phase and the actual subconfigure
separately, but it's not anymore after all the recent simplications.
Depends on D16381
Differential Revision: https://phabricator.services.mozilla.com/D16382
--HG--
extra : moz-landing-system : lando
Because we only ever run one subconfigure, the machinery to execute
several is not useful anymore. Inlining it allows to simplify the code
too, because it doesn't need to be generic anymore. This also removes
the last remaining bits of acwinpaths.m4.
Also remove now unused support for --list in build/subconfigure.py.
Depends on D16380
Differential Revision: https://phabricator.services.mozilla.com/D16381
--HG--
extra : moz-landing-system : lando
We only run it for js/src now, so get_config_files always returns empty lists.
Depends on D16378
Differential Revision: https://phabricator.services.mozilla.com/D16379
--HG--
extra : moz-landing-system : lando
This only affects the compiler version in the build system; not the
version reported by the compiler.
--HG--
extra : histedit_source : 37fc1a2090d872ca5e95d022e9f93b620b2d8dbb
When doing cross-compiles for Desktop, system pkg-config is going to be
looking for .pc files for the host, but systems setup for cross-compiles
come with a toolchain-prefixed pkg-config, which we should be looking
for.
Differential Revision: https://phabricator.services.mozilla.com/D16318
--HG--
extra : moz-landing-system : lando
When doing cross-compiles for Desktop, system pkg-config is going to be
looking for .pc files for the host, but systems setup for cross-compiles
come with a toolchain-prefixed pkg-config, which we should be looking
for.
Differential Revision: https://phabricator.services.mozilla.com/D16318
--HG--
extra : moz-landing-system : lando
MOZ_D3D_CPU_SUFFIX and MOZ_HAS_WINSDK_WITH_D3D are not used in the
build, and nothing includes d3d10.h except some angle code in a
preprocessed branch that is only taken for a macro we never define,
so we don't move the code corresponding to those. We also simplify the
detection code, which is convoluted now that it doesn't search for
multiple different DLLs.
Differential Revision: https://phabricator.services.mozilla.com/D16295
--HG--
extra : moz-landing-system : lando
In bug 1494022 there are problems using an SDK that is too new, which
can cause developers to spend many hours debugging an issue that we
could've checked up front.
This patch gets the version from the SDKSettings.plist file at the root
of the SDK directory, and checks it against min and max supported
versions.
The current min version is 10.11, since that is what we build with in
automation. The current max version is 10.13, since 10.14+ currently has
an issue with a black screen at startup.
Differential Revision: https://phabricator.services.mozilla.com/D16140
--HG--
extra : moz-landing-system : lando
Because MSVC compilers only support one architecture, we need to search
"cl" in different toolchain search paths for each of the host and
target, especially when they are different.
Likewise for the library paths for the linker. Ideally we'd pass
-LIBPATH both for host and target, but that has implications for rust
that I don't want to have to figure just now.
Depends on D15263
Differential Revision: https://phabricator.services.mozilla.com/D15264
--HG--
extra : moz-landing-system : lando
While it was preferable to use lld-link with msvc for performance
reasons when we defaulted to compile with msvc, now that we build with
clang-cl by default, it's now better to use msvc's linker when using
msvc's compiler, especially for aarch64, which is the last place where
we actually need msvc.
Differential Revision: https://phabricator.services.mozilla.com/D15262
--HG--
extra : moz-landing-system : lando
In bug 1259382, some workarounds were added to make the build system
alter PATH and not use absolute paths for toolchain programs, because
autoconf and the build system doesn't deal with spaces in those very
well. But later in bug 1290040, we made find_program return Windows
short paths (without spaces), which alleviates the need for those
workarounds.
We still, however, and unfortunately, need to alter PATH to account for
the fact that MSVC DLLs are not necessarily alongside the compiler
executables...
Depends on D15181
Differential Revision: https://phabricator.services.mozilla.com/D15182
--HG--
extra : moz-landing-system : lando
Like for other windows platforms. This currently doesn't make a
difference, but will with next change.
Differential Revision: https://phabricator.services.mozilla.com/D15181
--HG--
extra : moz-landing-system : lando
LLVM_PROFDATA needs the toolchain search dir, per bug 1515579.
Also, most of the options actually don't do anything useful with
artifact builds. In fact, the only one that artifact builds would need
is MOZ_PGO. So we move to options back to toolchain.configure, somewhere
late enough ; except MOZ_PGO, that we move to the top-level
moz.configure (because we don't need a separate file for one option).
Differential Revision: https://phabricator.services.mozilla.com/D16152
--HG--
extra : moz-landing-system : lando
Firefox uses multiple processes. It has intentional leaks, and when
running with ASAN, we have suppressions to eliminate those. When running
ASAN builds through CI tests, when Firefox exits, each of the processes
(parent and child) exits and goes through its leaks and when there are
(which is a given), the ASAN runtime runs llvm-symbolizer to symbolicate
and match against suppressions. So each process runs llvm-symbolizer. At
the same time.
Some of the addresses to symbolicate are in libxul. Which contains all
DWARF info, making it a ~1GB monster. Oh, and because you're lucky,
things align perfectly such that libxul size is a multiple of the page
size. That makes llvm-symbolizer pread() the file instead of mmap()ing
it. Did I say there are multiple processes? So suddenly you have n
processes simultaneously allocating and filling 1GB of memory each, on
CI machines that have enough memory for the job they usually run, but
not enough for a sudden rush of n GB.
And things go awry. When you're lucky and the OOM killer didn't take
care of killing the CI entirely, symbolication couldn't happen and the
suppressions are not matched, and leaks are reported.
This all turns out it originates in how llvm-symbolicate chooses between
pread() and mmap(), which turns out is just defaults not being made for
binary files.
Differential Revision: https://phabricator.services.mozilla.com/D16010
--HG--
extra : moz-landing-system : lando
In bug 1514288 we started to save the output from running Firefox during
PGO builds into log files, but they aren't correctly uploaded if the run
fails. This presents a problem for categorizing failed PGO builds if the
profileserver returns an error code (eg: bug 1517939), since the error
messages could be hidden away in log files that aren't uploaded.
Instead, we can generate them directly into the artifacts directory so
that they are always uploaded. Additionally, the log files are displayed
if the run fails so that the error messages are displayed in the log for
easier bug categorization.
Differential Revision: https://phabricator.services.mozilla.com/D15881
--HG--
extra : moz-landing-system : lando
Instead of per-document. This also allows to reuse this thread-pool if needed
for other stuff, like parallel CSS parsing (#22478), and to share more code with
Gecko, which is always nice.
This cherry-picks https://github.com/servo/servo/pull/22487, with a few minor
fixes to the build that are landing as part of the sync associated to this bug,
and an lsan exception tweak to point to the right module since it's moving.
The output from running the browser during PGO builds can have innocuous
error messages in them, but show up in treeherder as potential messages
to include when filing bugs. We can just save the output from these runs
as files and upload them as artifacts instead, so they don't show up in
the build log but are available for inspection if necessary.
MozReview-Commit-ID: 3VdVCKVkZNI
Differential Revision: https://phabricator.services.mozilla.com/D15154
--HG--
extra : moz-landing-system : lando
In bug 1259382, some workarounds were added to make the build system
alter PATH and not use absolute paths for toolchain programs, because
autoconf and the build system doesn't deal with spaces in those very
well. But later in bug 1290040, we made find_program return Windows
short paths (without spaces), which alleviates the need for those
workarounds.
We still, however, and unfortunately, need to alter PATH to account for
the fact that MSVC DLLs are not necessarily alongside the compiler
executables...
Depends on D15181
Differential Revision: https://phabricator.services.mozilla.com/D15182
--HG--
extra : moz-landing-system : lando
Mike Shal
Cosmin Sabou
Masatoshi Kimura
Chris Hartjes
Marco Castelluccio
Mike Hommey
Razvan Maries
Daniel Varga
Coroiu Cristina
Thomas Daede
Dorel Luca
Nathan Froyd
Chris Manchester
Emilio Cobos Álvarez
David Major
shindli
arthur.iakab
Kris Maglione
Csoregi Natalia
Andrew Halberstadt
Tom Ritter
Johann Hofmann
Wes Kocher
Sylvestre Ledru
Makoto Kato
Ted Mielczarek
Gurzau Raul
Alex Chronopoulos
Mark Banner
Sebastian Hengst
Tom Prince
Kartikaya Gupta
Ryan VanderMeulen
Jan Henning
Brindusan Cristian
Margareta Eliza Balazs
Ehsan Akhgari
Nick Alexander
Geoff Brown
Dustin J. Mitchell
Ciure Andrei