In some cases, we can end up linking some things with
--static-libstdc++. The notable (only?) example of that is for the
clang-plugin, and that happens because it gets some of its flags from
llvm-config, which contains --static-libstdc++ because clang itself is
built that way.
When that happens, the combination of --static-libstdc++ and
stdc++compat breaks the build because they have conflicting symbols,
which is very much by design.
There are two ways out of this:
- avoiding either -static-libstdc++ or stdc++compat
- work around the symbol conflicts
The former is not totally reliable ; we'd have to accurately determine
if we're in a potentially conflicting case, and remove one of the two in
that case, and while we can do that for the cases we explicitly know
about, that's not future-proof, and might fail just as much in the
future.
So we go with the latter. The way we do this is by defining all the
std++compat symbols weak, such that at link time, they're overridden by
any symbol with the same name. When building with -static-libstdc++,
libstdc++.a provides those symbols so the linker eliminates the weak
ones. When not building with -static-libstdc++, the linker keeps the
symbols from stdc++compat. That last assertion is validated by the
long-standing CHECK_STDCXX test that we run when linking shared
libraries and programs.
That still leaves the symbols weak in the final shared
libraries/programs, which is a change from the current setup, but
shouldn't cause problems because when using versions of libstdc++.so
that do provide those symbols, it's fine to use the libstdc++.so version
anyways.
Bump the minimum required version of the Rust toolchain to
the current stable release so we can take advantage of new
features.
Highlights of the 1.19.0 release:
* C-compatible `union` (untagged enums).
* Support for Visual Studio 2017.
* Non-capturing closures can be coerced to `fn` bindings.
* Numeric field names in tuple struct initializers.
* Higher macro recursion limit.
* `break` can return a value from `loop` expressions.
* Better error handling with mis-configured Visual Studio environments.
This change also enables 1.18.0 features. Some highlights:
* `pub(mod)` &c. for better control of symbol visibility.
* struct packing for better memory footprint in generated code.
* Faster build times.
MozReview-Commit-ID: 2OpUjAcytpE
--HG--
extra : rebase_source : 2ed0d7c4e7b78c26f7a7476e7b284bf1bdbe7c8b
Mostly removals, but also adding docs on how to create a
multilingual package right now.
I think I'd like to take another pass at those docs in a follow-up.
MozReview-Commit-ID: Dkw4MJ5DLyb
--HG--
extra : rebase_source : 4f79499e376cf6ddd23169a1c4525ed6b551a740
It becomes a library of some sort, so that multiple scripts can benefit
from it to build different versions of GCC.
The GPG key associated with GCC is also refreshed from keys.gnupg.net,
adding a new subkey, used to sign newer versions of GCC (and
postprocessed with pgpstrip to make it smaller).
We're soon going to build multiple versions of clang and gcc for linux,
and we need to differentiate them. Furthermore, there is a need for the
base-toolchains builds to use a fixed version of clang and gcc. So
rename the clang and gcc toolchain jobs to include their version, add
aliases to satisfy all existing jobs, and adjust the base-toolchains
jobs to use the explicit version.
--HG--
rename : build/build-clang/clang-linux64.json => build/build-clang/clang-3.9-linux64.json
rename : taskcluster/scripts/misc/build-gcc-linux.sh => taskcluster/scripts/misc/build-gcc-4.9-linux.sh
The lld linker creates separate segments for purely executable sections
(such as .text) and sections preceding those (such as .rel.dyn). Neither
gold nor bfd ld do that, and just put all those sections in the same
executable segment.
Since elfhack is putting its executable code between the two relocation
sections, it ends up in a non-executable segment, leading to a crash
when it's time to run that code.
We thus insert the elfhack code before the first executable section
instead of between the two relocation sections (which is where the
elfhack data lies, and stays).
--HG--
extra : rebase_source : ab18eb9ac518d69a8639ad0e785741395b662112
Since bug 635961, building with relro makes elfhack try to use the bss
data for a temporary function pointer. If there is not enough space for
a pointer in the bss, elfhack will complain it couldn't find the bss.
In normal circumstances, this is most likely fine. Libraries with a bss
so small that it can't fit a pointer are already too small to be
elfhacked anyways. In Firefox, the two libraries with the smallest bss
have enough space for two pointers, and aren't elfhacked (libmozgtk.so
and libplds4.so).
However, the testcase that is used during the build to validate that
elfhack works doesn't have a large enough bss on x86-64, making elfhack
bail out, and the build fail as a consequence.
This, in turn, is due to the only non-thread-local zeroed data being an
int, which is not enough to fit a pointer on x86-64. We thus make it a
size_t.
--HG--
extra : rebase_source : bca2ddbf9d4a5e8786881fc524d642c38d610227
Configure now detects VCS info. Configure now detects Watchman.
We can combine the two so configure can detect if Mercurial
is configured with Watchman enabled.
This commit does two things:
1) collects the Mercurial config so it is available to downstream checks
2) examines the config for presence and state of the fsmonitor
extension
We don't yet do anything with the fsmonitor state. But it should be
useful soon. Also, the return value is kinda wonky. This will almost
certainly be improved as soon as there is an actual consumer.
MozReview-Commit-ID: HyHZ2X8VI0h
--HG--
extra : rebase_source : e53928127470340275f0c0f07db72b536bba885b
extra : source : a8373914cbfd9b8595fc24f36c876cab0a26c02a
Often you only want to evaluate a function if all its dependencies
are true. Expressing this in a "when" can be difficult. So let's
add a convenience decorator for it.
The existing code for @depends_if() was refactored to take an
evaluation function as its first argument. This prevents some
duplicate code and turns @depends_if() and @depends_all() into
one-liners.
MozReview-Commit-ID: Jbugvf0lioM
--HG--
extra : rebase_source : 177741b80ac4fbfb547d6b36a6f5777fe514d91a
extra : source : 0c2bc12f4ebe44428385745266d2fd158e0c3382
Build Stylo (the styling system from servo) by default in all
builds for win32, win64, macOS and linux64 targets. It was
previously enabled for automation builds, so this just changes
the behaviour for local developer builds.
Note that this introduces a new dependency on libclang for the
binding generator. If you're developing on a tier-1 platform,
run `./mach boostrap` to install a working copy. Otherwise
llvm+libclang 4.0.1 is recommended.
Remove the explicit --enable-stylo=build in mozconfig.stylo
in favour of the configure default.
Add mozconfig.stylo to the hazard and debug-asan mozconfigs
so LLVM_CONFIG is defined properly for those builds.
Based on a patch by Bobby Holly in bug 1356991.
MozReview-Commit-ID: C2wRNl7JHpz
--HG--
extra : rebase_source : 1ed7c36a64e25b235a26864592cd7ea969a4cd25
Configure now detects VCS info. Configure now detects Watchman.
We can combine the two so configure can detect if Mercurial
is configured with Watchman enabled.
This commit does two things:
1) collects the Mercurial config so it is available to downstream checks
2) examines the config for presence and state of the fsmonitor
extension
We don't yet do anything with the fsmonitor state. But it should be
useful soon. Also, the return value is kinda wonky. This will almost
certainly be improved as soon as there is an actual consumer.
MozReview-Commit-ID: HyHZ2X8VI0h
--HG--
extra : rebase_source : d245d316cc8a27b2827b7824204549b91465bd34
Often you only want to evaluate a function if all its dependencies
are true. Expressing this in a "when" can be difficult. So let's
add a convenience decorator for it.
The existing code for @depends_if() was refactored to take an
evaluation function as its first argument. This prevents some
duplicate code and turns @depends_if() and @depends_all() into
one-liners.
MozReview-Commit-ID: Jbugvf0lioM
--HG--
extra : rebase_source : 38b4af7b668830589126e8a83f5d5ab73914d07c
In bug 1181040, we added ${var}_IS_SET variables for
mk_add_options-defined variables. In the two years since, that has never
been used for anything else than MOZ_PGO_IS_SET, and the only use for
that has now been removed, so remove those ${var}_IS_SET variables.
--HG--
extra : rebase_source : 2fc9abe0c3badbf06f3858fcf326237e67891fee
Note this will only enable it on try, autoland and mozilla-inbound,
which are the only branches where sccache are enabled at the moment.
Enabling on more builds (or not) is the subject of bug 1373460.
Also note that bug 1181040, that ensured PGO builds weren't using
sccache mentions that back then, link times went up when using sccache
(with -Z7) vs. without, but that was presumably with MSVC 2013. Try
suggests link times are the same using sccache now (still using -Z7,
pending bug 1318370).
--HG--
extra : rebase_source : 9f9d87714f3c841b790eb7d692ea0968b1755b70
We want most builds to be actually using sccache, so we include
mozconfig.cache from mozconfig.common. However, since the --with-ccache
configure option doesn't exist on non-compile jobs (e.g. artifact
builds), we move to using the CCACHE environment variable instead, which
allows us to unset it in mozconfig.no-compile.
And since mozconfig.no-compile is always included where no_sccache is
set, we can remove that variable.
--HG--
extra : rebase_source : a8c743de1fd7a3c0fbc53f7c233df36585897767
Some gaia-related code was also found and removed as part of the
cleanup.
MozReview-Commit-ID: DEjVSljzzu1
--HG--
extra : rebase_source : 58c4c34df44a258d90029853f29ea01338bd142b
Bug 1365859 introduced a dependency on the Visual Studio binary 'fxc'
to generate Shader bytecode. This was unavailable when compiling for
Windows on Linux as part of a MinGW build.
This commit adds a configure check for fxc, and also searches for
fxc2, which was written (https://github.com/tomrittervg/fxc2) to be
a tiny application that wraps D3DCompileFromFile and can produce
similar (but not exactly the same) output as fxc.
fxc2 is compiled using MinGW for Windows, and runs under wine, so
we need to check for wine also.
Finally, fxc outputs some include information fxc2 doesn't, so
we will just change that assert to not take effect.
MozReview-Commit-ID: 8LVxuODi6cV
--HG--
extra : rebase_source : 9116d266663284d6594e34aa53bd37eae01ba67f
The |mach try| implementation isn't really testing specific so figured while doing this refactor
anyway we could take the opportunity to move it to /tools. This renames the "autotry" module to
"syntax" as one of the 'selectors' under the newly created tryselect module.
MozReview-Commit-ID: DmnGU5sKuCk
--HG--
rename : testing/tools/autotry/autotry.py => tools/tryselect/selectors/syntax.py
extra : rebase_source : 52f2a61d97107b8d50e3894ea0f5cd9c24d4108d
This option was added ~decade ago; AFAICT from bug archaeology, the
option was added to prevent our servers from being overwhelmed.
Somewhere over the years, however, we obtained more capable servers and
the option disappeared from mozconfigs. It seems moderately unlikely
that we'll have a need for this option again, and we could reintroduce
this patch very easily in any event. Let's go ahead and remove it.
The values that we need to find in the registry can be inconsistent across
different installations, so we retrieve values from both views in our search
for a valid SDK. This also ensures this works for 32-bit and 64-bit python.
Mercurial installs on Windows often provide both a native hg.exe
and a hg shell script. This allows them to work from both Windows
and UNIX-like shells.
While the which Python package searches for .exe and other common
executable extensions on Windows automatically, it doesn't prefer
these results to an extension-less value. This meant that "hg"
was resolved to the shell script. When we went to run it, it
failed because it isn't a valid Win32 application.
We work around the problem by preferring "hg.exe" over "hg"
for the binary name.
Ideally this would be platform specific. But the harm should be
minimal.
--HG--
extra : rebase_source : 1354c739de1ff738812ad100176ad7e9d08fadde
This is now required as of 00ef8018730c.
Landing on a CLOSED TREE.
--HG--
extra : rebase_source : 410e4a677cdc1f1238f778a00078ac6e6de420bd
extra : amend_source : e81cf510948bedd76ac77790eb96aae1fe3cda1c
For reasons unknown to me, Windows CI is periodically failing to find
the Mercurial binary.
In addition, we've also reimplemented various VCS logic throughout
the build system. There is room to cut down on code complexity by
e.g. recording VCS info in configure instead of determining it
at run-time.
Also, for forensic purposes it is sometimes desirable to know which
VCS tool is in use by a build and which version of that tool is being
used.
This commit adds VCS type detection, binary searching, and version
resolution to configure.
substs now contains VCS_CHECKOUT_TYPE, HG, and GIT, which can be
consulted by downstream consumers.
If the Mercurial or Git versions could not be resolved, all variables
are not set. Otherwise, VCS_CHECKOUT_TYPE and one of HG or GIT is set.
If MOZ_AUTOMATION is set, we require that the VCS info be resolved.
This helps prevents weirdness in automation due to unexpected
environment configuration.
MozReview-Commit-ID: AMLy0Hfx5rD
--HG--
extra : rebase_source : edef9165d32dc47308a14b0fbabce3c1d3d28176
And include code to work around a bug on older Python versions.
MozReview-Commit-ID: 4pBnMQQJOGB
--HG--
extra : rebase_source : 6f7c5784230bd37b3496b9bb1781e8d342f741b4
developer_options (!--enable-release) implies to search for and prefer
the gold linker. As part of porting gold detection to moz.configure
in d0e782180741 (bug 1351109), the logic inadvertently changed to
require gold when either --enable-gold or !--enable-release were
present.
This commit relaxes the requirement to find gold when only --enable-gold
is true.
MozReview-Commit-ID: HTBicWNUkvy
--HG--
extra : rebase_source : dd3938a7914f5db6c315fb775e7cc5ea177bf600
check_prog, when used with a `when` argument, doesn't work the same way
as putting it under a `with only_when()` block, while it should. The
difference comes from the fact that `with only_when()` applies the
`when` to every option and depends used in the block (which check_prog
calls a bunch of).
So, we "manually" apply the `when` to all option and depends in
check_prog. An alternative solution would be to put the whole function
under a `with only_when()` block, but that would mean reindenting the
whole function.
Either way, as a consequence, this requires the `when` to have a
dependency on --help for "non-trivial" functions, which fortunately,
there's only one of.
--HG--
extra : rebase_source : d91eca9e303c7062394d92a526983714ef3e612f
LLVM_CONFIG, per the contents of toolkit/moz.configure, is tied to
--enable-stylo, but it currently is set on all types of builds. It
currently happens to work, but it's actually not meant to, and sure
enough, the fix for bug 1374727 exacerbates that.
So we create a new mozconfig.stylo file that enables stylo and sets
LLVM_CONFIG, such that only build types that do enable stylo have
LLVM_CONFIG set.
--HG--
extra : rebase_source : 01277a79951888046c0b8e29c61cfc3b049ee0f0
Wes Kocher
Mike Hommey
Ralph Giles
Axel Hecht
Sylvestre Ledru
Cameron McCormack
Nick Alexander
Gregory Szorc
Paul Adenot
Sebastian Hengst
Carsten "Tomcat" Book
Tom Ritter
Andrew Halberstadt
Dustin J. Mitchell
Nathan Froyd
Bob Owen
Ryan VanderMeulen