If -L flags get into CLikeCompiler::build_wrapper_args, they will be
correctly detected and the /LINK flag added to the list. However,
CompilerArgs::__iadd__ will reorder them to the front, thinking they're
GNU-style flags, and this will cause MSVC to ignore them after
conversion.
The fix is twofold:
1. Convert all the linker args into their compiler form, making sure the
/LINK argument is dropped (see 2)
2. Insert /LINK into extra_args if not already present
3. Execute in situ the unix_to_native replacement, ensuring no further
reordering occurs.
Fixes#11113
When C6000 support was added in #12246, TI compilers were given the correct version argument.
This broke the previous check which relied on an error being thrown by the compiler.
Now that we have access to Environment in get_assert_args, we can check what
the actual C++ stdlib provider is and only set relevant macros rather than
all possibly-relevant ones based on the compiler.
Also, while we're here, now that's sorted, wire up the GCC experimental
libc++ support in the macro emission given it doesn't uglify anything for
libstdc++ users now.
Bug: https://github.com/mesonbuild/meson/issues/12962
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
We're going to use it in some more places in a minute (for controlling assertions).
Bug: https://github.com/mesonbuild/meson/issues/12962
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
We'll need it in a moment for get_base_compile_args -> get_assert_args.
Bug: https://github.com/mesonbuild/meson/issues/12962
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
Followup to 90098473d5.
If the compiler already has one of these assertion macros [0] set, then don't
interfere.
e.g. a Linux distribution might be setting a stricter default than usual. This
is a pitfall many fell into with _FORTIFY_SOURCE and it's why AX_ADD_FORTIFY_SOURCE [1]
was contributed to autoconf-archive.
[0] _GLIBCXX_ASSERTIONS, _LIBCPP_HARDENING_MODE, or _LIBCPP_ENABLE_ASSERTIONS.
[1] https://www.gnu.org/software/autoconf-archive/ax_add_fortify_source.html
Bug: https://github.com/mesonbuild/meson/issues/12962
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
Followup to 90098473d5.
I changed my mind on this a few times. libcxx's documentation describes [0]
the hardening modes as:
"""
1) Unchecked mode/none, which disables all hardening checks.
2) Fast mode, which contains a set of security-critical checks that can be done
with relatively little overhead in constant time and are intended to be used
in production. We recommend most projects adopt this.
3) Extensive mode, which contains all the checks from fast mode and some additional
checks for undefined behavior that incur relatively little overhead but aren’t
security-critical. Production builds requiring a broader set of checks than fast
mode should consider enabling extensive mode. The additional rigour impacts performance
more than fast mode: we recommend benchmarking to determine if that is acceptable
for your program.
4) Debug mode, which enables all the available checks in the library, including
internal assertions, some of which might be very expensive. This mode is
intended to be used for testing, not in production.
"""
We chose 3) before because it felt like a better fit for what we're trying
to do and the most equivalent option to libstdc++'s _GLIBCXX_ASSERTIONS, but on
reflection, maybe we're better off picking a default with less overhead and
more importantly guarantees constant time checks.
[0] https://libcxx.llvm.org/Hardening.html#using-hardening-modes
Bug: https://github.com/mesonbuild/meson/issues/12962
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
By adding the option name to UserOption object, it is now possible to
display the name of the affected option when the given option value is
not valid.
Fixes#12635
When rustc cannot run host binaries it does an early return which
skipped getting native-static-libs. Move that code earlier to always run
it.
While at it, failing to determine those libs is a fatal error. We would
crash later when trying to access rustc.native_static_libs attribute
otherwise.
Places where compiler needs it already have access to Environment object
and can use it directly.
This fixes mypy complaining that not all compilers have self.exe_wrapper
in run() method that got moved to base class.
When doing a compiler check with a File object, there is no need to dump
the file into logs, we can just print the path. The file could
potentially be big.
The previous approach wasn't great because you couldn't control what the
detected C++ stdlib impl was. We just had a preference list we tweaked the
searched order for per OS. That doesn't work great for e.g. Gentoo with libc++
or Gentoo Prefix on macOS where we might be using libstdc++ even though the host
is libc++.
Jonathan Wakely, the libstdc++ maintainer, gave a helpful answer on how
to best detect libc++ vs libstdc++ via macros on SO [0]. Implement it.
TL;DR: Use <version> from C++20 if we can, use <ciso646> otherwise. Check for
_LIBCPP_VERSION as libstdc++ doesn't always define a macro.
[0] https://stackoverflow.com/a/31658120
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
The cuda compiler also executes the host compiler and generally needs to know
the host compiler flags. We did this for regular args but not for error args.
We use `-Xcompiler=` since that's how nvcc tells the difference between args
it uses itself for device code and args it passes to the host compiler.
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
[eli: fix quoting style]
When only a single input file shows up in an arglist, it makes no sense
to inject `-W,--start-group -lone -Wl,--end-group`, since there is
nothing being grouped together. It's just longer command lines for
nothing.
This is just a bad regex match, because it matches *.so as a prospective
filename input even though it is actually not an input but a linker
flag. --start-group is only relevant to input files...
* `-Wpedantic` creates useless churn due to its use of gcc-line directives:
../foo.cu:1:3: warning: style of line directive is a GCC extension
1 | namespace Foo {
| ^~
https://stackoverflow.com/a/31001220
Currently, not only Meson lacks a way to induce a "--fatal-warnings" on
LINK.exe, it is also unable to pass flags appropriately when using
clang-cl or Microsoft's stock clang.
This commit fixes it by implementing `fatal_warnings()` in the
MSVCDynamicLinker and ClangCLDynamicLinker classes, and by implementing
the requisite conversion steps in linker_to_compiler_args for
ClangCompiler.
Since commit abc7e6af01 it is not possible
for this set of methods to return None, which was an odd thing to return
to begin with. Cease to annotate it as such.
Currently both print `Detecting linker via`, which can be confusing.
Clarifying this by printing "archiver" instead of "linker" for the
static linker/archiver
With LTO, the linker will re-execute the compiler and various warnings may
be emitted. We must therefore pass -Werror to the linker as well when -Werror
is enabled to capture these. This is only required / useful with LTO.
Closes: https://github.com/mesonbuild/meson/issues/7360
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
For Clang, we now pass -D_GLIBCXX_DEBUG=1 if debugstl is enabled, and
we also pass -D_LIBCPP_HARDENING_MODE=_LIBCPP_HARDENING_MODE_DEBUG.
Per https://discourse.llvm.org/t/building-a-program-with-d-libcpp-debug-1-against-a-libc-that-is-not-itself-built-with-that-define/59176/3,
we can't use _LIBCPP_DEBUG for older Clang versions as it's unreliable unless
libc++ was built with it.
We choose MODE_DEBUG for stldebug while building with assertions will do
MODE_EXTENSIVE.
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
None of the options set here affect ABI and are intended for detecting constraint
violations.
For GCC, we simply need to set -D_GLIBCXX_ASSERTIONS.
For Clang, the situation is far more complicated:
* LLVM 18 uses a 'hardened mode' (https://libcxx.llvm.org/Hardening.html).
There are several levels of severity available here. I've chosen
_LIBCPP_HARDENING_MODE_EXTENSIVE as the strongest-but-one. The strongest
one (_DEBUG) doesn't affect ABI still but is reserved for stldebug.
* LLVM 15 uses a similar approach to libstdc++ called '_LIBCPP_ENABLE_ASSERTIONS'
Note that LLVM 17 while in development had fully deprecated _LIBCPP_ENABLE_ASSERTIONS
in favour of hardened, but changed its mind last-minute: https://discourse.llvm.org/t/rfc-hardening-in-libc/73925/4.
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
This reverts commit 53ea59ad84.
This breaks at least:
- frameworks/17 mpi
- frameworks/30 scalapack
The problem is that openmpi's pkg-config emitted link arguments
includes:
```
-Wl,-rpath -Wl,/path/to/libdir
```
The deduplication logic in meson doesn't contain sufficient information
to tell when the compiler is passing an argument that requires values,
and definitely cannot tell when that argument is split across argv. But
for arguments that *can* do this, it is not possible to deduplicate a
single argument as standalone, because it is not standalone.
The argument for deduplicating rpath here was that if you have multiple
dependencies that all add the same rpath, the Apple ld64 emits a
non-fatal warning "duplicate -rpath ignored". Since this is non-fatal,
it's not a major issue. A major issue is when builds fatally error out
with:
```
FAILED: scalapack_c
cc -o scalapack_c scalapack_c.p/main.c.o -Wl,--as-needed -Wl,--no-undefined -Wl,--start-group /usr/lib64/libscalapack.so /usr/lib64/liblapack.so /usr/lib64/libblas.so -Wl,-rpath -Wl,/usr/lib64 -Wl,/usr/lib64 -Wl,--enable-new-dtags /usr/lib64/libmpi.so -Wl,--end-group
/usr/libexec/gcc/x86_64-pc-linux-gnu/ld: error: /usr/lib64: read: Is a directory
```
Tristan Partin
Xavier Claessens
Ben Corby
William D. Jones
L. E. Segovia
Will Toohey
Junior Rantila
Sam James
Charles Brunet
Petr Machacek
Elliot
David Seifert
Eli Schwartz
Jussi Pakkanen
dragonmux
Dylan Baker
U2FsdGVkX1
H. Vetinari
Alyssa Ross
Simon Ser
wrvsrx
Denis Drakhnia