Currently C++ inherits C, which can lead to diamond problems. By pulling
the code out into a standalone mixin class that the C, C++, ObjC, and
Objc++ compilers can inherit and override as necessary we remove one
source of diamonding. I've chosen to split this out into it's own file
as the CLikeCompiler class is over 1000 lines by itself. This also
breaks the VisualStudio derived classes inheriting from each other, to
avoid the same C -> CPP inheritance problems. This is all one giant
patch because there just isn't a clean way to separate this.
I've done the same for Fortran since it effectively inherits the
CCompiler (I say effectively because was it actually did was gross
beyond explanation), it's probably not correct, but it seems to work for
now. There really is a lot of layering violation going on in the
Compilers, and a really good scrubbing would do this code a lot of good.
When the '-Wl,-rpath-link' option refers to several folders, we can
either use one single entry, like this:
-Wl,-rpath-link,/path/to/folder1:/path/to/folder2:/path/to/folder3
...or we can use multiple entries, like this:
-Wl,-rpath-link,/path/to/folder1
-Wl,-rpath-link,/path/to/folder2
-Wl,-rpath-link,/path/to/folder3
Because the '-rpath-link' requires full folder paths, using the one
single entry option can result in a very long argument.
While this shouldn't be a problem, at least *one* toolchain (the latest
version of the Tensilica toolchain for Xtensa processors) segfaults when
using arguments that are too long.
Because other toolchains might be affected and because using multiple
entries instead of a very long one doesn't seem to have any drawback,
this patch changes the arguments building logic to use multiple
'-Wl,-rpath-link' entries.
I recall that @jpakkane never wanted this, but @nirbheek did, but then
@nirbheek changed his mind.
I am fine either way except for the cross inconsistency that exists
today: There is no `c_preproc_args` or similar one can put in the cross
file, so no way to replicate the effect of CPPFLAGS during cross
compilation.
OpenBSD does not link libc into shared libraries by default to avoid binding
libraries to specific libc majors. So prevent passing "-Wl,--no-undefined" or
"-zdefs" when linking (those options have always suffered false positives on
this operating system).
Before, the logic initialization compiler options from environment
variables vs config files was strewn about. Now, it is consolidated. We
leverage the new `envconfig.py` module to expose the configuration data
to `compilers.py` without creating an import cycle.
This does two things:
* On windows GCC-like compilers, the subsystem is always explicitly
specified (either -mwindows or -mconsole). MSVC is already explicit.
* The gui_app linker flags are now added after those mandated by
external dependencies. This is because some misguided libraries (such
as SDL) think that hijacking `main()` and forcing `-mwindows` in link
flags is clever. We must unconditionally override such misuses to let
gui_app work as intended.
Instead use coredata.compiler_options.<machine>. This brings the cross
and native code paths closer together, since both now use that.
Command line options are interpreted just as before, for backwards
compatibility. This does introduce some funny conditionals. In the
future, I'd like to change the interpretation of command line options so
- The logic is cross-agnostic, i.e. there are no conditions affected by
`is_cross_build()`.
- Compiler args for both the build and host machines can always be
controlled by the command line.
- Compiler args for both machines can always be controlled separately.
This allows each implementation (gnu-like) and msvc to be implemented in
their respective classes rather than through an if tree in the CCompiler
class. This is cleaner abstraction and allows us to clean up the Fortran
compiler, which was calling CCompiler bound methods without an instance.
ICC doesn't use the same -fprofile-generate/-fprofile-use that GCC and
Clang use, instead it has -prof-gen and -prof-use. I've gone ahead and
added the threadsafe option to -prof-gen, as meson currently doesn't
have a way to specify that level of granularity and GCC and Clang's
profiles are threadsafe.
Because we need to inherit them in some cases, and python's
keyword-or-positional arguments make this really painful, especially
with inheritance. They do this in two ways:
1) If you want to intercept the arguments you need to check for both a
keyword and a positional argument, because you could get either. Then
you need to make sure that you only pass one of those down to the
next layer.
2) After you do that, if the layer below you decides to do the same
thing, but uses the other form (you used keyword by the lower level
uses positional or vice versa), then you'll get a TypeError since two
layers down got the argument as both a positional and a keyword.
All of this is bad. Fortunately python 3.x provides a mechanism to solve
this, keyword only arguments. These arguments cannot be based
positionally, the interpreter will give us an error in that case.
I have made a best effort to do this correctly, and I've verified it
with GCC, Clang, ICC, and MSVC, but there are other compilers like Arm
and Elbrus that I don't have access to.
Dylan Baker
Daniel Mensinger
Michael Hirsch, Ph.D
Fernando Ramos
makise-homura
Jon Turney
Jussi Pakkanen
John Ericson
Nirbheek Chauhan
Antoine Jacoutot
John Ericson
jml1795
Andrei Alexeyev
Beau Johnston
Martin Storsjö
Phillip Cao