We need to pass any generated sources down the CustomTarget
inititalizers so that they will generate a dependency correctly,
otherwise we get race conditions.
...Update 2, to be exact, since the Visual Studio linker only gained the
`/WHOLEARCHIVE:` feature since Visual Studio 2015 Update 2.
This checks whether we have the corresponding `cl.exe` which is
versioned at or after Visual Studio 2015 Update 2 before we try to apply
the `/WHOLEARCHIVE:xxx` linker flag. If we aren't, use built-in methods
in Meson to grab the object files of the dependent static lib's, along
with the objects that were `link_whole:`'ed into them, and feed this
list into the linker instead.
This would make `link_whole:` work on Visual Studio 2015 Update 1 and
earlier, including previous Visual Studio versions.
This partially reverts commit add502c648.
In 'linkshared' test, annotate cppfunc() as imported, so an indirection
through an import stub is generated, avoiding a relocation size error
when building using gcc for Cygwin with LTO on.
Align with the example of how to write this portably in [1].
The 'c' language part of that test already gets this right.
[1] http://gcc.gnu.org/wiki/Visibility
This is so dumb, we can just insert C for you without you having to know
that you're using C under the hood. This is nicer because:
1) Meson doesn't make the user add a language they're not explicitly
using
2) If there was ever an implementaiton of Vala that didn't use C as
it's assembly language, this wouldn't make any sense.
The build level shouldn't be deal with interpreter objects, by the time
they leave the intpreter they should be in the Meson middle layer
representaiton
We need to escape space in variables that gets into cflags or libs
because otherwise we cannot split compiler args when paths contains
spaces. But custom variables are unlikely to be path that gets used in
cflags/libs, and escaping them cause regression in GStreamer that use
space as separator in a list variable.