Add new add_project_[link]_args functions #1030
Conversation
thiblahute
force-pushed
the
project_args
branch
2 times, most recently
from
605ca1d
to
b05b83a
Compare
| if cc.get_id() == 'msvc' | ||
| add_global_arguments('/DGLOBAL_ARGUMENT', language: 'c') | ||
| add_project_arguments('/DPROJECT_OPTION', language: 'c') | ||
| add_project_arguments('/DPROJECT_OPTION_1', language: 'c') |
There was a problem hiding this comment.
This is not required. Just use -DGLOBAL_ARGUMENT etc. MSVC understands that.
There was a problem hiding this comment.
Well it was failling with MSVC and suspected this was the reason.
|
|
||
| cc = meson.get_compiler('c') | ||
| if cc.get_id() == 'msvc' | ||
| add_project_arguments('/DSUBPROJECT_OPTION', language: 'c') |
thiblahute
force-pushed
the
project_args
branch
2 times, most recently
from
f7f4460
to
7aa3ef5
Compare
|
Travis build failed because Appveyor build is stuck, and the Internet tells me that when this happens you're supposed to cancel the stuck build and kick off a new one. @jpakkane needs to do that. |
|
Please re-push to trigger a rebuild :) |
thiblahute
force-pushed
the
project_args
branch
from
7aa3ef5
to
8ef7cf0
Compare
Done.. let's see how it goes :) |
thiblahute
force-pushed
the
project_args
branch
2 times, most recently
from
7242baa
to
99e77c8
Compare
|
There should be a test that uses C++ and which checks that the project options set for C have not leaked into other languages. |
|
It would also be nice to test that target-specific c/cpp arguments override project arguments. |
| @@ -1944,6 +1945,7 @@ def generate_link(self, target, outfile, outname, obj_list, linker, extra_args=[ | |||
| commands = [] | |||
| if not isinstance(target, build.StaticLibrary): | |||
| commands += self.build.get_global_link_args(linker) | |||
| commands += self.build.get_project_link_args(linker, target.subproject) | |||
There was a problem hiding this comment.
At least -L arguments should be the other way around. The first one is used so for those project link args should be before global ones. But what about other linker arguments? I think that for other ones the last one is the deciding one. Maybe we should just stick to this order and fix things if they actually break in the wild?
There was a problem hiding this comment.
What we should be doing is gathering all the -L arguments from all sources into one OrderedDict and then generating the command from that. What you're pointing out is already a problem with our linker argument list since global link -L arguments override target-specific ones.
|
Args froze is not tested at all (I thought I had written a test for that but grepping says no). This needs a failing test that rougly does the following:
Apart from these two fixes this looks good. |
Done
Done |
thiblahute
force-pushed
the
project_args
branch
from
99e77c8
to
ac0fca0
Compare
thiblahute
force-pushed
the
project_args
branch
from
ac0fca0
to
af28abd
Compare
Fixes 979