Closed olbat closed 6 years ago
But they are binding to the same value in linux and darwin (64 bits). From the CI message
Expected: "lib LibSome\nfun just_size_t : LibC::Int\nend"
got: "lib LibSome\nfun just_size_t : LibC::SizeT\nend"
It might be the case that the expectation is wrongly translated. The goal is to have that LibC::SizeT
in the output so the output is portable.
Disclaimer: I am not really familiar with crystal_lib though.
But they are binding to the same value in linux and darwin (64 bits).
Are you sure of that ? because on the CI this test is either failing on Linux or macOS ...
I think the problem on Linux is maybe related to this warning ... I'm trying to find a way to fix it.
I pushed a fix however I'm not sure it's the best way to do it.
Great work @olbat ...looking forward to this getting merged!
Ugh I guess you do need that fix 😕 ☝️ https://github.com/crystal-lang/crystal_lib/pull/52#discussion_r202545832
I tried to improve the fix a little bit ... Now I'm looking for the builtin headers' directory in directories of clang's default search path.
Does it look OK to you ?
It looks good to me, but i'd like a review from someone who's worked on this codebase before.
@olbat yes, this project could be based off clang.cr if needed.
@ysbaddaden okay, nice :)
I think it could be nice centralize the efforts if it's okay to add an external dependency to this lib ... If the maintainers (@bcardiff @RX14 ?) are ok with that, I'll try to find some time to take a look (in another PR ofc).
I'm unfamiliar with this codebase. Judging by the graphs, it's mainly @asterite and @ysbaddaden who should decide
Thanks for your efforts!
No problem, thank you for the help and the review ! :)
@olbat yes, this project could be based off clang.cr if needed.
see #56
The CI is failing because
SizeT
is not binded on the same type when running on Linux and MacOS.I guess the difference can be explained by the different libc (glibc/clang) ...
Any idea on how it should be fixed ?