Implement -d inlines support for Mach-O.

[rjmansfield]

No description provided.

[EricRahm]

Overall lgtm, nice and straightforward. Thanks for including the test as well. Only small request is to add some details to the commit message and include the version of clang you used to generate the binary.

[rjmansfield]

Overall lgtm, nice and straightforward. Thanks for including the test as well. Only small request is to add some details to the commit message and include the version of clang you used to generate the binary.

Sounds good. FWIW I tried to make this into a yaml test instead of using a prebuilt binary but I think I've found a obj2yaml/yaml2obj bug writing out the dsym. Since I wasn't sure how long it'll take to fix, and how long before the test could rely on updated yaml2obj, I opted for using a binary instead. However if you want, I can look at the yaml2obj bug more and then update the test with a comment to point to the llvm bug report (if needed)

[EricRahm]

However if you want, I can look at the yaml2obj bug more and then update the test with a comment to point to the llvm bug report (if needed)

Having an upstream bug filed and referenced would be great.

[rjmansfield]

However if you want, I can look at the yaml2obj bug more and then update the test with a comment to point to the llvm bug report (if needed)

Having an upstream bug filed and referenced would be great.

I was able to use generate a minimal YAML test and work around the upstream bug. llvm/llvm-project#166993 for reference.

[EricRahm]

That test file is wild :) Given the complexity of the setup I think we're fine as-is, we can always come back and attempt to reduce it. Just a small request to more explicitly document the steps to generate it.

[EricRahm]

For completeness can we include the rough set of commands, I'm guessing vaguely:

# on macOS
# clang <some flags> tests/testdata/macho/test_inlines.c
# obj2yaml <some flags>
# manually change thing because bug xyz

[rjmansfield]

@EricRahm can you have another look at this when you have a chance?

[EricRahm]

@EricRahm can you have another look at this when you have a chance?

Thanks for the reminder! I put some time on my schedule tomorrow. I'd like to also loop in @haberman since this has potential user facing changes.

[haberman]

This looks good to me. The test file is a little bigger than we generally like, but maybe that's necessary to account for the large amount of DWARF data generated for inlines.

[rjmansfield]

This looks good to me. The test file is a little bigger than we generally like, but maybe that's necessary to account for the large amount of DWARF data generated for inlines.

OK, well I was able to trim it down a bit by ~27% by generating the yaml from the inline c test with -g1 without effecting the test behaviour.

[EricRahm]

Thanks @rjmansfield, lgtm!