I'm not sure how to deal with missing env -u.

• We could do env CLANG_MODULE_CACHE_PATH= and change the compiler's interpretation of empty string for this variable. I'm not sure if the current behaviour (there will be no module cache in the cc1 at all) is intentional or useful. Hesitant to change this behaviour.

Awesome!

Thanks for the fast review! Good suggestions, will apply and land.

LGTM

Empty lines handling should also make it skipped, fix the 3 failures.

LGTM, thanks for the detailed explanation/comments. One minor nitpick inline.

Just noticed that 3 test cases failed. Please fix them before landing.

Hi, nice to see this getting in. Comments inline!

Nice!

Forgot to ask, is there anything preventing a testcase to be added here?

LGTM

This is pretty cool, I enjoy the idea of getting a tar out of a crash. I'm also a +1 for having this group of behaviors as a more official -femit-reproducer=<option> flag. In future work, do you plan to change the default crash mode to output a tar instead of multiple files?

Why isn't D23934 sufficient? What do these flags with very long names do that ffixed-date-time doesn't?

It would be nice to have some mechanism to notify developers that includes are still performed regardless of requires

LGTM

Nice cleanup, fixes and testcases, just pending test failures and this looks good.

LGTM

LGTM

Could the test be tagged with keywords that would prevent the linting to kick in? The amount of changes for this and following patches seems too big to be worth the reformatting.

The right solution to the paren problem is to add some notion of precedence (and associativity) to Nodes, but that's a larger change that would become simpler once the refactoring I'm doing is completed.

LGTM

LGTM!

Given context from further patches in the set, I believe ptrauth operand bundle has been a good way to represent this and works well. LGTM

This is good to go, sorry for the delay.

LGTM

Hi @modimo, can you also add a driver test to check (clang/test/Driver/...) that nothing is added to cc1 cmdline when -fno-new-infallible is used or take precendence?

Thanks for working on upstreaming this @ab. Overall looks good to me, I see clang-format issues, are those legit? One more comment inline.

Seems like the build failures are unrelated, LGTM with the suggested clang-format changes.

Nice catch, thanks for working on this!

Overall looks great but there are some test failures, looks like you also need to update line 223?

Ping

@dexonsmith and @jansvoboda11 thanks for the fast reply and the extra testing.

Assuming I'm answering the correct question that the returns_nonnull is preserved through a PCH, the answer is yes.

Thanks for working on this, comments inline. @vsapsai @jansvoboda11 @dexonsmith any headermap related concerns on your side?

Ping

Ping

Thanks for adding more tests! LGTM.

LGTM!

LGTM

Ping.

@yln sorry, I didn't get notifications for this somehow. This got fixed as part of the more general support in https://github.com/llvm/llvm-project/commit/819e0d105e84c6081cfcfa0e38fd257b6124553a

Overall looks good, few remaining nitpicks.

LGTM

LGTM. An additional review here would be nice though.

Thanks for adding the tests. LGTM after some remaining nitpick.

Update after last round of reviews.

Nothing else of substance from me, but I don't feel confident in giving final sign off on this. LG, but please wait until one of the other reviewers accepts as well.

Overall looks good, sounds like it's failing tests though?

LGTM

Sounds reasonable to me! Can you double check whether this attribute gets correctly serialized/deserialized in face of CXXNewExpr? An example of how to test that would be in clang/test/PCH/cxx-method.cpp.

Update patch after @aaron.ballman review!

Thanks for the review @aaron.ballman!

Make this work for c++14 as well, update testcase to cover all possibilities mentioned. @mdreseler thanks for bringing this up.

Looks good to me with the remaining clang-format linting on the assert.

Overall looks good, one more comment below.

You should also update the title with [NFC] or similar.

The approach looks good. There are some test failures - you need to update some bits in clangd and probably anywhere else in the repo that you see uses of getUsingDecl().

Update patch to reduce code dup as suggested by @dvyukov

Cool, thanks for the input!

Thank you @dvyukov for another round of reviews! A have more questions for you.

Cover more cases in atomic.ll.