[Clang][Sema] Diagnose indeterminately sequenced accesses

[rZhBoYao]

Summary

Add -Windeterminately-sequenced and fix -Wunsequenced according to P0400R0 and CWG2571.

Fixes #63935

[rZhBoYao]

Put -Windeterminately-sequenced under -Wsequence-point's control

[Endill]

Thank you for working on this!

clang/test/SemaCXX/warn-unsequenced.cpp

2–6

Checking for cxx17 prefix in C++23 mode is misleading. I suggest to rename is to since-cxx17, and have a new run line for C++23 mode, leaving check in C++17 mode intact.

[Endill]

I'm sorry for hand-waving, but I also remember @aaron.ballman saying that code related to -Wunsequenced or around it was missing checks for C++17.
I can't say whether you addressed that or not, so just leaving it here for you and reviewers.

[rZhBoYao]

Improve the test.

Not sure what the lack of C++17 checks was referring to in your conversation with Aaron.
Can @aaron.ballman confirm whether it is addressed by this patch?

[rZhBoYao]

BTW, I am not sure if CWG2571 is implemented by @cor3ntin? If so, can we mark it as done on https://clang.llvm.org/cxx_dr_status.html#2571? This patch handles the warning around it tho.

[Endill]

In D156057#4527787, @rZhBoYao wrote:

BTW, I am not sure if CWG2571 is implemented by @cor3ntin? If so, can we mark it as done on https://clang.llvm.org/cxx_dr_status.html#2571? This patch handles the warning around it tho.

As a reminder, we don't change cxx_dr_status.html directly. Instead, we add a special comment in clang/test/CXX/drs/dr25xx.cpp (for the case of CWG2571), which should be accompanied by test case(s) if possible. You can follow the pattern of existing tests. After that clang/www/make_cxx_dr_status script should be run to update the HTML.

[rZhBoYao]

Thanks for the reminder. I am aware of that. Browsing through 762672a73a1e and a560ccf2af7a, I believe the indeterminately sequenced requirement is met, neither of which test the codegen so I might just put the example shown in the release note in dr25xx.cpp.

In D156057#4527821, @Endill wrote:

In D156057#4527787, @rZhBoYao wrote:

BTW, I am not sure if CWG2571 is implemented by @cor3ntin? If so, can we mark it as done on https://clang.llvm.org/cxx_dr_status.html#2571? This patch handles the warning around it tho.

As a reminder, we don't change cxx_dr_status.html directly. Instead, we add a special comment in clang/test/CXX/drs/dr25xx.cpp (for the case of CWG2571), which should be accompanied by test case(s) if possible. You can follow the pattern of existing tests. After that clang/www/make_cxx_dr_status script should be run to update the HTML.

[rZhBoYao]

Refactor VisitCXXOperatorCallExpr and mark CWG2571 as done in Clang 15 (deliberately the same version P2128R6 was implemented) since this patch only fix the diagnostics around it.

[cor3ntin]

@aaron.ballman that's been waiting for a while - i don't feel comfortable reviewing it, mind looking at it?