Rebased on latest master;
Addressed the comments;

Addressed the comments;
Rebased the patch on latest master;

Add pre-committing tests with the incorrect behaviour as part of the NFC patch.

Hi, you commit failed clang-ppc64le-rhel buildbot: http://lab.llvm.org:8011/#/builders/57/builds/2938. Please take a look.

• used memcpy to burn AlignPackInfo into uint32_t and vice versa;
• named XL on AIX stack effect and related options after "xl" instead of "aix"
• adjusted natural align related code

Rebase on renaming NFC patch

LGTM. Thanks!

Split testcases;
Simplified code;
Don't need to respect attribute align within typedef when the alignment value is less than bitcontainer size;

Adjusted the logic of operator == for class AlignPackInfo;
Removed mac68k sentinel;

As we discussed offline, the strategy to handle long long type bitfield under 32bit mode would be the bitcontainer shall be 32-bits unless if a single long long bitfield has length greater than 32 bits (what would have been a diagnostic in xlc/xlclang/xlclang++), then the bitcontainer shall be 64-bits.

Just a minor suggestion: Maybe Enabling return and argument vector types for AIX is more accurate. Otherwise, LGTM.

Update the logic to tell if two AlignPackInfo are equal;

Addressed the comments;
Handle PCH for AIX;

LGTM. Thanks!

Addressed comments;

Addressed the comments;

Update testcases;

Remove the if condition and tweak the comments;

Fix wrong sinit/sterm function formatting issue;

Fix minor issues in the testcase;

Add testcases to both CodeGen and CodeGenCXX folder;

In D90892#2386908, @aaron.ballman wrote:

I think this generally seems reasonable, but I'm far from an AIX expert so you should wait a few days in case other reviewers have feedback.

Address comments;
Move testcases to CodeGen folder instead of CodeGenCXX since ctor/dtor attribute should work in both C&C++ mode;

Remove unused function parameters;
Move testcases to CodeGenCXX folder;

Fixed the duplication in the testcase;

Manually close the revision after landing it upstream.

Add one case to the test;

Fix the formatting issues;

Fixed the testcase formatting issues;
Fixed the BIGGEST_ALIGNMENT value in testcases;

Edit the definition of SuitableAlign;

Hi @jyknight , are you okay with us changing the "definition" of SuitableAlign without sending a note to the mailing list?

I am wondering can we split the option related changes to a separate patch for reviews? That would make current patch a bit easier to review and faster to be committed as two small pieces.

Hi Zarko, the patch is out of sync with master branch, could you update it?

ping.

Use regex to match AIX layout dumping format.

In D89064#2320133, @hubert.reinterpretcast wrote:

Can we use a regex to make this also work in AIX?

Remove emit errors for oversized long long bitfield and related testcase;