In D99250#2682598, @DavidSpickett wrote:

This change has caused a test failure in lldb. On our Arm bots: http://lab.llvm.org:8011/#/builders/96/builds/6582
Also on Linux x86_64: http://lab.llvm.org:8011/#/builders/68/builds/10277

Not windows though because test/API/commands/frame/language/TestGuessLanguage.py is an expected failure there already.

Is there a mapping in lldb that needs to be updated as well?

Good catch, thanks!

In D97656#2676095, @shchenz wrote:

Seems the functionality of the llvm-objcopy is not right. Could you please help to check? And maybe we need to add some other tests for the file format.

1:

$ llvm-objcopy 32.o copy_nodebug.o
$ dump -X32 -tv copy_nodebug.o      
...
dump: copy_nodebug.o: dump: 0654-103 Cannot read from the specified file.
[6]     m   0x00000000     undef     1  extern                    **Invalid Symbol Name**

Gentle ping.

Removed unnecessary checks.

Addressed comments.

If there are no more comments, I'll commit the patch soon.
What do you think? @aprantl

Addressed @shchenz's comments and replace Buffe with streams based on D91028.

Added const

Addressed Zheng's comments.
Thank you all for your comments.
Since no DWARF consumers has dialect-specific behavior, the patch will proceed.

Thank you for your review! @jhenderson and @Higuoxing.
Looking forward to more comments from @hubert.reinterpretcast and MaskRay

Addressed James's comments.

Thx! @aprantl The motivation of the patch came from the crash of tag name mismatching when using DBX under AIX. And modifying the debugger doesn't seem to make sense?

Only the .text, .data, .tdata, and STYP_DWARF sections have relocations.

Addressed jhenderson's comments.

Gently ping.

Gently ping.

Gently ping.

Addressed @jhenderson 's comments.

Update the parsing of address and flag of Section.

Updated the test case.

Thank you so much for your comments! @jhenderson They are very useful.

Addressed James's comments.

Thank you so much for reviewing this patch @jhenderson @MaskRay @hubert.reinterpretcast and sorry for the late reply. I was on vacation the last few weeks.
Here is the doc about the structure of XCOFF files: https://www.ibm.com/support/knowledgecenter/ssw_aix_72/filesreference/XCOFF.html
I am working on adding 64-bit support but the other tools this patch may depend on do not support 64-bit for now. All of this has to be done step by step.

1. Make more fields as possible optional.
2. Test the behaviours when the key is omitted.
3. Use the larger type for 32-bit and 64-bit.

In D94054#2479070, @nemanjai wrote:

It seems like this patch depends on another patch. Please don't forget to add the dependencies to the Phabricator reviews.

Addressed nemanjai's comments. :)

D93894 covers this patch.

D93894 covers this patch.

Leave the ISel of intrinsics to td.

Addressed nemanjai's comments.

Rebase on D89855.

Check if the register used can be forward.

Ping.

Replaced some checks by isInt<>().

Updated documentations.

In D92089#2441712, @steven.zhang wrote:

I see most of the benefit is from llvm/test/CodeGen/PowerPC/constants-i64.ll. Does the benefit come from the same pattern or several ? And you'd better visit all the f64 and compare the new algorithm with old to make sure that, there is no deg.