Filter out DbgUsers that will not be sunk prior to sorting.

Nice find!

I think this looks mostly good, but it looks like it's missing a unittest for the DIBuilder function that you added.

In D95045#2522531, @TWeaver wrote:

I would argue this is minor enough to go in without one, but I'm happy to add one if you feel strongly in favour for a test going in too.

In D95045#2522488, @StephenTozer wrote:

This looks good to me - does this need a test case?

Added more context to diff

This looks good to me - does this need a test case?

In D82363#2506919, @StephenTozer wrote:

W.r.t the replacement of DBG_VALUE with DBG_VALUE_LIST, the current plan is to make that all happen in a patch following this stack. The patch itself should mostly be deletion of old code and simplifying where possible, renaming DBG_VALUE_LIST to DBG_VALUE, adding an extra operand to the new DBG_VALUE to represent directness, adding several tests, and updating existing tests as appropriate.

Paging @vsk -- it was a while ago, but do you still have the bitcode from D7406 handy, and would you have spare cycles to test it against this (combined patch in D94631)?

I know this change is old but I'd like to get it committed if possible. (Got waylaid by the pandemic) Or maybe abandon it and resubmit?
Anyway, are there any reviewers out there?

Ping - this one has had a lot of discussion, and I think all of the substantial concerns have been addressed; are there any further comments?

Ping.

Ping.

Ping - any further comments?

Thanks @rnk

Fixed issue in which DBG_VALUE_LIST expressions were being emitted without subregister masking.

In D94685#2504510, @atanasyan wrote:

@ntesic Do you have commit access?

@ntesic Do you have commit access?

Addressed comments:

• Use shouldUpdateCallSiteInfo() predicate
• Use ToErase->eraseFromParent()
• Reduce function attributes in test

some comments inline, lgtm otherwise

lgtm

How does it look now, with some renaming I did?

Rebased onto recent master (resolving clashes with the DW_OP_implicit_ptr patch); rewrote dbg-value-list-spill.mir test.

This fixes 32r6{,el}_defconfig and does not regress any of my other MIPS kernel builds.

LGTM

Patch fixes the reported bug; added test cases produces a crash before patch is applied. LGTM with minor nit.

In D78105#2487169, @nickdesaulniers wrote:

Compiler crash reported in: https://bugs.llvm.org/show_bug.cgi?id=48695

As @bjope points out on D91425, it should be mentioned that while solution will improve the debugging experience in a number of cases it isn't bulletproof. For example a pass could insert a non-debug instruction between the dbg.value+deref and the to-be-inlined call.

Address feedback from @bjope:
+ Change comment for dbg.value+derefs in ShouldRemove lambda.
+ Update the test to check that the backward scan removes adjacent duplicate dbg.value+derefs still.

Rebase