[PATCH] D147696: [Assignment Tracking][SROA] Fix fragment when slice size equals variable size

[Orlando Cazalet-Hyams via Phabricator via llvm-commits]

Orlando added inline comments.


================
Comment at: llvm/test/DebugInfo/Generic/assignment-tracking/sroa/var-sized-fragment.ll:21
+
+; CHECK: dbg.value(metadata i32 %.sroa.0.0.extract.trunc, metadata ![[#]], metadata !DIExpression())
+
----------------
jmorse wrote:
> Orlando wrote:
> > jmorse wrote:
> > > Does the upper 32 bits not get a dbg.value(poison) at least? Best to test for that not having an expression if it does get generated, as that's the purpose of this test.
> > Unfortunately the upper 32 bits (which are tracked with a dbg.declare) get incorrect debug-info. See https://github.com/llvm/llvm-project/issues/61981. This issue is independent of assignment tracking.
> > Unfortunately the upper 32 bits (which are tracked with a dbg.declare) get incorrect debug-info. See https://github.com/llvm/llvm-project/issues/61981. This issue is independent of assignment tracking.
> 
> Sure, but the fragment information for that is orthogonal surely, we can test for the wrong location but right fragment information, with a "FIXME" saying the location is wrong and a link to the bug? That'll be updated when the bug is fixed, and means we still achieve coverage of the upper bits of this alloca that's being split up either way.
SGTM, how does this look?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D147696/new/

https://reviews.llvm.org/D147696