[PATCH] D143334: [clang][Interp] Fix diagnosing uninitialized ctor record arrays

[Aaron Ballman via Phabricator via cfe-commits]

aaron.ballman accepted this revision.
aaron.ballman added a comment.
This revision is now accepted and ready to land.

LG with a request for an additional test.



================
Comment at: clang/lib/AST/Interp/Interp.cpp:390
 
-  if (isa<RecordType>(ElemType.getTypePtr())) {
+  if (ElemType->isRecordType()) {
     const Record *R = BasePtr.getElemRecord();
----------------
tbaeder wrote:
> shafik wrote:
> > aaron.ballman wrote:
> > > The difference between these two is that `isRecordType()` is looking at the canonical type whereas `isa<>` is looking at the type under inspection rather than the canonical type. I'd expect these to have the same behavior in most cases, but only matter for cases involving typedefs.
> > > 
> > > I think you're correct about the test case below not needing these particular changes -- at least, I'm not seeing what's changed that should impact the test. Should this be split into two changes? 1) Expose the test, 2) Make this functional change + add a new test where the canonical type is different to demonstrate the fix.
> > +1
> Can you come up with a small test case that would show the difference? You mentioned typedefs, but if the array is of a typedef type, the old `isa<>` version doesn't work either.
Ah, I see what you mean, yeah. With a typedef type, the `isa` would have also failed. That's a good test case to add to this patch as well.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D143334/new/

https://reviews.llvm.org/D143334