[PATCH] D132236: [analyzer] Fix liveness of LazyCompoundVals

[Artem Dergachev via Phabricator via cfe-commits]

NoQ added a comment.

So to be clear, I think your solution is probably an ok stop-gap solution. False negatives aren't that bad, I'm more worried that existing leak reports may become less understandable because they'll be reported later than necessary, which may obscure the reason why we think they're finally leaking. But I really want us to agree on how this facility is supposed to work, and explore the perfect solution, before implementing an imperfect solution.

Maybe we should introduce "weak region roots" that aren't necessarily live themselves but anything derived from them (any `SymbolRegionValue<R'>` or `SymbolDerived<R', $x>` for arbitrary sub-region `R'` of a weak-region-root `R`) is live. This could be pretty straightforward.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D132236/new/

https://reviews.llvm.org/D132236