[PATCH] D98726: [analyzer] Enabling MallocChecker to take up after SmartPtrModelling

Thu Jun 3 23:37:29 PDT 2021

NoQ added a subscriber: vrnithinkumar.
NoQ added a comment.
Herald added a reviewer: teemperor.
Herald added a subscriber: manas.

Ok so if I was to debug this further, I'd set a breakpoint on `markInvalid()`, figure out which invocation corresponds to the right bug report, and that'd lead me to an explanation why this bug report was suppressed.

This could be an interesting investigation because it may uncover a bug in the suppression mechanism which causes us to have false negatives all over the place. False negatives are hard to notice because they only manifest as an absence of warning which isn't very in-your-face.

On the other hand, this specific issue will most likely be eliminated as soon as you replace `unique_ptr` methods' inlining with direct modeling.

---

Coming back to the patch in general, even though I had some ideas in D98726#2669315 <https://reviews.llvm.org/D98726#2669315>, I have a feeling that I'm forgetting something. @xazax.hun @vrnithinkumar, IIRC last year we added this FIXME not only because we wanted to find more bugs but also because we actually had some correctness problems here, do you remember what they were?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D98726/new/

https://reviews.llvm.org/D98726