[PATCH] D75432: [analyzer][NFC][MallocChecker] Convert many parameters into CallEvent

[Balázs Kéri via Phabricator via cfe-commits]

balazske added inline comments.


================
Comment at: clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp:1194
-  if (!FD)
+  if (!Call.getOriginExpr())
     return;
 
----------------
Szelethus wrote:
> Szelethus wrote:
> > balazske wrote:
> > > Szelethus wrote:
> > > > Szelethus wrote:
> > > > > balazske wrote:
> > > > > > This should be added to avoid later crash (probably not needed for every check kind?):
> > > > > > ```
> > > > > >   const FunctionDecl *FD = C.getCalleeDecl(CE);
> > > > > >   if (!FD)
> > > > > >     return;
> > > > > > ```
> > > > > Not all `CallEvent`s have a corresponding `FunctionDecl` or a `CallExpr`, for instance, `CXXAllocatorCall` corresponds with `CXXNewExpr`, which is not a `CallExpr`, but it is handled by this checker. For this reason, I decided to move this check to the individual modeling functions.
> > > > Oh I'm sorry, do we have an actual crash resulting from this? 
> > > I did not look into it by detail but the problem is in `MallocChecker::checkOwnershipAttr` with a null `FD`. Probably it is enough to insert a return at that point (makes the crash gone on that analyzed project).
> > I'm sorry but I don't have enough to follow up on this. On the master branch, everything seems to work smoothly for me (all tests pass) and I didn't get any buildbot failures. Can you post a code snippet?
> If you have a project name, I can analyze that myself to save you the trouble.
Could not debug the problem because debug info was not there for unknown reason (no exact line locations were printed). But the project was emacs 26.3 `configure --with-xpm=no --with-gif=no --with-tiff=no --with-gnutls=no`, failed at one of the the first files with CodeChecker analyze.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75432/new/

https://reviews.llvm.org/D75432