[PATCH] D78374: [Analyzer][StreamChecker] Added evaluation of fread and fwrite.

[Balogh, Ádám via Phabricator via cfe-commits]

baloghadamsoftware added a comment.

> Is there a significant gain behind doing that instead of doing a state split immediately? It feels unnatural to delay the split. It doesn't accurately represent how the code would run. Are we supposed to do this at every precall? What about other checkers that may rely on this one?

The actual gain is that we somewhat reduce the exponential growth of the exploded graph. With such a Schrödinger's cat pattern we delay it to the point where it is actually checked, it checked at all.



================
Comment at: clang/lib/StaticAnalyzer/Checkers/StreamChecker.cpp:39
+  /// There is an execution path with none of the error flags set.
+  bool NoError = true;
+  /// There is an execution path with EOF indicator set.
----------------
`bool NoError:1 = true` etc. Why not use bit fields for a struct of boolans?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D78374/new/

https://reviews.llvm.org/D78374