[PATCH] D75356: [Analyzer][StreamChecker] Introduction of stream error state handling.

Thu Mar 5 00:35:37 PST 2020

balazske marked 2 inline comments as done.
balazske added inline comments.

================
Comment at: clang/lib/StaticAnalyzer/Checkers/StreamChecker.cpp:37-40
+  // NoError: No error flag is set or stream is not open.
+  // EofError: EOF condition (feof returns true)
+  // OtherError: other (non-EOF) error (ferror returns true)
+  // AnyError: EofError or OtherError
----------------
balazske wrote:
> Szelethus wrote:
> > balazske wrote:
> > > Szelethus wrote:
> > > > These too. Also, I'm not yet sure whether we need `OtherError` and `AnyError`, as stated in a later inline.
> > > I plan to use `AnyError` for failing functions that can have **EOF** and other error as result. At a later `ferror` or `feof` call a new state split follows to differentiate the error (instead of having to add 3 new states after the function, for success, EOF error and other error). If other operation is called we know anyway that some error happened.
> > I think it would still be better to introduce them as we find uses for them :) Also, to stay in the scope of this patch, shouldn't we only introduce `FseekError`? If we did, we could make warning messages a lot clearer.
> This change is generally about introduction of the error handling, not specifically about `fseek`. Probably not `fseek` was the best choice, it could be any other function. Probably I can add another, or multiple ones, but then again the too-big-patch problem comes up. (If now the generic error states exist the diffs for adding more stream operations could be more simple by only adding new functions and not changing the `StreamState` at all.) (How is this related to warning messages?)
For now, the `EofError` and `OtherError` can be removed, in this change these are not used (according to how `fseek` will be done).

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75356/new/

https://reviews.llvm.org/D75356