NagyDonat wrote: Offtopic @Xazax-hun > there are many ways to get to undef during constant folding (signed overflows > ...
I was surprised to read this because as far as I know (knew) the analyzer models signed operations as if overflow was completely natural for them. (See e.g. the method `assumeInBound` which IIRC relies on this to convert a "within this interval" check to a ">= some threshold" check. I also vaguely recall that there are many FPs where the analyzer assumes wraparond of signed values.) I would be very grateful if somebody pointed out some logic in the analyzer source which ensures that a signed overflow produces an `UndefinedVal` (or the overflow is disallowed by some other mechanism). If I was right and we don't have such code yet, then it might be good to add some... https://github.com/llvm/llvm-project/pull/114222 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
