martong added a comment.
Here is a smaller reproducer:
void bar(short k) {
++k; // k1 = k0 + 1
assert(k == 1); // k1 == 1 --> k0 == 0
(long)k << 16; // k0 + 1 << 16
}
And the explanation is the following. With this patch, when the analyzer
evaluates the `(long)k << 16` expression then it can properly deduce the value
of `k` being `1`. However, it was not possible in the baseline. Since we do not
model neither promotion nor explicit casts we get the false positive report.
This issue highlights the importance of the patch stack to implement the
modeling of casts (starting with https://reviews.llvm.org/D99797).
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D113753/new/
https://reviews.llvm.org/D113753
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
