NoQ added a comment.
I think there's still this problem i've outlined in the comment above: you can
step into an integer-underflow if your analysis begins with unlock(). You could
just ignore all unlocks that move you below 0, which would be ok.
Could you add this test?
std::mutex m;
void foo() {
m.unlock(); // MutexCount = 4294967295, should be 0, just ignore this
unlock.
sleep(1); // no-warning
m.lock(); // MutexCount = 0, should be 1, woohoo we're sure we're in the
section now.
// What's the current mutex count?
sleep(1); // expected-warning{{}}
}
Repository:
rL LLVM
https://reviews.llvm.org/D21506
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
