Issue 87 in memory-sanitizer: False negative with stack slot reuse

[memory-s...@googlecode.com]

Status: New
Owner: ----
Labels: Type-Defect Priority-Medium

New issue 87 by euge...@chromium.org: False negative with stack slot reuse
https://code.google.com/p/memory-sanitizer/issues/detail?id=87

int main() {
for (int i = 0; i < 10000; ++i) {
int a;
int * volatile p = &a;
if (i < 9000)
*p = i;
else
return *p;
}
return 0;
}

Local variable "a" should be poisoned every time it goes into scope (on
llvm.lifetime.start?).


--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[memory-s...@googlecode.com]

Comment #1 on issue 87 by euge...@google.com: False negative with stack
slot reuse
https://code.google.com/p/memory-sanitizer/issues/detail?id=87

An unfinished patchset.
It fixes this issue, but runs into another - lifetime intrinsics break
debug location for the return instruction in some cases.

Attachments:
lifetime.patch 2.2 KB
lifetime-cfe.patch 717 bytes