Print current stack on CHECK violation to aid debugging and
match other sanitizers functionality.
Details
Details
Diff Detail
Diff Detail
- Repository
- rCRT Compiler Runtime
Event Timeline
Comment Actions
Could you reliably trigger a check failure, with a wild free for example?
I'm thinking a test would be nice.
This has potential for deadlocks. Slow unwinder may malloc AFAIK, which is not good if the CHECK failure is itself in malloc. On the other hand, fast unwinder would require frame pointers in internal msan and sanitizer_common functions, which we try to avoid.
I'm tempted to land this anyway and see if the deadlocks actually happen in practice or they are purely theoretical.
Comment Actions
Yes, that's how I tested it, with unaligned pointer (added the same test).
MSan does make an effort to block interceptors during unwind (see SymbolizerScope), so if the problem arises, we'll fix it.