Revise hit counts when hit a false positive
Details
Diff Detail
- Repository
- rL LLVM
Event Timeline
Greg's out this week, he or Jim may want to comment.
I was curious about this method IncrementFalseAlarmsAndReviseHitCount, there weren't any calls to it. A little digging shows that the one use of it was removed in r166472.
What does it mean if you get a watchpoint exception on your machine, but you didn't have a watchpoint set on that address. Is there something else going on here?
Jason, in MIPS watch registers, last 3 bits of watchpoint addresses are masked by IRW(instruction, read, write types) flags,
so machine gives exception for any address having other bits same to the actual watch address execept last 3 bits.
Comment at line number 749 in below code gives an such example.
Zachary, TestMyFirstWatchpoint.py already covers this case.
Ahh, so the test was already failing and you're just fixing it? That's
fine then, thanks.