Handle zero-sized unwind fragment candidates #62931
Conversation
Every unwind fragment should be non-zero sized. If we consider splitting on a boundary that would create a zero-sized fragment, don't report that one. This only occurs when setting the stress mode JitSplitFunctionSize to something small.
ghost
added
the
area-CodeGen-coreclr
|
Tagging subscribers to this area: @JulieLeeMSFT Issue DetailsEvery unwind fragment should be non-zero sized. If we consider
|
| @@ -1693,7 +1693,12 @@ void UnwindFragmentInfo::Dump(int indent) | |||
| printf("%*sUnwindFragmentInfo #%d, @0x%08p, size:%d:\n", indent, "", ufiNum, dspPtr(this), sizeof(*this)); | |||
| printf("%*s uwiComp: 0x%08p\n", indent, "", dspPtr(uwiComp)); | |||
| printf("%*s ufiNext: 0x%08p\n", indent, "", dspPtr(ufiNext)); | |||
| printf("%*s ufiEmitLoc: 0x%08p\n", indent, "", dspPtr(ufiEmitLoc)); | |||
| printf("%*s ufiEmitLoc: 0x%08p ", indent, "", dspPtr(ufiEmitLoc)); | |||
There was a problem hiding this comment.
This was a useful change when I was debugging.
| @@ -2736,13 +2736,29 @@ void emitter::emitSplit(emitLocation* startLoc, | |||
| reportCandidate = false; | |||
| } | |||
|
|
|||
| // Don't report a zero-size candidate. This will only occur in a stress mode with JitSplitFunctionSize | |||
There was a problem hiding this comment.
Can we also add an assert in Alloc() for
assert(this->ufiEmitLoc->GetIG()->igSize > 0);There was a problem hiding this comment.
Which Alloc are you referring to?
There was a problem hiding this comment.
sorry, I meant Allocate() method, somewhere around here:
runtime/src/coreclr/jit/unwindarm.cpp
Line 1631 in d5f5950
There was a problem hiding this comment.
We can't have that assert, or any assert about the size of a particular IG, because the IG could be zero sized, as long as there are other IGs in the fragment that have code. That's what the endOffset > startOffset is asserting.
There was a problem hiding this comment.
Hhm, that's true. There could be other non-zero IGs in the fragment.
|
I actually forgot to ask you, but can you add a test case for this? |
ghost
locked as resolved and limited conversation to collaborators
Every unwind fragment should be non-zero sized. If we consider
splitting on a boundary that would create a zero-sized fragment,
don't report that one. This only occurs when setting the
stress mode JitSplitFunctionSize to something small.