Kristianerik wrote:

> > Thanks for the guidance. After reviewing CompilerInstanceTest.cpp and 
> > tracing all callers of addRemappedFile(StringRef, MemoryBuffer*) across 
> > clang/lib/, clang/tools/, and clang-tools-extra/, every call site passes a 
> > buffer constructed via getMemBuffer() or getMemBufferCopy() null is 
> > impossible by construction. The fuzzer was generating an invalid state that 
> > can't occur through any legitimate code path. I'll close this PR.
> 
> If you had time, maybe you could take a look at #201179? This one can be 
> triggered directly on godbolt and I am not sure if it shares the same reason 
> with this one.

Sorry I was a bit busy this weekend, but I took a look at #201179 and it 
appears to be a different issue. #201188 was a fuzzer-only state that couldn't 
occur in practice, while #201179 has a real reproducer on godbolt and looks 
like a regression introduced between clang 20 and 21, possibly related to the 
ModuleManager in-memory buffer changes in that timeframe. I'll investigate 
further and follow up on that issue directly.

https://github.com/llvm/llvm-project/pull/201289
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to