uweigand wrote: > So, like @nikic hinted at previously, it seems that with any other calling > convention than "C", the extension attribute is present either on both or > neither of the caller/callee arguemnts. > > Patch updated to check only calls/functions with the C calling convention and > remove the previous isInternal() heuristic. > > It would be possible to also check a call to an internal function to see that > the call and definition extensions match. However, when trying this with a > simple (fastcc) test case the caller will actually do the extension even if > there only is the extension attribute in the function definition. I suppose > this makes sense as they are assumed (per comment above) to exactly match, > and if the Callee has the attribute that would be the reasonable assumption > to follow.
This does make sense to me, yes. I assume the various calls to sanitizer helper routines still use the "C" calling convention, so those still have to be fixed before this can be enabled? https://github.com/llvm/llvm-project/pull/223755 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
