rsundahl added a comment. In D143675#4160084 <https://reviews.llvm.org/D143675#4160084>, @vitalybuka wrote:
> Usually freezing signatures is not a big concern, we can agree to preserve > existing functions. > The stuff like ASanStackFrameLayout is the concern. compiler and runtime must > agree on data layout. The same for global. Yes. Presently, the address sanitizer code generator is very aware of the stack frame layout and how it will be poisoned up to and including the values to be stored in the shadow memory. I don't think that the runtime actually shares ASanStackFrameLayout at the moment but certainly would be supportive if under this proposed flag (where everything gets outlined), we actually passed some canonicalized/serialized ASanStackFrameLayout object to the runtime where it would be free to implement as appropriate. We're not proposing this kind of change at this time but may be worth doing in some future effort to separate instrumentation from implementation. (Similar arguments apply under this flag for globals). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D143675/new/ https://reviews.llvm.org/D143675 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
