ilovepi wrote: > Moved metadata emission from Clang to `ModuleSymbolTable`. Now other tools > can integrate `EmitModuleFlags` to set symbol table metadata. With the new > function we don't really need to change interface of `CollectAsmSymbols` and > `CollectAsmSymvers`. Without `CPU` and `Features` parameters, these functions > still have to use a "default" asm parser as they used to. Addressed code > style issues and improvements suggested in the comments. > > RISC-V issue #67698 is indeed fixed (added a LIT test), but I don't think > that any of #50591, #65090, #69780 are fixed. I don't have an environment to > reproduce them quickly. This patch is only about symbols defined in inline > assembly, so miscompilation issues or inconsistent flags for the actual > assembler are likely related to something else (`riscv-isa` metadata from > #80760 perhaps?).
Thanks for looking. I don't think its a problem that some of the other issues aren't addressed by this PR. They all interact to some extent, which is why I mentioned them. For at least one of them, I think handling this bit correctly is a prerequisite for it to be addressed. We can re-evaluate their status once this is finished and determine where the remaining gaps are (if any). https://github.com/llvm/llvm-project/pull/174995 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
