jrtc27 wrote: > > When DynamicReloc::Kind was introduced, I was concerned of the many Kinds, > > but that was still better than the previous state. Thanks for the > > simplification. > > Yeah I think Alex was a bit confused (and also there were some bugs that he > faithfully replicated and since fixed, or that I'm fixing in the special case > of #150729). I've been delving into how all this works to clean up some > horrors in CHERI LLD (and many more in Morello LLD, where there is actually a > legitimate use case for AgainstSymbol(WithTargetVA) with something other than > R_ADDEND, due to ABI weirdness), and as part of that discovered the confusing > and overly-complex nature of all this. With the exception of the MIPS GOT > page (and Computed), the end state of this stack (see #150796, #150799, > #150797) is everything consistently gets funnelled through getRelocTargetVA. > I might also tackle that at some point, though I've already spent too much > time on MIPS the past few days, despite not even caring about it any more > downstream in CHERI LLVM...
Eh, I decided to do it in https://github.com/llvm/llvm-project/pull/150810 because then I could do https://github.com/llvm/llvm-project/pull/150813, which is the true sensible (IMO) end state. https://github.com/llvm/llvm-project/pull/150798 _______________________________________________ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
