labath wrote: > > Ok, I see now. It's an ld.bfd vs ld.lld thing. You probably have your clang > > configured to use lld. LLD does not put relocation addends into the data > > section (on both arm and intel). ld.bfd does, which is why this sort of > > happens to work there. Was your intention to test with LLD? > > Yep, I run it with lld built together with clang and lldb. Should the > condition be narrowed to affect only builds with lld?
No, I think this is fine. I don't believe we have the ability to detect the linker used at the moment, and I'd like to avoid adding new dimensions to the test suite. Plus, the `PIE` flag captures the problem (that lldb depends on linker-specific relocation behavior, at least for ELF) these test expose fairly well. To fix this, we'd need to change ObjectFileELF::RelocateSection to relocate non-debug info sections as well. It may be worth linking these comments to a bug that provides more context. It's possible this could break some targets (like android) that do not support building/running non-PIE executables, but if that happens, we should just get those targets to set the PIE flag. https://github.com/llvm/llvm-project/pull/93808 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
