https://sourceware.org/bugzilla/show_bug.cgi?id=32232
--- Comment #2 from Fangrui Song <i at maskray dot me> --- > /tmp/p/elfutils/out/debug/src/elflint --quiet --gnu-ld > /tmp/p/elfutils/out/lld/src/addr2line > section [ 6] '.gnu.version_r': entry 3 has invalid offset to next entry In `.gnu.version_r`, places all Vernaux entries after all Verneed entries . It seems that elflint expects that a Vernaux is immediately followed by its associated Vernaux entries. I think this is overly restrictive and should be dropped. > section [23] '.relro_padding' has wrong type: expected RELR, is NOBITS > section [23] '.relro_padding' has wrong flags: expected none and possibly > ALLOC, is WRITE|ALLOC .relro_padding has to be WRITE|ALLOC to be part of PT_GNU_TLS (also the first RW PT_LOAD). https://maskray.me/blog/2020-11-15-explain-gnu-linker-options#z-relro GNU ld uses DATA_SEGMENT_ALIGN/DATA_SEGMENT_RELRO_END and has size penalty. > section [11] '.rodata': merge flag set but entry size is zero lld keeps the SHF_MERGE flag in the output section to show that at least one input section has the SHF_MERGE flag. This is extra information, which helped me debug some weird linker scripts. -- You are receiving this mail because: You are on the CC list for the bug.
