On Sat, Aug 08, 2015 at 10:58:15AM +0200, Kai Wasserbäch wrote: > And there *IS* a difference vs. your output: for you the relocations in > 794488_elfs/libelf1/dump.elf.J4EnbO look fine, for me the second relocation is > botched with libelf1 while it works with libelfg0. > > libelf1: > relocations: 2 > 0: 10, SCRATCH_RSRC_DWORD1 > 1: 200000081, > > libelfg0: > relocations: 2 > 0: 10, SCRATCH_RSRC_DWORD1 > 1: 2c, SCRATCH_RSRC_DWORD0
Awesome. That should explain why the application of that relocation crashes and burns. Odd I couldn't replicate locally against elfutils libelf 0.163. It might be some subtle compiler code generation issue. Or maybe debian applies a patch that isn't upstream? Yep! https://sources.debian.net/src/elfutils/0.163-4/debian/patches/0003-Add-mips-n64-relocation-format-hack.patch/?hl=34#L34 Note how that replaces the cast and sizeof Elf64_Rel with Elf64_Rela in the memcpy. Those are not the same size! Could someone rebuild the debian package without that patch applied (or correctly replace the wrong Elf64_Rela with Elf64_rel) and see if that helps? Thanks, Mark -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
