http://bugs.winehq.org/show_bug.cgi?id=19931
--- Comment #21 from Paul Bolle pebolle@tiscali.nl 2009-10-28 07:37:00 --- 0) I think I have found the problem(s). There seem to be two issues here.
1) On Fedora /usr/bin/wine is a shell script that (basically) either calls /usr/bin/wine32 or /usr/bin/wine64. Somehow this confuses winedbg, at least it feels like winedbg behaves better if [...]/wine32 is copied to [...]/wine. Packaging issue, I'd guess.
2) Another issue is an apparent bug in elf_load_file(). I pinpointed the winedbg repeating crashes (with a lot of printfs) to: esm->fmap->sect[esm->sidx].mapped = ELF_NO_MAP; in elf_unmap_section().
The last steps in the call chain always looked like: elf_load_file() elf_unmap_file() elf_unmap_section()
It turns out elf_load_file() calls elf_unmap_file() if elf_map_file() failed. That is incorrect (and really should be incorrect, given the names of those two functions).
3) Trivial patch is attached.