[Bug gold/23258] Linker cored dump at process_one_opcode

[ccoutant at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=23258

--- Comment #11 from Cary Coutant <ccoutant at gmail dot com> ---
(In reply to Lan from comment #10)
> Finally managed to get back to this issue.  (I had reverted to use clang
> 3.8, so needed to reinstall clang 6.0 to try to reproduce the problem.)
> 
> After changing GseAdmTempFile() to replace tempnam() with mkstemp(), I no
> longer hit the trap in the loader.
> 
> So the trap is related to the warning message.

Yes, in producing the warning, the linker tries to parse the debug info (if
present) in order to give source file name and line number. If that fails, it
should fall back gracefully to the object file name.


> I can fix all the warning to work around the trap in the loader, but it is
> still a defect that the loader shouldn't trap, right ?

Yes, I'd like to find and fix it, but with the objects you provided, I wasn't
able to reproduce the problem. It seems like it should be a problem with the
debug info in GseAdEnableDB.o, but my version of gold is able to parse the
debug info with no problem. It's possible that the segfault is triggered by a
combination of factors, and without the same libraries that you have, it's not
hitting the same conditions.

If you add the -t option to the link, the linker will list all the files it
reads. If you can then package all of those up, I'll give it another try.

-- 
You are receiving this mail because:
You are on the CC list for the bug.