------- Comment #15 from davek at gcc dot gnu dot org 2010-09-13 14:41 -------
(In reply to comment #14)
> Well, scans definitely pass on x86_64 AND i686 linux without -fpic.
>
> Why it fails for the -fpic targets should be clear from the assembly dumps.
>
> The fix you are referring to added "(%rip)" with and "?" operator, so it
> matches either previous string ("previously passed") or a new one.
Doh. Sorry Uros, it was late at night and my eyes were somewhat glazed over
and I missed that. The problem is entirely caused by the leading underscore
that symbols don't use on linux/ELF but do use on windows/COFF. (I was also
wrong about it previously passing, it just didn't fail before because it wasn't
there before, I didn't spot that your patch was within a day of the testcase
first being added.)
The fix is obvious (s/obj/_?obj/ in the match patterns) and I'll send a patch
to the list shortly.
--
davek at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|REOPENED |RESOLVED
Resolution| |FIXED
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45361
