--- Comment #8 from aldot at gcc dot gnu dot org 2008-04-17 12:03 ---
fixing the constraint and all is well. Thanks, Andrew!
*** This bug has been marked as a duplicate of 35102 ***
--
aldot at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #7 from aldot at gcc dot gnu dot org 2008-04-17 11:32 ---
Changing the attached testcase to use either =q or =R makes no difference, it
still tries to emit the wrong thing, thus reopening.
--
aldot at gcc dot gnu dot org changed:
What|Removed
--- Comment #6 from pinskia at gcc dot gnu dot org 2008-04-17 11:17 ---
(In reply to comment #4)
> Sounds a little bit like PR6142 popped up again via IRA.
If you read what the patch does there, it changes the constraints of the
internal patterns from "r" to "q" so this is not a bug :).
--- Comment #5 from pinskia at gcc dot gnu dot org 2008-04-17 11:16 ---
(In reply to comment #3)
> It compiles just fine with the "old" RA.
> If the "=r" is wrong, then shouldn't it be rejected also with -fno-ira?
Why, the old RA is choosing just happens to be the right one.
"=q" or "
--- Comment #4 from aldot at gcc dot gnu dot org 2008-04-17 11:13 ---
Sounds a little bit like PR6142 popped up again via IRA.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35963
--- Comment #2 from pinskia at gcc dot gnu dot org 2008-04-17 10:33 ---
I think the inline-asm is wrong.
asm volatile ("movb %%fs:%1,%0":"=r" (v):"m"(*(unsigned char *) addr)
);
"=r" is wrong here.
--
pinskia at gcc dot gnu dot org changed:
What
