http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51986
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |ASSIGNED
Last reconfirmed| |2012-01-25
AssignedTo|unassigned at gcc dot |jakub at gcc dot gnu.org
|gnu.org |
Ever Confirmed|0 |1
--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-01-25
14:33:39 UTC ---
Created attachment 26463
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26463
gcc47-pr51986.patch
I think it isn't a false positive, at least from compiler's POV.
The compiler can't know that PATTERN (insn) is always non-NULL.
I think it is always non-NULL, thus I'm proposing just removing the IMHO
unneeded check, if it was actually needed (but then it would need to be
scheduler internal, I think most of RTL code assumes everywhere that an insn
has non-NULL PATTERN and PATTERN shouldn't be used on non-insns like notes or
barriers), we'd want to swap the pat == 0 check with initialization of *rev.
