http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51323
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Priority|P3 |P1
AssignedTo|unassigned at gcc dot |jakub at gcc dot gnu.org
|gnu.org |
Target Milestone|--- |4.6.3
Summary|g++ confuses this with |[4.6/4.7 Regression] g++
|function argument in |confuses this with function
|optimized call |argument in optimized call
--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-11-29
10:28:48 UTC ---
Actually, it fails on 4.6 vanilla branch too, just not in GCC 4.6-RH, because
we don't enable -fipa-sra by default for -O2/-Os for debug info quality reasons
in 4.6, only for -O3 (that is something that is solved in GCC 4.7).
The regression started with
http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=164135 which started building
MEM_REF in the IPA-SRA codepath.
As for your workaround, if you only do it for GCC 4.6.* and not also for 4.7
and later, it is fine, all I would like is that such workarounds don't stick
around forever for future versions of GCC.
