[Bug debug/54796] [4.8 Regression] Non-addressable stack parameter debug quality regression

jakub at gcc dot gnu.org gcc-bugzilla@gcc.gnu.org
Wed Oct 3 16:40:00 GMT 2012


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54796

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.8.0

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-10-03 16:40:04 UTC ---
I think the problem is that by the time these vt_canon_true_dep calls are made,
mloc's addr is some VALUE for which alias.c unfortunately no longer can figure
out that they are sp based, as sp REG has been removed from all the VALUE locs.
 If we could somehow somewhere preserve the information that some VALUEs are
sp based (i.e. find_base_term (val) ==
static_reg_base_value[STACK_POINTER_REGNUM]), then we could disambiguate at
least that the sp based stores can't clobber global vars or hard frame pointer
based MEMs.



More information about the Gcc-bugs mailing list