[Bug tree-optimization/51039] [4.7 Regression] ICE: in estimate_function_body_sizes, at ipa-inline-analysis.c:1977 with -finline-small-functions -fno-ipa-pure-const and passing incompatible function ptr

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51039

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2011-11-09
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
   Target Milestone|---                         |4.7.0
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-11-09 
09:43:51 UTC ---
The code in ipa-inline-analysis.c that ICEs seems out-of-place.  It was
probably
meant to compensate for the lack of updates that I fixed with the other
patches.
In fact, it doesn't check for conservative consistency between edge and
stmt flag but asserts we optimistically keep the stmt flag correct which we
do not do during early inlining.

I'll just remove the offending code and see if something breaks -- iff, then
such check should be in the gimple IL verification code (also checking for
conservativeness, not optimality).  I'll check if such verifier code trips
over sth during bootstrap and otherwise will add it.