[Bug ipa/83619] [8 Regression] ICE in inliner: caller edge count does not match BB count

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83619

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2018-01-03
     Ever confirmed|0                           |1

--- Comment #2 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Returning 1 from cgraph_edge::sreal_frequency is equivalent of what we always
did - when there is no profile we assume every call have same weight (and also
disable recursive inlining).

Counts should be initialized after the estimation. If not, we should get ICE in
the verifier we see here.  However profile estimation is local only, so if you
convert it to ipa (), then you get unintialized as we do not know the global
count of calls within the whole program (don't to whole program propagation).

I will take a look what is going on here.
Honza