[Bug target/66047] [5/6 Regression] vlc compilation failure with target attribute

["jakub at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66047

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |rejects-valid
           Priority|P3                          |P1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-05-07
                 CC|                            |hubicka at gcc dot gnu.org
   Target Milestone|---                         |5.2
     Ever confirmed|0                           |1

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Is callgraph still usable at the time of expansion?  If yes, perhaps we should
for the can_change_signature case if we're contemplating to switch on
sseregparm calling convention automatically walk all the callers and see if
they support it.
If not, bet somewhere around where we compute can_change_signature we'd need to
also compute target_can_change_signature bit that would be computed using some
target hook (perhaps with a shortcut if target node for both caller and callee
is the same, don't call the hook).