[Bug lto/44463] whopr does not work with weak functions

["hubicka at ucw dot cz" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44463

--- Comment #11 from Jan Hubicka <hubicka at ucw dot cz> 2011-01-10 13:20:36 UTC ---
> Yep.  It's a mess.
> 
> I'd really really like to have aliases be handled in a unified
> IPA symbol table entry facility before adding even more hacks.
> Thus we'd have proper resolution information for them (hopefully)

Me too, but not for 4.6.  I spent some time thinking about implementation
of the symbol table and it is clear it won't be easy project at all.
Starting from "simple" problems like that our current target machinery makes
it impossible by not understanding ASM names to more involved issues how
to represent calls to aliases and thunks nicelly in cgraph without all
the problems we have.

We will also need to extend toplevel ASM syntax for a way specifying what
symbols are defined in them.

I plan to branch as soon as I will have some rest from fixing all the current
LTO issues.

Anyway my plan for this PR is to move streaming of alias pairs into lto-cgraph
(it is de-facto part of cgraph anyway) so they are read early enough.  Then
after decl merging but before fixup we can take care to remove the
non-prevailing weak aliases from the pair list.

I will do that once the other alias change settle unless someone stops me.

Honza