Re: [PATCH][RFC] PR78035

2016-10-20 Thread Richard Biener
On Thu, 20 Oct 2016, Jan Hubicka wrote: > > > > The following tries to address the inconsistency with folding > > two addresses based on two decls vs. one pointer vs a decl address > > caused by the latter not honoring interposition. > > > > Honza, is > > > > + /* If obj1 is interposable g

Re: [PATCH][RFC] PR78035

2016-10-20 Thread Jan Hubicka
> > The following tries to address the inconsistency with folding > two addresses based on two decls vs. one pointer vs a decl address > caused by the latter not honoring interposition. > > Honza, is > > + /* If obj1 is interposable give up. */ > + if (VAR_P (obj1) > + && (TRE

[PATCH][RFC] PR78035

2016-10-20 Thread Richard Biener
The following tries to address the inconsistency with folding two addresses based on two decls vs. one pointer vs a decl address caused by the latter not honoring interposition. Honza, is + /* If obj1 is interposable give up. */ + if (VAR_P (obj1) + && (TREE_STATIC (obj1) || D