This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] RIP lang_hooks.tree_inlining.convert_parm_for_inlining
On Fri, 22 Jun 2007, Richard Guenther wrote:
> On Fri, 22 Jun 2007, Richard Guenther wrote:
>
> > On Fri, 22 Jun 2007, Joseph S. Myers wrote:
> >
> > > On Fri, 22 Jun 2007, Richard Guenther wrote:
> > >
> > > > * gimplify.c (gimplify_call_expr): Verify the call expression
> > > > arguments match the called function type signature. Otherwise
> > > > mark the call expression to be not considered for inlining
> > > > using TREE_STATIC flag.
> > > > * ipa-inline.c (cgraph_mark_inline): Honor TREE_STATIC on the
> > > > edges call expression.
> > >
> > > If this is a new use of TREE_STATIC, could you define a new checking macro
> > > (CALL_CANNOT_INLINE_P or similar) which expands to TREE_STATIC after
> > > checking the argument is a CALL_EXPR, rather than using TREE_STATIC
> > > directly, and update the comments in tree.h listing the uses of
> > > static_flag?
> >
> > Sure.
>
> Btw, I noticed we now fail to warn for gcc.dg/warn-1.c and
> gcc.dg/assign-warn-3.c. Both are old-style function declarations
> where the frontend doesn't warn, but we only warned if we inlined
> the functions because of the langhook. I don't consider this
> a problem. [We also mark them not for inlining because the frontend
> doesn't set TYPE_ARG_TYPES correctly, only DECL_ARGUMENTS are filled.
> If this is a problem I have to add another scanning loop variant to
> gimplify.c]
I was wrong, we still inline them but no longer warn.
> Shall I xfail the testcases or simply remove them?
>
> Thanks,
> Richard.