Postpone expanding va_arg until pass_stdarg

[Tom de Vries]

On 10-02-15 14:46, Richard Biener wrote:
>> This patch is a way to achieve that gimplification doesn't call the actual
>> >gimplify_expr langhook, and it fixes the failure. But I'm guessing that's
>> >not the proper way to fix this.
> More like
>
> Index: gcc/tree.c
> ===================================================================
> --- gcc/tree.c  (revision 220578)
> +++ gcc/tree.c  (working copy)
> @@ -5815,6 +5815,7 @@ free_lang_data (void)
>        still be used indirectly via the get_alias_set langhook.  */
>     lang_hooks.dwarf_name = lhd_dwarf_name;
>     lang_hooks.decl_printable_name = gimple_decl_printable_name;
> +  lang_hooks.gimplify_expr = lhd_gimplify_expr;
>     /* We do not want the default decl_assembler_name implementation,
>        rather if we have fixed everything we want a wrapper around it
>        asserting that all non-local symbols already got their assembler
>

That worked, and allowed me to do a bootstrap on x86_64 for all languages and 
regtest for unix/ and unix/-m32 without any issues (other than scan-dump-tree 
failures for stdarg, since the va_list_gpr/fpr_size optimization is switched off).

That leaves just teaching the va_list_gpr/fpr_size optimization to recognize 
ifn_va_arg.

Thanks,
- Tom