[lto] PATCH: misc bug fixes

[Mark Mitchell]

Sandra Loosemore wrote:

> 2006-10-05  Sandra Loosemore <sandra@codesourcery.com>
> 
> 	* gcc/dwarf2out.c (lto_void_type_die): New.
> 	(AT_string_form): Always use DW_FORM_string when emitting LTO
> 	information, because the LTO front end doesn't (yet) understand
> 	DW_FORM_strp.
> 	(dwarf2out_init): Initialize lto_void_type_die.
> 	(lto_init_ref): Use debug_info_section_label instead of section name.
> 	(lto_type_ref): Make it work.
> 
> 	* gcc/lto/lto.c (lto_read_unspecified_type_DIE): New.
> 	(lto_read_DIE): Add entry for lto_read_unspecified_type_DIE.
> 	(lto_resolve_type_ref): Make sure new_context gets initialized.

> +   /* The LTO front end DWARF reader doesn't yet support DW_FORM_strp, so
> +      if we're emitting LTO information, always use DW_FORM_string.  */
> +   if (flag_unit_at_a_time)
> +     return node->form = DW_FORM_string;

I'm supportive of the hack of using DW_FORM_string in LTO mode (for now) 
but checking flag_unit_at_a_time is pretty hackish.  We need to add an 
-flto option (with a corresponding flag_lto flag) anyhow; we might as 
well do that now.  Kenny could use your hack in his local tree until 
that's done.  It should be easy to add the option; add it to common.opt, 
and invoke.texi.

Everything else in this patch is fine.

Thanks,

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713