This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Re: dwarf2out change breaks ada and libstdc++ onsolaris too
On Tue, 3 Feb 2004 14:39:25 -0800, Devang Patel <dpatel@apple.com> wrote:
> On Feb 3, 2004, at 2:23 PM, Jason Merrill wrote:
>
>> On Tue, 3 Feb 2004 09:44:24 -0800, Devang Patel <dpatel@apple.com> wrote:
>>
>>> It seems I broke all targets that do not supply target specific
>>> routine to build builtin va_list. These targets use standard
>>> builtin va_list that is just void *.
>>
>> It's not clear to me why that's a problem.
>>> /* For TYPE_DECL, lookup TREE_TYPE. */
>>> if (TREE_CODE (decl) == TYPE_DECL)
>>> ! {
>>> ! tree v = root_type (TREE_TYPE (decl));
>>> ! if (v && TREE_CODE (v) == VOID_TYPE)
>>> ! /* Can not force die VOID_TYPE. */
>>> ! return;
>>> ! at_import_die = force_type_die (TREE_TYPE (decl));
>>> ! }
>>
>> Why is this necessary? We can cope with void* in other contexts.
>
> typedef __builtin_va_list __gnuc_va_list;
> typedef __gnuc_va_list va_list;
> namespace std
> {
> using ::va_list;
> }
>
> base_type_die() and add_type_attribute() do not generate DIE for
> VOID_TYPE. In other contexts we just don't generate DIE. Here we
> need DIE for DW_TAG_imported_decl. Which means force_type_die() fails.
You're missing my point. Didn't you just say va_list is void*? That's
different from plain void.
In any case, it's wrong to ignore the TYPE_DECL. The decl is what we're
interested in. We should be importing the appropriate typedef die.
Jason