Re: [PATCH v2] [gdb/symtab] Fix DW_TAG_member regression

[Tom Tromey <tom@tromey.com>]

Tom> +      else if (cur_abbrev->tag == DW_TAG_member && has_const_value
Tom> +	       && has_external)

I'm a little surprised by this test.

I guess I would assume we'd see this same problem for a static class
variable even if it doesn't have an initializer.  And I'm not sure
has_external is needed?

I would have expected something like "has const value or has a location"
instead.  Could

Tom> +	{
Tom> +	  /* For Dwarf v4, GCC generates a DW_TAG_member for a static const
Tom> +	     member.  */

... but OTOH if this is the only case where DW_TAG_member is generated
then it seems fine.  Though I still don't understand the "external" bit.

So if this is correct could you add some explanation for it?

Tom> --- a/gdb/dwarf2/cooked-index-shard.c
Tom> +++ b/gdb/dwarf2/cooked-index-shard.c
Tom> @@ -65,6 +65,15 @@ cooked_index_shard::create (sect_offset die_offset,
Tom>    else if (tag_is_type (tag))
Tom>      flags |= IS_STATIC;
 
Tom> +  if (tag == DW_TAG_member)
Tom> +    {
Tom> +      /* A cooked index entry generated for a DW_TAG_member should be treated
Tom> +	 the same as one generated for a DW_TAG_variable.  Normalize to
Tom> +	 DW_TAG_variable, to simplify code dealing with cooked index
Tom> +	 entries.  */
Tom> +      tag = DW_TAG_variable;
Tom> +    }

I think this should be in the scanner and not here.

Tom