Re: [rfa] generate symbols associated to namespaces

[David Carlton <carlton@kealia.com>]

On Tue, 9 Sep 2003 15:51:37 -0400, Elena Zannoni <ezannoni@redhat.com> said:
> David Carlton writes:

> Ok, but, could you check in the algorithm change (the nested_level
> stuff) for scan_partial_symbols w/o any namespace changes? So that
> we can do a cvs -D if something goes wrong? I have decided that it
> is best to have a change per 24hours period to a file, after
> spending a couple of days tracking down some other breakage in
> symfiles, a few weeks back.  Then the rest of the changes.

That should be fine.

>> +	  psym = add_psymbol_to_list (pdi->name, strlen (pdi->name),
>> +				      VAR_DOMAIN, LOC_STATIC,
>> +				      &objfile->static_psymbols,
>> +				      0, addr + baseaddr,
>> +				      cu_language, objfile);

> I don't like this too much. I.e. changing add_psymbol_to_list to
> return a value so you can get a hold of the psym. After all,
> add_psymbol_to_list doesn't change any of the fields of the psymbol,
> and cp_check_possible_namespace_symbols uses 2 fields that you
> already have around.

No, add_psymbol_to_list does change the fields: it calls
SYMBOL_SET_NAMES to set the demangled name, which I need in
cp_check_possible_namespace_symbols.  So having add_psymbol_to_list
return the psymbol seems the best solution; I suppose I could break
out the demangled-name-caching code that SYMBOL_SET_NAMES uses into a
separate function, but it doesn't seem worth it here.

>> +static char *
>> +add_partial_enumeration (struct partial_die_info *enum_pdi, char *info_ptr,
>> +			 struct objfile *objfile,
>> +			 const struct comp_unit_head *cu_header,
>> +			 const char *namespace)
>> +{
>> +  bfd *abfd = objfile->obfd;
>> +  struct partial_die_info pdi;
>> +
>> +  if (enum_pdi->name != NULL)
>> +    add_partial_symbol (enum_pdi, objfile, cu_header, namespace);
>> +  
>> +  while (1)
>> +    {
>> +      info_ptr = read_partial_die (&pdi, abfd, info_ptr, cu_header);
>> +      if (pdi.tag == 0)
>> +	break;
>> +      if (pdi.tag != DW_TAG_enumerator || pdi.name == NULL)
>> +	complaint (&symfile_complaints, "malformed enumerator DIE ignored");
>> +      else
>> +	add_partial_symbol (&pdi, objfile, cu_header, namespace);
>> +    }
>> +
>> +  return info_ptr;
>> +}
>> +

> Any enumeration related tests that fail before and pass after this
> change?

Nope.  I don't think it should change GDB's behavior on well-formed
debug info.

>> +  /* Add a symbol associated to this if we haven't seen the namespace
>> +     before.  */
>> +
>> +  if (dwarf2_extension (die) == NULL)

> Could you use this function in the check before this, in
> read_namespace (not in this patch, but already in the file)?

I think so; I don't see why the using directive bit couldn't be moved
into this block.

> I think it would make sense, in a next pass, to have all the creation
> of blocks go through the new function.

That would be my instinct as well.  I can't remember why I didn't do
that in this patch, but I don't think I had a good reason for why it
wouldn't work.

David Carlton
carlton@kealia.com