Is language_lookup_primitive_type_by_name needed?

[Andrew Cagney <cagney at gnu dot org>]

If you look at language_lookup_primitive_type_by_name it appears to do a 
simple name -> type lookup based on the language.  Where name is 
something like "double", "float", ...

Looking at c-exp.y:yylex, though, it doesn't appear to be useful. The 
problem is that earlier code vis:

     case 6:
      if (strncmp (tokstart, "struct", 6) == 0)
        return STRUCT;
      if (strncmp (tokstart, "signed", 6) == 0)
        return SIGNED_KEYWORD;
      if (strncmp (tokstart, "sizeof", 6) == 0)
        return SIZEOF;
      if (strncmp (tokstart, "double", 6) == 0)
        return DOUBLE_KEYWORD;
      break;

will have already identified those types and exited the routine. 
language_lookup_primitive_type_by_name only being called for types it 
doesn't know about.

Was this doing something else that was later lost?

Is the same true for other languages?

- f-exp.y appears to use f77_keywords

- objc-exp.y looks like C

- p-exp.y might do something with it, but I can't remember enough pascal

- ada-exp.l, I'm not sure

ideas?

Deleting it greatly simplifies multi-arch.

Andrew