Re: [rfa] generate symbols associated to namespaces

[David Carlton <carlton@kealia.com>]

On Tue, 5 Aug 2003 14:08:39 -0400, Daniel Jacobowitz <drow@mvista.com> said:
> On Tue, Aug 05, 2003 at 11:06:28AM -0700, David Carlton wrote:
>> On Tue, 5 Aug 2003 13:54:29 -0400, Daniel Jacobowitz <drow@mvista.com> said:

>>> Is there any particular reason not to use the global block?  If so
>>> might want to comment what it is.

>> Yes: we never want these symbols to be found by a normal search of
>> all symtabs' global blocks, because we don't trust these symbols.
>> We only want them found as a last resort, once we've looked every
>> place that symbols associated to classes should live.  So that's
>> why I stash them in a local block.  I'll add a comment to that
>> effect.

> OK.  I would have thought that the fake symtab would never have been
> searched normally, though - I guess it gets chained into
> objfile-> symtabs?

Yeah, allocate_symtab does that automatically.

I'm a little tempted not to have a symtab there at all, instead just
having a special block (or even a special dictionary) that isn't
associated to a symtab.  That, however, runs into the issue of what to
do with the 'symtab' argument to lookup_symbol in that case; we could
set it to NULL, but in that case GDB might crash if some silly user
typed 'break NamespaceName'.  Or something.

David Carlton
carlton@kealia.com