Re: [rfa] annotate blocks with C++ namespace information

[David Carlton <carlton@math.stanford.edu>]

On Tue, 11 Mar 2003 16:23:13 -0500, Daniel Jacobowitz <drow@mvista.com> said:
> On Tue, Mar 11, 2003 at 01:14:16PM -0800, David Carlton wrote:

> In this case Stanford is not contributing the code; contribution is a
> copyright related action, and they never had the copyright.

Fair enough.

>> Yes; it should be easy enough to rewrite, say, find_last_component
>> using cp_find_first_component.  The only reason why I didn't do
>> that (other than laziness) was that then I'd want to go figure out
>> a situation where find_last_component actually gets called, to make
>> sure I didn't make a boneheaded mistake while doing so, and I
>> didn't feel like doing that.  But I'll definitely add a FIXME
>> comment about that?

> OK.  find_last_component gets called for dealing with stub methods, if
> I recall correctly why I wrote it.

Yes, that's right.  And all I know about them is that I just spent an
afternoon investigating a bug dealing with them, didn't succeed in
fixing the bug, and felt dirty afterwards.

>> Do any demanglers put in spaces after 'operator'?  I hope not...

> I thought one of them did, but I might have been mistaken.

You could easily be right: I haven't done a lot of GCC v2 checking.
I'll look into that, and, if so, add a guard for a possible extra
space in cp_find_first_component.  (Otherwise, the gdb_assert in
cp_find_first_component could leave people unable to load C++ files
with certain constructs mixing templates and operators.)  Not that
there's any justification for having the different demanglers
producing different output, but that's another issue entirely...

David Carlton
carlton@math.stanford.edu