Re: [PATCH v5 2/3] [gdb/exp] Fix ignoring of incorrect namespace prefix

[Tom de Vries <tdevries@suse.de>]

On 5/1/26 9:44 PM, Tom Tromey wrote:
>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
> 
> Tom> +	      const char *current_scope = (block->function_block () != nullptr
> Tom> +					   ? block->scope ()
> Tom> +					   : nullptr /* Don't know.  */);
> 
> Calling block::scope would yield "" even for the static block case.
> So I was wondering why not use that rather than check function_block?
> But like I said elsewhere, I don't really follow this function that well.

AFAIU, you're proposing the following:
...
  	  if (len == 0)
  	    {
	      const char *current_scope = block->scope ();
	      directive_match = streq (scope, current_scope);
  	    }
...

That runs into the following regressions:
...
FAIL: gdb.cp/namespace.exp: print X
FAIL: gdb.cp/namespace.exp: print 'G::Xg'
FAIL: gdb.cp/namespace.exp: print G::Xg
FAIL: gdb.cp/namespace.exp: print G::XgX
...

The first can be reproduced like this:
...
$ gdb -q -batch outputs/gdb.cp/namespace/namespace \
     -ex "b marker2" -ex run \
     -ex "print X"
   ...
No symbol "X" in current context.
...

The relevant bit of test-case source is:
...
namespace
{
   int X = 9;
}

namespace C
{
   namespace D
   {
     void marker2 (void)
     {
        ...
     }
   }
}
...

So we're in cp_lookup_symbol_via_imports, with:
- scope == "C::D"
- name == "X",
- block == static block
and when looking at (AFAIU) implicit import:
...
(gdb) p *current
$3 = {import_src = 0x348dac0 "(anonymous namespace)",
       import_dest = 0x1b477a0 "", alias = 0x0,
       declaration = 0x0, next = 0x0, decl_line = 0,
       searched = 0, excludes = {0x0}}
...
we're trying to figure out directive_match.

In this case directive_match needs to be true for the test to pass, but 
we have scope == "C::D" and current_scope == "", so streq (scope, 
current_scope) == false.

So I've left this patch as is in a v6.

Thanks,
- Tom