Re: Help using the GDB C++ STL pretty-printers / xmethods

[Hannes Domani <ssbssa@yahoo.de>]

 Am Sonntag, 8. Mai 2022, 22:26:37 MESZ hat Paul Smith <paul@mad-scientist.net> Folgendes geschrieben:

> On Sun, 2022-05-08 at 15:44 -0400, Paul Smith wrote:
> > On Sun, 2022-05-08 at 09:16 +0100, Jonathan Wakely wrote:
> > > > gdb.set_convenience_variable('mgr', val['mgr'])
> > > > init = gdb.parse_and_eval('$mgr->initialized')
> > > >
> > > > This will use the xmethod to evaluate the expression.
> > >
> > > And then:
> > >
> > > if init:
> > >     return gdb.parse_and_eval('*$mgr')
> >
> > Unfortunately, this doesn't work :(.  I can't do it from the GDB
> > command line or python (I have tried both with the same results).
> > Something about convenience variables doesn't play well with xmethods
> > (or maybe this xmethod implementation specifically?)
>
> Hm.  I have done a fair amount of work writing GDB Python convenience
> functions and commands, but I've only done a small amount of pretty-
> printer work and no xmethod implementations.
>
> But I don't understand this from the C++ STL xmethods.py:
>
>   class UniquePtrGetWorker(gdb.xmethod.XMethodWorker):
>     ...
>       def __call__(self, obj):
>           impl_type = obj.dereference().type.fields()[0].type.tag
>
> Why are we using dereference() here?  Aren't we expecting to receive a
> gdb.Value of type std::unique_ptr here, not _pointer to_
>
> std::unique_ptr?
>
>
> But, it definitely doesn't work to remove the dereference(), and also
> the value we get normally IS a pointer; adding debugging above I see:
>
>   type: 'std::unique_ptr<Foo, std::default_delete<Foo> > *'
>
> I don't really grok xmethods so I'm not sure how calling "foo->val"
> when "foo" is std::unique_ptr<> results in the __call__ method being
> invoked with a gdb.Value of type std::unique_ptr<>*.
>
> My suspicion (not based on anything) is that whatever the reason is, is
> why I can't use these xmethods with a convenience variable.

This is documented in the XMethod API, see XMethodWorker.__call__ in [1]:

The first element is always the this pointer value.


Why don't you just set the convenience variable to the address?:

(gdb) set $xx=&foo
(gdb) p (*$xx)->val
$7 = 0


Hannes