Re: debug_on_error and plotting

[Rik]

On 11/10/2011 01:35 AM, Michael Goffioul wrote:
> On Thu, Nov 10, 2011 at 1:57 AM, Rik <address@hidden> wrote:
>> I just fixed the problem in this changeset
>> (http://hg.savannah.gnu.org/hgweb/octave/rev/b4b8e525dee0).
> I had a quick look through gdb and detected that __myhandle__ is
> causing the problem (is there any other?). It comes from the fact that
> this is not a "real" property, in the sense that it does not inherit
> from base_property, so it cannot be returned by get_property(), though
> it is returned by get(true). This is a kind of weird situation, and
> IMO what's returned by get() should also be return-able by
> get_property().
>
> What about turning the __myhandle__ property into a handle_property
> object? Would that cause any other issue? I think it would be a
> cleaner solution than what's currently being done in
> remove_all_listeners() method.
Michael,

I found the same thing when debugging the issue, but went with the less
dramatic fix of just extending the existing unwind protect frame.

I don't mess with the graphics code much, but it appears another solution
would be to have get (true).map_value filter out __myhandle__.  Or
__myhandle__ could be filtered out explicitly in the for loop before
calling get_property.  These are just alternatives and I don't have a
recommendation or leaning since, as stated, I don't spend much time with
the graphics code.

--Rik