Octaviz segfaults

[Jonathan Stickel-5]

On 5/5/08 octave-maintainers-request@... wrote:

> Date: Mon, 5 May 2008 14:00:03 -0500
> From: " Jordi Guti?rrez Hermoso " <jordigh@...>
> Subject: Octaviz segfaults
> To: "Octave Maintainers List" <octave-maintainers@...>
> Message-ID:
> <9543b3a40805051200y2e1723e5w1b0a3e346afd315@...>
> Content-Type: text/plain; charset="iso-8859-1"
>
> Octaviz segfaults on the attached sample script and data.
>
> I'd simplify the test case, but I've honestly have been unable to
> figure out what's so particular about this data to make it segfault.
> I've tried other similar approaches without getting a segfault.
>
> Also attached you will find a stack trace.
>
> Now, I know that Octaviz isn't really part of Octave proper, but I
> really have no idea where else to ask for help. Dragan Tubic, the
> former lead developer of Octaviz, has abandoned Octaviz.  :-(  It's also
> hard to tell, as you can see from the tangled web of a stack trace, to
> know if the segfault is Octave's, Octaviz's or VTK's fault. I'm
> leaning towards Octaviz, as a segfault in the other two would probably
> be noticed by members of their respective communities.
>
> I'm not asking for a complete solution... I'm happy to dig through the
> code myself and see what's wrong, and in fact, I've already tried. I
> am just asking for a little bit of help. My current guess is that
> Octaviz is somehow not keeping a proper reference count (#21 in the
> stack trace below looks like a likely candidate to look for problems).
>
> Maybe I'm also asking if it's worthwhile rescuing Octaviz. I do think
> that VTK produces some of the prettiest graphics all around, but so
> much effort is now spent on other graphics backends for Octave, that
> perhaps it's ok to let Octaviz die.
>
> Thanks,
> - Jordi G. H.

Jordi

Octaviz isn't dead!  It just never reached a critical level of interest
where development is consistently ongoing.  Why do you think Dragan has
abandoned the project?  I know that he is rarely involved now, but he
usually responds if prompted.  Anyway, there are a couple of us who
still try to keep Octaviz working, although there has not been any
recent development of new features and functions.  You can ask about
Octaviz specific issues on the Octaviz forum or mailing list (from the
sourceforge project page).  I should probably update the web page
pointing to those.

About your specific problem, I do not get a segfault on my system (Mac
OS X 10.4, octave-3.0.0, octaviz-0.4.7).  I can send you the two
resulting png files offlist if you like.  Does Octaviz work for you
otherwise?  For example, does vtk_sombsurf work?

Jonathan

[Jordi Gutiérrez Hermoso-2]

On 05/05/2008, Jonathan Stickel <jjstickel@...> wrote:
>  About your specific problem, I do not get a segfault on my system (Mac OS X
> 10.4, octave-3.0.0, octaviz-0.4.7).

Same versions here. I'm thinking it may be a 64bit issue. Are you
using a 32 bit OS?

- Jordi G. H.

[Jonathan Stickel-5]

Jordi Gutiérrez Hermoso wrote:

> On 05/05/2008, Jonathan Stickel <jjstickel@...> wrote:
>  
>>  About your specific problem, I do not get a segfault on my system (Mac OS X
>> 10.4, octave-3.0.0, octaviz-0.4.7).
>>    
>
> Same versions here. I'm thinking it may be a 64bit issue. Are you
> using a 32 bit OS?
>
> - Jordi G. H.
>
>
>  

I'm not sure.  I think the "kernel" is 64 bit, but most everything else
is 32 bit, including vtk, octave, and octaviz.  I'll be out of town for
a couple days without email.  I'll try to look in more detail when I get
back.

Jonathan

[Jordi Gutiérrez Hermoso-2]

On 05/05/2008, Jonathan Stickel <jjstickel@...> wrote:
> Why do you think Dragan has abandoned the project?

---------- Forwarded message ----------
From: Dragan Tubic <dragan.tubic.1@...>
Date: 5 May 2008 07:42
Subject: Re: Octaviz segfault
To: Jordi Gutiérrez Hermoso <jordigh@...>

[snip]

 Unfortunately I don't work actively on octaviz anymore.

[snip]

:-(

On 05/05/2008, Jonathan Stickel <jjstickel@...> wrote:

> Jordi Gutiérrez Hermoso wrote:
>
> > On 05/05/2008, Jonathan Stickel <jjstickel@...> wrote:
> >
> >
> > >  About your specific problem, I do not get a segfault on my system (Mac
> OS X
> > > 10.4, octave-3.0.0, octaviz-0.4.7).
> > >
> > >
> >
> > Same versions here. I'm thinking it may be a 64bit issue. Are you
> > using a 32 bit OS?
> >
>  I'm not sure.

Well, how about this. What's the sizeof a pointer for you? On my
system, the following C++ code outputs "8". What does it output on
yours?

     #include <iostream>

     int main(){
       using namespace std;
       cout << sizeof(int*) << endl;
     }

If it's a problem with 64 bit misalignment of pointers, that would be
a good clue to go bug squashing.

Also, looking a bit more closely at VTK's code, I'm starting to think
that I wouldn't put a segfault beyond them. Manually namespacing all
of their functions instead of using C++ namespaces? #including
obsolete C++ headers? Certainly not the cleanest C++ I've seen. The
problem may well lie within VTK proper, not even Octaviz.

Thanks,
- Jordi G. H.

[Thomas Weber-8]

On 05/05/08 15:44 -0500, Jordi Gutiérrez Hermoso wrote:
> On 05/05/2008, Jonathan Stickel <jjstickel@...> wrote:
> >  About your specific problem, I do not get a segfault on my system (Mac OS X
> > 10.4, octave-3.0.0, octaviz-0.4.7).
>
> Same versions here. I'm thinking it may be a 64bit issue. Are you
> using a 32 bit OS?

Okay, I can confirm this on 64 bit, but I'm currently busy. Please open
a bug report against Octaviz in Debian, severity normal (so that I don't
forget to look into this).

        Thomas

[Jordi Gutiérrez Hermoso-2]

On 09/05/2008, Thomas Weber <thomas.weber.mail@...> wrote:
> Okay, I can confirm this on 64 bit,

Nice. Now that we know it's a 64 bit problem, we have a good clue on
where to go bughunting.

I do hope the bug is in Octaviz and not in VTK, because I'd really
hate to go bughunting in the second one.

I'll file a bug report in Debian. Looks like this package is ours now
that Dragan orphaned it.

- Jordi G. H.