Re: Segmentation Fault with Octave MXE under Windows

[Daniel J Sebald]

On 05/29/2014 12:22 PM, Daniel J Sebald wrote:
> On 05/29/2014 12:14 PM, John W. Eaton wrote:
[snip]
> > It is supposed to be initialized properly in every constructor, never 0.
> > If that's not the case, can you point to a constructor where it is not
> > initialized?
>
> The test would indicate that "rep" is being used out of scope. The
> problem might lie elsewhere. I'm looking at the function
> __magick_read__.cc at the moment for any "delete" called twice or
> something similar. Array<> is used in __magick_read__.cc, but I'm not
> seeing anything odd just yet.

The destructor could be a red herring--simply the place where a problem 
shows up.  This __magick_read__.cc is somewhat hard to follow.  It is 
doing some tricky stuff disposing of image frames (multiple I suppose, 
in the case of GIF).  One question I have is how "disposalmethod" option 
is assured to have nFrames entries:

  const Array<std::string> disposalmethod = options.getfield 
("disposalmethod").cellstr_value ();
  for (octave_idx_type i = 0; i < nFrames; i++)
    {
      imvec[i].quality (quality);
      imvec[i].animationDelay (delaytime(i));
      imvec[i].gifDisposeMethod (disposal_methods[disposalmethod(i)]);
    }

Dan