Re: 2 items about fltk backend

[Shai Ayal]

On Wed, Dec 2, 2009 at 6:26 PM, Michael D. Godfrey
<address@hidden> wrote:
> I have discovered:
>
> 1. The problem of no data appearing when a plot command
>    such as plot([1:200]) was issued came back on my system.
>    I tried running glxgears to see if X11 was OK.  It ran. I
>    retried the Octave plot([1:200]) and it worked.  This has
>    happened several times more (while I was working on
>    problem 2 below) so it is definitely a reliable connection.
>    This means that something that Octave does to openGL and
>    X11 gets them confused.  Whatever glxgears does straightens
>    them out.  Ugh.
>
> 2. I tried with fltk backend set:
>    plot([1:200])
>    print  "test.pdf"
>
>    This fails do to gs not finding the /tmp/xxx.ps  file which
>    print.m should have written.  The reason for this is a timing
>    window in print.m.  The window does not seem to cause gnuplot
>    usage to fail, but it is still, I think, unsafe.  fltk takes longer
>    to produce the .ps, so it fails all the time.  My kludge fix was to
>   insert  at line 692  (approx)
>        while(!exist(name))
>        pause(5);
>      endwhile;
> The next line after this should be:
>      [errcode, output] = system (command);
>
> This fixes the problem, but is not a fix suitable for inclusion
> in the code.  There should be a way of determining if draw_now()
> has actually completed and generated its output.  Testing on this
> completion is what should go in place of my patch, I think.
What you report is strange since in the fltk_backend drawnow is not
done in a separate thread (like gnuplot), so it should end only after
the file does exist. I'll try to investigate later.

Shai