Bug#598547: gnuplot: Gnuplot batch mode task stops when called from a backgrounded script

[Ron]

Hi again :)

On Fri, Apr 20, 2012 at 05:50:12AM +0930, Ron wrote:
> On Thu, Apr 19, 2012 at 09:57:09PM +0200, Anton Gladky wrote:
> > Hi Ron,
> > 
> > what is the reason of marking this bug as serious?
> 
> You can take your pick of the RC severities, but I don't think we should
> release with a bug that makes gnuplot useless from other scripts/applications
> that might call it.  Especially not if it's a regression of a bug that was
> already supposedly fixed upstream.
> 
> Do you happen to have a reference to the fix or the bug report there?
> I couldn't find it in the Changelog, or in their tracker so far.



Ok, so before anybody else also goes on a wild goose chase over this
latter part:
 
> I actually stumbled on this while trying to debug another issue, namely that
> gnuplot silently does nothing if called from another application.
> eg. if you do:
> 
>   system("gnuplot foo.script");
> 
> then gnuplot runs, and does exit, but doesn't actually appear to execute the
> script ...  I'm not sure yet if these two things are related (still digging
> into that) ...

This bit isn't a bug in gnuplot, it was me being silly and just not flushing
the script properly before calling gnuplot on it (and then being confused by
the odd behaviour that ensued, and the original bug of this report).

So ignore this part.  But I can definitely reproduce the original bug.
I presume it shouldn't be too hard to fix (again), gnuplot really has no
reason to be reading from stdin when it's writing to a non-interactive
'terminal' such as one of the image formats.  It just needs someone who
knows the code well enough to know which of the bazillion places that it
tries to do that need to be disabled in non-interactive modes.

If there was a ready-made -dbg package I'd have possibly narrowed that
down and sent a patch, but since it's easy to reproduce, I'll leave that
to someone who has a debug build already available :)

Cheers,
Ron