Re: non-critical problem...

[Daniel J S. <dan...@ie...>]

Hans-Bernhard Broeker wrote:

>On Tue, 13 Apr 2004, Daniel J Sebald wrote:
>
>  
>
>>Rather than doing a full install, I built the latest tarball and decided 
>>to just test it within the "src" subdirectory.  I realize "you're not 
>>supposed to do that", but I ran across something a bit strange with the 
>>"gnuplot_x11" terminal.  
>>    
>>
>
>That's exactly the reason you're not supposed to do that.  It's explained
>in the INSTALL file (under "How to test gnuplot"), too.
>
>  
>
>>A couple things.  First, it might be nice if gnuplot would check the 
>>current active directory, i.e., the directory from which gnuplot is 
>>launched if it can't find gnuplot_x11 in the expected place.  
>>    
>>
>
>No way.  The security guys would immediately hang us by the toes for such
>a thing.  They're nervous enough about the setuid install if the libvga
>terminal is compiled in as it is.
>

OK, no argument there.

>>If nobody likes that idea, fine.  However, the second thing is that
>>after running gnuplot without "gnuplot_x11" present, the command line
>>hung.  
>>    
>>
>
>Which one: the gnuplot prompt or the shell, after you returned from
>gnuplot?
>
>  
>
>>CNTRL-C broke from the hang, 
>>    
>>
>
>Broke to where?
>

It broke back to the gnuplot prompt, i.e.,

gnuplot>

>>but afterward typing any keyboard keys no longer produces a response at
>>the command line.
>>    
>>
>
>Which command line? Did you try to clean up the terminal's state typing
>"stty sane" blindly?
>

It can't get back to the terminal.  It is stuck at "gnuplot>". 
 Actually, the sequence is stranger than I originally thought.  After 
the Exec failure, the screen will respond to keyboard hits but not 
process anything, i.e.,

Expected X11 driver: /usr/local/libexec/gnuplot/4.0/gnuplot_x11
Exec failed: No such file or directory
See 'help x11' for more details

asdfasdf

asd
asf


Then from there, CNTRL-C will bring back "gnuplot>", but the terminal is 
then unresponsive to any keyboard hits... except for a CNTRL-C which 
then displays another "gnuplot>" prompt as though CR was hit.

I get two process IDs showing up attributed to "gnuplot"

 1081 pts/4    00:00:00 gnuplot
 1082 pts/4    00:00:00 gnuplot <defunct>

You know, it's similar to what happens when "gnuplot_x11" gets killed 
externally and "gnuplot" relaunches a new "gnuplot_x11".  Could there be 
confusion in the code that handles relaunching processes?

>>That may be a bug.
>>    
>>
>
>Could be.  Ethan, I guess this one's in your ballbark --- I'm almost
>willing to bet it's caused by the mouse or font feedback stuff, so
>please see if you can do something about it.
>

I will write a bug report for it.  If the user comes across this 
situation, they have bigger problems to deal with.

>I seriously got to get some sleep now.
>

8:35 + 6:00 or 7:00 = screen hypnosis.

Dan