From: Craig D. <def...@bo...> - 2013-02-13 23:02:37
|
Hmmm, it seems that the INT signal doesn't make gnuplot do what I want it to do for the pipe case. In particular, if I'm sending binary data over a pipe and issue an INT, it doesn't make gnuplot start accepting new commands over the pipe. As near as I can tell, gnuplot continues to read from the pipe regardless. So INTing seems to work fine over a tty but not over a pipe control link. On Feb 12, 2013, at 3:37 PM, Craig DeForest <def...@bo...> wrote: >> >> I am dubious that you would see slow communications between your perl script and gnuplot just because DISPLAY is remote. The actual x11 rendering is done by a separate process, gnuplot_X11, and no matter how slowly it may be progressing gnuplot itself should be able to accept new input on stdin. > > Sorry if I wasn't clear here -- the communication is still fast of course, but the render can take a long time if X11 communications is slow. Gnuplot doesn't issue any indication of state until the whole process completes. Same goes for wxt, of course. > > >> >> - Slow rendering jobs can be interrupted by the user, and there's no documented way (that I've found) to let gnuplot know to stop what it's doing and give a prompt back. >> >> In "normal" (i.e. interactive) mode you can get a prompt back by hitting ^C >> (INT signal). You should be able to achieve the same by sending INT. I've just tested that this works from the command line using "kill -INT" >> >> That doesn't specifically address your question about binary data being sent on stdin. I'm not certain what state you will be left in if there is a binary read pending when you send an INT signal. >> >> Ethan > > Thanks. Early experiments with "kill -INT" hosed up the pipe, but I'll run another round and see what happens. > > |