Debian Bug report logs - #444699
gnuplot: -persist takes inordinate amounts of cpu

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Matthijs Melchior <mmelchior@xs4all.nl>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: gnuplot: -persist takes inordinate amounts of cpu

Date: Sun, 30 Sep 2007 14:22:10 +0200

[Message part 1 (text/plain, inline)]

Package: gnuplot
Version: 4.2.0-4
Severity: normal


gnuplot takes an inordinate amount of cpu time when invoked with the
-perstsit option.  On the machine where the problem is detected,
one "gnuplot -persist" process takes 33% of the cpu, 2 "gnuplot -persist"
invocations make the system unresponsive with 100% cpu and a load of
~20.

strace shows the gnuplot process sits in a loop doing
  nanosleep({0, 10000}, NULL).

usecase:
  I have a script to plot NTP behavoiur, and it uses 'gnuplot -persist'
  to show a graph.  I use multiple invocations of 'gnuplot -perstis'
  to compare different days.

A generated gnuplot script and input file are attached.
To see the behavour give the following command, twice:
  gnuplot -persist /tmp/ntpgraph.debug

-- System Information:
Debian Release: lenny/sid
  APT prefers testing
  APT policy: (990, 'testing'), (500, 'oldstable'), (500, 'unstable'), (500, 'stable'), (1, 'experimental')
Architecture: i386 (i686)

Kernel: Linux 2.6.21-2-686 (SMP w/1 CPU core)
Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968)
Shell: /bin/sh linked to /bin/bash

Versions of packages gnuplot depends on:
ii  gnuplot-nox                   4.2.0-4    A command-line driven interactive 
ii  gnuplot-x11                   4.2.0-4    X11-terminal driver for gnuplot

gnuplot recommends no packages.

-- no debconf information

[ntpgraph.debug (text/plain, attachment)]

[loopstats.20070929 (text/plain, attachment)]

Message #10 received at 444699@bugs.debian.org (full text, mbox, reply):

From: sacrificial-spam-address@horizon.com

To: 444699@bugs.debian.org

Subject: Gnuplot wakes up 100x/second for no good reason.

Date: Sat, 29 Mar 2008 16:48:49 -0400

Package: gnuplot
Version: 4.2.2-1

After popping up a plot window (even if the window has been closed),
ust sitting there at the prompt, gnuplot is doing an infinite stream of
select(1, [0], NULL, NULL, {0, 10000})  = 0 (Timeout)

Give that it isn't doing anything ELSE, that should have a NULL final
parameter.  I confess I've been lazy and haven't tracked it down to a
particular line in the source code; unfortunately, running it under ltrace
kills it with a SIGTRAP.

This eats laptop batteries unnecessarily and should be fixed.

(Found via powertop, but I expect this is the same as the existing bug.)

Message #15 received at 444699@bugs.debian.org (full text, mbox, reply):

From: Bastien ROUCARIES <roucaries.bastien@gmail.com>

To: 444699@bugs.debian.org, 444701@bugs.debian.org, Debian bugs control server <control@bugs.debian.org>

Date: Mon, 13 Sep 2010 18:11:35 +0200

forcemerge 444699 444701
tag  444701 + upstream
forwarded 444701
https://sourceforge.net/tracker/?func=detail&aid=3065389&group_id=2055&atid=102055
thanks

Message #24 received at 444699@bugs.debian.org (full text, mbox, reply):

From: Bastien ROUCARIES <roucaries.bastien@gmail.com>

To: 444699@bugs.debian.org, Matthijs Melchior <mmelchior@xs4all.nl>, Nicolas Caval <nicolas.caval@wanadoo.fr>, Matt Blissett <matt+debian@blissett.me.uk>

Subject: [bug] gnuplot -persist bug

Date: Tue, 14 Sep 2010 09:59:28 +0200

tag 444699 + unreproducible
tag 444699 +  moreinfo
thank

I could not reproduce the bug.

From upstream:

>The loop in question deals with events from the window manager, so I think
>that in order to pursue this we would need to know what desktop and window
>manager is in use.  Note that there was a similar report that turned out to
>involve a bug in the Ion window manager:
>  http://www.mail-archive.com/ion-general@lists.berlios.de/msg01047.html
>
>Also this thread from 2006 might possibly be relevant
>  http://blog.gmane.org/gmane.comp.graphics.gnuplot.devel/month=20060201

Send a report that this bug log contains spam.