Hi, When I wrote the following mail, I totally had forgotten where I initially read that the 100% CPU bug wasn't completely fixed in 1.4.18.
So here's the mail I sent just a few hours ago to liferea-de...@lists.sourceforge.net. ----- Forwarded message from Axel Beckert <a...@deuxchevaux.org> ----- From: Axel Beckert <a...@deuxchevaux.org> Date: Tue, 24 Mar 2009 13:08:00 +0100 Subject: [Liferea-devel] 100% CPU issue(s) revisited To: liferea-de...@lists.sourceforge.net Hi, I'm a big fan of Liferea, using it since 1.0.27 from Debian Etch and with over 700 feeds. (I love the "search all feeds feature", it's my personal Google News. ;-) I read in the Liferea blog that the 100% CPU issue is or at least was a well known problem. Last year in July 1.4.18 was announced and was said to fix the 100% CPU issue[1]. I also have the feeling that it got better since then, but for me there is still some random 100% CPU usage issues left, as also reported in the comments to this [1] http://liferea.blogspot.com/2008/07/fix-for-100-cpu-usage-problem.html 1.4.18 is also the version in Debian Lenny, but occasionally I still have 100% CPU usage. It starts suddenly after hours or sometimes first after a day of usage or so. My workaround was to restart Liferea until it happened again. In case it was at 100% CPU usage, the window closed, then an empty X window of the same size pops up and vanishes after a few to maybe 15 seconds. Once or twice I had to kill that ghost window and the liferea-bin process, since it didn't vanish after minutes. I'm involved in the development of Conkeror[2] (don't ask about the name[3]), a XULRunner based web browser with Emacs look and feel. With conkeror, we also had a strange more or less random 100% CPU issue which mostly showed up on Debian, probably since Debian has disabled JEMALLOC in XULRunner by default. [2] http://conkeror.org/, http://packages.debian.org/conkeror [3] http://conkeror.org/FAQ#Wheredoesthename.22Conkeror.22comefrom.3F Using gdb, Nick, one of our developers, found out[4] that this was caused by the idle timeout for the "frecency" update ([5], places.frecency.updateIdleTime). Setting this to 0, e.g. via about:config or startup files, solved the issue. [4] http://repo.or.cz/w/conkeror.git?a=commitdiff;h=c92814d3c8b72f70e7a0b3fe1708d7c8c275977b [5] http://kb.mozillazine.org/Places.frecency.updateIdleTime Nick also says that this variable isn't properly initialized if the XULRunner application doesn't care about it (in comparison to Firefox et al.). This would explain why it appears at random Conkeror uptimes (and Liferea uptimes :-), because it could have any, also very small values on startup which means the 100% CPU usage can start quite early even when only showing about:blank. I also added this setting to my productive 1.4.18 Liferea on Debian Lenny via about:config and haven't had the above described 100% CPU usage problem anymore since then. Liferea runs smooth for about a week without any problems. Haven't used 1.4.26 from Debian Sid that much yet, so I can't say anything about more recent versions. Hope this helps someone -- if not even the Liferea developers :-) Regards, Axel -- [Signature, SF ads and Mailman footer moved to /dev/null] ----- End forwarded message ----- Since it was also a nearly Debian-only issue with Conkeror due to Debian's XULRunner has JEMALLOC disabled by default, this is probably also the reason why Liferea has these problems mostly on Debian (and perhaps also the reason why upstream hasn't noticed them as we did). Regards, Axel -- Axel Beckert - a...@deuxchevaux.org, a...@noone.org - http://noone.org/abe/ -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
