| DBO | any trunk testers present for Unity? | 00:59 |
|---|---|---|
| DBO | I need a second opinion | 00:59 |
| GiulioC | hi all | 07:34 |
| GiulioC | can i ask a question about the latest update of unity (and unity-2d)? | 07:34 |
| davidcalle | jcastro_, ping | 08:05 |
| andyrock | hi all | 11:17 |
| htorque | andyrock: hi! i'm sure you can help me with bug 829961 - what would i need to do to get some useful valgrind output? | 11:24 |
| ubot5 | Launchpad bug 829961 in indicator-application (Ubuntu) "indicator-applications leaks memory" [Undecided,New] https://launchpad.net/bugs/829961 | 11:24 |
| andyrock | htorque, hi | 11:44 |
| andyrock | htorque, G_SLICE=always-malloc G_DEBUG=gc-friendly valgrind -v --tool=memcheck --leak-check=full --num-callers=40 --log-file=valgrind.log compiz --replace | 11:45 |
| andyrock | and attach valgrind.log | 11:46 |
| htorque | can't i just write a wrapper for the indicator to check and restart that one? | 11:46 |
| htorque | and: do i need to let it run for some time or is it done after finishing the startup? | 11:47 |
| andyrock | mhhh... you should "valgrind" unity-panel-service | 11:48 |
| andyrock | not the entire compiz | 11:49 |
| andyrock | so you know unity-panel-service is a dbus service | 11:49 |
| htorque | yeah... indicator-datetime is a small thing, indicator-applications definitely should be looked at, and unity-panel-service goes haywire here. while idling it's fine, but when using it there are some weird jumps adding lots of ram to the process. | 11:50 |
| htorque | yes | 11:50 |
| andyrock | so go here /usr/share/dbus-1/services | 11:50 |
| andyrock | now you should comment (with a #) the Exec line in unity panel service file | 11:50 |
| andyrock | i don't remember the right name of the file | 11:51 |
| andyrock | now you can killall unity-panel-services | 11:51 |
| htorque | com.canonical.Unity.Panel.Service.service ;-) | 11:51 |
| andyrock | yep :) | 11:51 |
| andyrock | btw you can killall unity-panel-services and run the command | 11:52 |
| htorque | yeah, and then i'd start it with valgrind - when that's done can i kill it or will it finish? | 11:52 |
| andyrock | G_DEBUG=gc-friendly valgrind -v --tool=memcheck --leak-check=full --num-callers=40 --log-file=valgrind.log unity-panel-service | 11:52 |
| andyrock | you can kill it... | 11:52 |
| andyrock | or just logout and login | 11:52 |
| andyrock | using this method you should get useful information about the leaks... | 11:53 |
| htorque | and for those processes that leak over time, should i let valgrind run for a while or doesn't that matter? | 11:53 |
| andyrock | use it for a while... | 11:53 |
| andyrock | try to do something with the indicators | 11:54 |
| andyrock | such as opening menu, scrolling | 11:54 |
| andyrock | etc. | 11:54 |
| andyrock | 4-5 min should be enough | 11:54 |
| htorque | with indicator-applications and -datetime i actually don't need to do anything - mem consumption raises all the time :-( | 11:54 |
| andyrock | so wait a pair of minutes than kill unity-panel-service or just logout/login | 11:55 |
| htorque | unity-panel-service can be upped by opening and closing the network indicator (i think that's still called nm-applet), but i've seen leaks without doing that (just normal desktop usage)... | 11:55 |
| andyrock | ok... update your bug description too :) | 11:57 |
| htorque | sure, will do | 11:57 |
| htorque | thanks for your help :-) | 11:57 |
| andyrock | i will be online for a while... if you need help ping me ;) | 11:57 |
| htorque | here's an updated plot: http://img.xrmb2.net/images/731666.png - indicator-datetime and -application show the same issue after a unity restart, so at least that's reproducible here. :) | 12:00 |
| Omega | Hmm, I found a very weird bug | 12:00 |
| Omega | you cannot use media buttons in some applications while a right clight menu is open | 12:01 |
| Omega | and in others (like firefox) it dismisses the right click menu | 12:01 |
| Omega | click* | 12:01 |
| htorque | Omega: can you give an example on such an application? i can reproduce the firefox behavior. | 12:02 |
| Omega | banshee | 12:03 |
| Omega | I'll submit a bug report | 12:03 |
| andyrock | htorque, i think that valgrind unity-panel-service should valgrind also indicator-datetime, etc. | 12:03 |
| andyrock | if i'm not wrong they are "dlopened" in unity-panel-service | 12:04 |
| andyrock | btw tedg can give you more info ;) | 12:04 |
| htorque | Omega: post it here, i'll confirm | 12:04 |
| htorque | andyrock: thanks again! :-) | 12:05 |
| andyrock | yw | 12:08 |
| Omega | htorque: can you reproduce it with banshee, or nautilus or gnome-terminal? | 12:12 |
| Omega | (it might affect all gtk apps) | 12:13 |
| Omega | https://bugs.launchpad.net/indicator-sound/+bug/830037 | 12:17 |
| ubot5 | Ubuntu bug 830037 in The Sound Menu "media buttons do not function correctly when right click menu is open in some applications" [Undecided,New] | 12:17 |
| Omega | htorque: ^ | 12:17 |
| htorque | Omega: confirmed, and yes, i'm seeing the same results in those examples | 12:20 |
| Omega | htorque: are you on oneiric? | 12:21 |
| htorque | yes | 12:21 |
| Omega | I am on natty | 12:21 |
| htorque | ok, i'll add that to the report | 12:21 |
| Omega | (was reporting it against indicator-sound the best thing to do?) | 12:22 |
| htorque | i got no idea myself against which package to report that | 12:23 |
| htorque | someone will sort it out, no worries | 12:23 |
| === andyrock_ is now known as andyrock | ||
| === alecu is now known as ext3 | ||
| === ext3 is now known as alecu | ||
| manzabud | hi | 21:02 |
| manzabud | some one there | 21:02 |
| === yofel_ is now known as yofel | ||
| manzabud | hi some one | 21:42 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
