Re: Navit memory leak
Le 02/02/2010 18:51, vancel35 a écrit : > Thought there might be someone that's interested in this thread. > It's always interesting to know that an issue has been solved :) Thanks for the report. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Navit memory leak
On Tue, Feb 02, 2010 at 09:51:01AM -0800, vancel35 wrote: > The other possibility that I considered is that the display driver may have > had a memory leak, and two xserver packages were updated as well: > > $ grep xserver pkgs.upgradable.20100201 > xserver-xorg-conf - 0.1-r18.4 - 0.1-r19.4 > xserver-kdrive-common - 0.1-r38.4 - 0.1-r41.4 > > I don't know if either of those xserver updates have anything to do with the > screen buffer. I don't even know if any of these packages actually fixed No nothing to do with buffer.. it's xinput-calibrator in Xsession.d and config for gta01.. > the problem, but Navit doesn't crash (whatever the cause), and that's what's > important. :) but that's always good :) Cheers, -- uin:136542059jid:martin.ja...@gmail.com Jansa Martin sip:jama...@voip.wengo.fr JaMa ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Navit memory leak
Update: I did an opkg update & opkg upgrade yesterday, and there were 92 packages that were upgradable. An updated version of Navit was included, and the problem is fixed. $ grep navit pkgs.upgradable.20100201 navit - 1:0.1.0+svnr2902-r5.4.4 - 1:0.1.0+svnr2934-r5.4.4 navit-icons - 0.1.0+svnr2871-r1.4 - 0.1.0+svnr2927-r1.4 The other possibility that I considered is that the display driver may have had a memory leak, and two xserver packages were updated as well: $ grep xserver pkgs.upgradable.20100201 xserver-xorg-conf - 0.1-r18.4 - 0.1-r19.4 xserver-kdrive-common - 0.1-r38.4 - 0.1-r41.4 I don't know if either of those xserver updates have anything to do with the screen buffer. I don't even know if any of these packages actually fixed the problem, but Navit doesn't crash (whatever the cause), and that's what's important. :) Thought there might be someone that's interested in this thread. -Laura -- View this message in context: http://n2.nabble.com/Navit-memory-leak-tp4481548p4502355.html Sent from the Openmoko Community mailing list archive at Nabble.com. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Navit memory leak
Xavier Cremaschi writes: > :( didn't know that. Don't be so sad, it's much easier to do such development on x86 since valgrind will slow things down a lot. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Navit memory leak
On 30/01/2010 11:36, Timo Juhani Lindfors wrote: > Xavier Cremaschi writes: >> The best way is to launch a memory profiler I think (valgrind), or maybe > > Valgrind's ARM support is not in mainline yet so you need to do your > development on x86. :( didn't know that. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Navit memory leak
Xavier Cremaschi writes: > The best way is to launch a memory profiler I think (valgrind), or maybe Valgrind's ARM support is not in mainline yet so you need to do your development on x86. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Navit memory leak
Hi, I am not sure we can use htop to profile that. If a given software does lots of malloc/free, I could be a clever choice for the operating system to let it occupy a big chunk of memory. For example : you are the OS, you have 50mb of free memory and a user launch navit, which does lots of malloc/free... it seems ok for me to allocate ~45Mb for it (after 5 minutes of usage), since it seems to do lots of things in memory and you have 50Mb free. It does not mean this is a memory leak. The best way is to launch a memory profiler I think (valgrind), or maybe another big program when navit is launched. If its memory consumption decreases, then it's fine. If not, it's a memory leak. My 2 cents, Xavier. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Navit memory leak
I recently did an opkg update / upgrade, and a new version of Navit was installed. I am a new user to Navit, so I have it turned on even when I know where I'm going to learn how its routing engine works in practice... and because new programs are fun. :) Since I've been taking the same route every day, I know Navit with my current configuration should be able to handle my daily commute. However, after the update, Navit seems to have a SEVERE memory leak. I tested my theory by running an ssh session from my desktop and running htop within it. I started Navit and let it sit for a few minutes to watch memory. On start it was using about 10% of my total RAM, and it didn't change for several minutes. My FR had 79MB in use before Navit start and 82MB in use after Navit start. I suspected that it had to do with a malloc() call to render part of the map without freeing the memory that is now off-screen, and that appears to be the problem. I tested it by dragging the map and letting it redraw that part of the screen, and each time, Navit consumed more RAM. After dragging the map for only 1/2 of my commute distance, Navit has increased RAM usage to 17%, and my total memory is now at 92MB used. I'm sure the memory fills more quickly when I'm driving, because I have auto-zoom enabled, so it's getting multiple zoom levels in memory for the same map area. Is this a known issue? For such a short trip (13 miles / ~20Km), I wouldn't expect such a severe memory leak. The only configuration items I've changed from default is that I enabled autozoom. Other changes were after this memory leak appeared in attempts to make it stop such as downloading different size map regions from OSM (city level, region, state). An interesting note is that as I've been sitting here typing this email (about 15 minutes), it's slowly been releasing memory. Releasing it at such a slow pace seems silly when I'm using the "car" configuration. I'd actually rather have it not try to cache those rendered sections and just re-render if I return to the area. Hopefully I've given enough information to be useful in debugging. -Laura -- View this message in context: http://n2.nabble.com/Navit-memory-leak-tp4481548p4481548.html Sent from the Openmoko Community mailing list archive at Nabble.com. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
