| Wizzup | https://nlnet.nl/news/2025/20250422-announcement-grants-CommonsFund.html | 13:16 |
|---|---|---|
| sicelo | 💯 | 13:26 |
| arno11 | Wizzup: sicelo: i tried more testing with sphone/rakia, the ringtone trick doesn't work and when sphone crashes, vcm crashes too. what a mess | 14:32 |
| arno11 | if i kill/restart vcm from dsmetool, sphone works again | 14:34 |
| arno11 | are sphone/rakia calls still working on d4 daedalus btw ? | 14:38 |
| arno11 | (just to be sure all recent troubles are n900 specific...) | 14:39 |
| Wizzup | arno11: I think we want to understand why things crash | 14:45 |
| Wizzup | regarding rakia, I haven't gotten that to work well for me yet, I need to revisit that | 14:45 |
| Wizzup | but I would oh so love for that to work reasonably well | 14:45 |
| arno11 | yeah indeed, could be really cool | 14:46 |
| Wizzup | but maybe we should fix any issues with tp-ring first and then move onto tp0-rakia | 14:50 |
| Wizzup | tp-rakia | 14:50 |
| arno11 | sure | 14:50 |
| arno11 | ah @tp-ring calls, seems the ringtone randomly doesn't work only with idle/screen off | 15:06 |
| Wizzup | so I see this on my d4 | 16:05 |
| Wizzup | 2025-04-22T12:28:57.821929+02:00 maindroid pulseaudio[3307]: [pulseaudio] alsa-util.c: Error opening hctl device: No such device | 16:05 |
| Wizzup | 2025-04-23T15:48:34.610504+02:00 maindroid pulseaudio[3307]: [alsa-sink-40124000.mcbsp-cpcap-hifi cpcap-hifi-0] alsa-mixer.c: Assertion 'm' failed at ../src/modules/alsa/alsa-mixer.c:1479, function pa_alsa_path_select(). Aborting. | 16:05 |
| Wizzup | followed by | 16:05 |
| Wizzup | 2025-04-23T15:48:35.058654+02:00 maindroid pulseaudio[6374]: [pulseaudio] pid.c: Stale PID file, overwriting. | 16:05 |
| Wizzup | 2025-04-23T15:48:35.668609+02:00 maindroid pulseaudio[6374]: [pulseaudio] module-alsa-card.c: Failed to find a working profile. | 16:06 |
| Wizzup | 2025-04-23T15:48:35.684143+02:00 maindroid pulseaudio[6374]: [pulseaudio] module.c: Failed to load module "module-alsa-card" (argument: "device_id="0" name="platform-omap-hdmi-audio.3.auto" card_name="alsa_card.platform-omap-hdmi-audio.3.auto" namereg_fail=false tsched=yes fixed_latency_range=no ignore_dB=no deferred_volume=yes use_ucm=yes avoid_resampling=no | 16:06 |
| Wizzup | card_properties="module-udev-detect.discovered=1""): initialization failed. | 16:06 |
| Wizzup | 15:48 is when I started mpv | 16:06 |
| Wizzup | so it seems to me like pulseaudio is aborting | 16:06 |
| Wizzup | The assertion failure indicates that the mixer argument passed to pa_alsa_path_select is invalid | 16:10 |
| Wizzup | sicelo: I wonder if we ought to merge this mce fix for unexpected shutdown soon, is it mixed in in the pending mr? | 17:50 |
| sicelo | Wizzup: a quick and dirty fix is the one i shared yesterday ... you would have to create that MR since i didn't submit it: | 17:59 |
| sicelo | 22/04 12:53 < sicelo> if we're not ready for mce/#64, then a quick workaround/solution would be to change https://git.maemo.org/leste/mce/src/branch/master/src/modules/battery-upower.c#L194 to `if ((upowbat.state == UP_DEVICE_STATE_EMPTY && !mcebat.charger_connected) || ...` | 17:59 |
| Wizzup | right | 18:00 |
| Wizzup | well, I can make that change today | 18:00 |
| sicelo | but yes, mce/#64 already does this, and in fact does not use UP_DEVICE_STATE_EMPTY at all ... so even without a charger, as long as either voltage or capacity_level match, it would remain powered on | 18:01 |
| Wizzup | ok | 18:32 |
| freemangordon | on my n900 with chimaera, starting calculator takes about 5 seconds (warm start) | 20:27 |
| freemangordon | OMP takes about 6 | 20:28 |
| freemangordon | starting dist-upgrade | 20:29 |
| freemangordon | this is stock clock speeds ets | 20:32 |
| freemangordon | *etc | 20:32 |
| Wizzup | freemangordon: I don't know if you were gone for a bit or not, but I went over the strace extensively | 20:39 |
| Wizzup | and we realised that Arno was disabling some qt plugins unintentionally, and this gave him a speedup | 20:40 |
| freemangordon | I saw you did, but didn;t get the conclusion | 20:40 |
| Wizzup | We have three qt plugins, he only had two enabled | 20:40 |
| freemangordon | yes, I understand, but why would loading maemo platform plugin take so long? | 20:40 |
| freemangordon | or, did I miss something? | 20:41 |
| Wizzup | they all load hildon and gtk, which seems to do some theme and icon loading | 20:42 |
| Wizzup | the warm start is similar on daedalus to chimaera | 20:42 |
| Wizzup | for me on n900 at least | 20:42 |
| freemangordon | right, but loading some icons should not tgake 3-4 seconds, no? | 20:43 |
| Wizzup | if I recall correctly | 20:43 |
| Wizzup | well, you can look at strace and do the same work :) | 20:43 |
| Wizzup | if I recall correctly at least my conclusion was that there didn't seem to be a regression there | 20:43 |
| Wizzup | there might be in the pvr/dri mesa initialisation | 20:43 |
| freemangordon | I think there is no regression | 20:44 |
| Wizzup | but yes, out platform plugins do add significant startup overhead | 20:44 |
| freemangordon | but it is still slow | 20:44 |
| freemangordon | mhm | 20:44 |
| Wizzup | maemo-launcher support would probably help here. | 20:44 |
| freemangordon | yes | 20:44 |
| freemangordon | that's what I was thinking about | 20:44 |
| freemangordon | but I want to dist-upgrade my device to daedalus first | 20:44 |
| freemangordon | also, I wonder what it is so much faster on d4 | 20:45 |
| arno11 | freemangordon: i think there is a regression: warm start is definitely a lot slower on daedalus from a FRESH install. | 20:55 |
| arno11 | seems different from dist-upgrade | 20:56 |
| freemangordon | arno11: are you sure your dist-upgarded system is the same as FRESH install, like, you did lots of tweaks, no? | 20:56 |
| arno11 | just to be clear: i made some tweaks to workaround the main issue form a afresh install | 20:57 |
| freemangordon | anyway, lets see *my* dist-upgraded system performance | 20:58 |
| freemangordon | because in chimaera it takes lots of time, comparable to what you report for daedalus | 20:58 |
| freemangordon | *here* | 20:58 |
| arno11 | ah i see | 20:59 |
| freemangordon | also, I'll make measurements with a very basic QApplication, to see why our plugins add so much overheas | 21:05 |
| freemangordon | *overhead | 21:05 |
| Wizzup | some of this might be solved with some gtkrcs, specifying specific fonts and not loading everything else, that kind of stuff I guess | 21:05 |
| Wizzup | but we also load loads and loads of icons | 21:05 |
| freemangordon | I think the style plugin is inefficient, as it loads all kinds of gtk widgets, even if not needed | 21:06 |
| freemangordon | but have to confirm it | 21:06 |
| freemangordon | also, why do we load all those icons, do you have an idea? | 21:07 |
| Wizzup | I think this is the gtk2 qt style plugin | 21:07 |
| freemangordon | actually we should not load them, as they are served by sqpwood server | 21:07 |
| freemangordon | *sapwood | 21:07 |
| Wizzup | I am not sure that that works in qt | 21:08 |
| freemangordon | why not? ah, you think gtk2 qt plugin ignores pixbif engines? | 21:08 |
| freemangordon | *pixbuf | 21:08 |
| freemangordon | wait, it does not | 21:09 |
| freemangordon | you you try to start from ssh shell, it complains about missing sapwood-server if you don't export proper DISPLAY | 21:10 |
| Wizzup | freemangordon: no, I did: | 21:22 |
| Wizzup | on device, start osso-xterm, screen -SU work | 21:22 |
| Wizzup | close term | 21:22 |
| Wizzup | ssh, screen -Uxr | 21:22 |
| Wizzup | then you have a term with same env | 21:23 |
| freemangordon | Wizzup: btw, do we have noatime set for our ext4 partitions? | 21:35 |
| Wizzup | no idea | 21:35 |
| Wizzup | we have relatime | 21:36 |
| freemangordon | ugh | 21:36 |
| freemangordon | that's slow | 21:36 |
| freemangordon | where is rottfs mounted? | 21:36 |
| freemangordon | *rootfs | 21:36 |
| Wizzup | what do you mean | 21:36 |
| Wizzup | tea i create --repo leste/bugtracker -a MerlijnWajer -t 'Switch to noatime instead of relatime for root fs' | 21:36 |
| Wizzup | https://git.maemo.org/leste/bugtracker/issues/774 | 21:37 |
| freemangordon | I don't see rootfs mounted in /etc/fstab | 21:37 |
| freemangordon | so, where are mount options? | 21:37 |
| Wizzup | probably defaults to relatime | 21:41 |
| Wizzup | oh, I see what you mean | 21:41 |
| Wizzup | I think it's probably just ... not there | 21:42 |
| freemangordon | with noatime calendar loads in 4 seconds (warm) | 21:46 |
| Wizzup | on daedalus? | 21:46 |
| freemangordon | chimaera | 21:47 |
| Wizzup | ok | 21:47 |
| Wizzup | btw you might be able to do mount -o remount,noatime / | 21:47 |
| Wizzup | (saves a reboot) | 21:47 |
| freemangordon | that's what I did | 21:47 |
| Wizzup | :) | 21:47 |
| freemangordon | I think we shall set noatime everywhere | 21:48 |
| freemangordon | that's a great improvement more or less for free | 21:48 |
| freemangordon | not to say we *must* set it for flash storage | 21:49 |
| arno11 | i just tried noatime on daedalus (no qt5 tweaks), calculator starts in 12 sec warm | 21:54 |
| freemangordon | I just started dist-upgrade, will take a while | 21:55 |
| arno11 | freemangordon: on daedalus, the biggest issue i see with qt5 is when you launch more than one qt5 app: need 20-30 sec to start calculator or qalendar | 22:07 |
| arno11 | i mean, with only 2 apps, it doubles the time | 22:08 |
| sicelo | where will that noatime be set? want to boot with it going forward. | 23:23 |
| sicelo | i suppose it means defining it in fstab, which currently doesn't have definition for / ? | 23:26 |
| sicelo | oh, kernel has `rootflags=` | 23:30 |
| Wizzup | I'd put it in fstab yeah | 23:34 |
| Wizzup | but yes, you can use rootflags | 23:34 |
| sicelo | can you share how your / is currently mounted (my N900 is out of battery ... about to boot, and i want to make this change before booting) | 23:37 |
| sicelo | iirc we're ext4? | 23:38 |
| sicelo | done :-) | 23:51 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!