| Wizzup | sicelo: as noted, you can make this change live using mount -o remount,noatime / | 00:08 |
|---|---|---|
| Wizzup | fwiw | 00:08 |
| sicelo | i wanted to make it permanent, on every boot | 06:16 |
| freemangordon | sicelo: please make a PR that makes it permanent for everyone | 07:17 |
| freemangordon | we would like data=writeback too, perhaps | 07:20 |
| freemangordon | hmm: | 07:23 |
| freemangordon | dpkg: warning: downgrading hildon-base from 1.14+m7.1 to 1.10+4m7 | 07:23 |
| freemangordon | ok, that's fine | 07:24 |
| freemangordon | Wizzup: sicelo: see https://developer.ridgerun.com/wiki/index.php/High_performance_SD_card_tuning_using_the_EXT4_file_system | 07:38 |
| freemangordon | so yeah, I think we want both noatime and data=writeback | 07:39 |
| freemangordon | https://community.progress.com/s/article/linux-filesystem-tuning---performance-datawriteback-vs-dataintegrity-dataordered | 07:46 |
| freemangordon | https://www.infradead.org/~mchehab/kernel_docs/admin-guide/ext4.html | 07:51 |
| sicelo | mmm, the progress.com website seems to advise against data=writeback | 08:31 |
| freemangordon | sure, but what is the context - "Product: MarkLogic" | 08:33 |
| freemangordon | we're not making server OS here :) | 08:35 |
| sicelo | freemangordon: Wizzup: might be unrelated causes, but seems that with the icon blinking forever, power draw is a bit high | 09:47 |
| sicelo | freemangordon: try that data=writeback yourself ... seems to not be working correctly. at least it sent me to recovery login | 10:57 |
| Wizzup | sicelo: yeah I think it's searching for wifi all the time | 10:58 |
| Wizzup | the main effect of the patch it as I understand it that it doesn't 'give up' on connecting | 10:59 |
| Wizzup | this works around the kernel/firmware issues where it sometimes rejects the association, but it also has this effect | 10:59 |
| Wizzup | freemangordon: this writeback is mostly for writes yeah, not for reads? | 11:01 |
| Wizzup | I am not sure if disabling journalling is a good idea | 11:01 |
| Wizzup | sicelo: building status area applet | 11:03 |
| sicelo | tx | 11:04 |
| sicelo | i wouldn't mind the writeback that much myself, if it would boot :-) | 11:04 |
| Wizzup | I don't think disabling journalling is a good idea tbh, especially on devices that run on battery | 11:06 |
| Wizzup | you probably have to remove the journalling bit or something | 11:06 |
| Wizzup | and run fsck | 11:06 |
| Wizzup | see the wiki that fmg linked, it has like 3-4 steps | 11:06 |
| Wizzup | but again, they're trying to optimise for writing 40 fps 1080p video | 11:07 |
| uvos__ | disabeling journalling on a battery powered and genneraly pretty unstable platform is a terrible idea imo | 14:05 |
| uvos__ | i also like my atime, but i can see disableing this | 14:07 |
| g4rrgl3n0m4d | .11 | 14:16 |
| Wizzup | uvos__: I agree @ no journalling | 14:30 |
| Wizzup | uvos__: can you look at the mce mr and see if it's ready? I'm living without battery icon for some days now :D | 14:30 |
| freemangordon | wait, where did you find that data=writeback means journaling is comp[letely disabled? | 15:20 |
| Wizzup | the links you sent say it | 15:21 |
| freemangordon | hmm, I think they say something else | 15:21 |
| Wizzup | The optimization consists on selecting the data mode for our fileystem as data=writeback. The other two possible data modes (data=ordered and data=journal) do not have as good of bandwidth performance as writeback. The EXT4 journaling is disabled when using writeback. For details on the attributes of the various EXT4 data modes, read the "Data Mode" section from the Ext4 Filesystem kernel | 15:21 |
| Wizzup | documentation. | 15:21 |
| freemangordon | sec | 15:22 |
| freemangordon | "Data ordering is not preserved, data may be written into the main file system after its metadata has been committed to the journal." | 15:22 |
| freemangordon | this is from https://www.infradead.org/~mchehab/kernel_docs/admin-guide/ext4.html | 15:22 |
| freemangordon | ok, my n900 powered off with fully charged battery | 15:23 |
| freemangordon | I was under the impression this was already fixed | 15:23 |
| freemangordon | hmm, no, wait | 15:24 |
| freemangordon | it is up | 15:24 |
| freemangordon | but very busy | 15:24 |
| freemangordon | hmm, what it is doing? | 15:25 |
| sicelo | mining bitcoin :-) | 15:27 |
| freemangordon | looks like | 15:27 |
| freemangordon | maybe no swap | 15:27 |
| freemangordon | yeah, no swap | 15:29 |
| freemangordon | ok, qalendar starts (warm) in 5 seconds | 15:36 |
| freemangordon | wait, but only when started from ssh | 15:37 |
| freemangordon | if started from launcher, it takes > 10 s | 15:38 |
| freemangordon | ok, for sure there is a regression | 15:49 |
| arno11 | freemangordon: yeah and it is even worse if you launch another qt5 app | 15:53 |
| freemangordon | unfortunately looks like a regression in qt | 15:53 |
| freemangordon | maybe a comparison between strace from chimaera will reveal something | 15:54 |
| arno11 | yep | 15:55 |
| Wizzup | freemangordon: the description 'over ssh' is not a very useful one unless it means 'not the standard environment' | 15:55 |
| freemangordon | well, "over ssh" means - ssh client session | 15:55 |
| freemangordon | but even there it is slower than on chimaera | 15:56 |
| freemangordon | not to say that it takes ages to boot | 15:56 |
| Wizzup | I use GNU screen and then it's 100% the same | 15:56 |
| Wizzup | if it takes ages to boot then it doesn't seem like it'd be qt related almost | 15:56 |
| freemangordon | mhm | 15:57 |
| freemangordon | [ 209.983886] wlan0: authenticate with 10:fe:ed:dd:f7:61 | 15:57 |
| freemangordon | connection is set to "auto connect to wifi" | 15:59 |
| freemangordon | ok, calculator took > 20s to start | 16:01 |
| arno11 | indeed... | 16:01 |
| freemangordon | hmm, and 30% of the memory, what the? | 16:02 |
| arno11 | yeah that's crazy | 16:03 |
| freemangordon | it takes 75MB RES | 16:03 |
| arno11 | but seems not so leste specific, other distros are affected, that's how i found the qt5ct raster widget tweak | 16:05 |
| uvos__ | "Data ordering is not preserved, data may be written into the main file system after its metadata has been committed to the journal." | 16:06 |
| arno11 | btw forcing raster from env var to launch an app doesn't help. it is also very slow | 16:06 |
| uvos__ | iiuc that breaks restoreing from journal | 16:06 |
| uvos__ | so its as good as not haveig it | 16:06 |
| freemangordon | arno11: try export MESA_SHADER_CACHE_DISABLE=true | 16:07 |
| arno11 | ok | 16:07 |
| Wizzup | if it's not the same from the maemo loader and from regular env you have another problem | 16:07 |
| Wizzup | I think you first want to make sure everything is on the same env | 16:07 |
| Wizzup | for example, I don't think arno's setup loads our qt style plugin with special maemo widgets | 16:07 |
| freemangordon | sure, but there still might be some effect | 16:08 |
| Wizzup | yup | 16:08 |
| freemangordon | anyway, it is not normal for qalendar to take 70MB | 16:08 |
| arno11 | Wizzup: it is easy to disable qt5ct stuff, i test without it on the fly | 16:09 |
| Wizzup | you can check /proc/pid/maps | 16:09 |
| Wizzup | see if there's other libs loaded that you do not expect | 16:09 |
| arno11 | freemangordon: same result 10 sec | 16:09 |
| Wizzup | raster should be disabled for this | 16:10 |
| freemangordon | sorry, have to run, ttyl | 16:10 |
| freemangordon | but | 16:10 |
| freemangordon | qalendar -style=windows -platform=xcb | 16:11 |
| freemangordon | still takes lots of memory (even more that by default) and takes ages to load | 16:11 |
| freemangordon | *than by | 16:11 |
| Wizzup | freemangordon: there's also the input plugin | 16:30 |
| Wizzup | I would run env|grep QT | 16:31 |
| Wizzup | and unset them all | 16:31 |
| uvos__ | "1745497821 <Wizzup> uvos__: can you look at the mce mr and see if it's ready? I'm living without battery icon for some days now :D" | 16:51 |
| uvos__ | done | 16:51 |
| Wizzup | thanks :) | 17:02 |
| Wizzup | I'll finish some work things and then build/test it | 17:02 |
| sicelo | thanks. | 17:05 |
| sicelo | let us know if you end up with magic smoke ;-) | 17:08 |
| sicelo | but yeah, i've been running this for weeks now - on D4 while the battery was still working a bit, and on both my N900, the one which charges normally (thus reports normally), and also on the one with broken USB (thus always 0%) | 17:11 |
| Wizzup | merged & building | 17:26 |
| freemangordon | Wizzup: with everything QT unset, and MESA SW, it is still ~75MiB | 19:46 |
| freemangordon | calculator uses 65 MiB | 19:46 |
| freemangordon | that's not normal | 19:46 |
| freemangordon | will make a minimalistic qt app when I have time | 19:47 |
| Wizzup | calculator might load qt5maemo | 19:49 |
| sicelo | Wizzup: silly q - so what's the plan with the homepage? will it remain on github.io? | 20:12 |
| Wizzup | I don't know :) | 20:16 |
| Wizzup | not planning any major disruption, maybe keep it there, maybe mirror, maybe redirect | 20:17 |
| sicelo | cool | 20:38 |
| Wizzup | sicelo: just rebooted after upgrading, looks like there's still some reboot thing going on | 20:47 |
| Wizzup | but the battery applet showed the battery at ~30% or so, basin on how green-filled the battery weas | 20:47 |
| Wizzup | before I rebooted the device was at 87% | 20:47 |
| sicelo | mmm, | 20:48 |
| sicelo | share /sys/class/power_supply/battery whenever you can, and `upower -d` | 20:49 |
| Wizzup | booting with charged worked | 20:50 |
| Wizzup | charger* | 20:50 |
| Wizzup | https://pastebin.com/raw/KHWYgFwE | 20:51 |
| Wizzup | it rebooted again | 20:52 |
| Wizzup | $ | 20:52 |
| Wizzup | Broadcast message from root@maindroid (Thu Apr 24 20:51:43 2025): | 20:52 |
| Wizzup | The system is going down for system halt NOW! | 20:52 |
| Wizzup | heh | 20:52 |
| sicelo | rebooted when you disconnected charger? | 20:53 |
| Wizzup | no | 20:53 |
| Wizzup | with charger connected | 20:53 |
| Wizzup | it really seems to take a while though, quite interesting | 20:53 |
| Wizzup | I am going to try to ssh in and run upower -d | 20:54 |
| Wizzup | https://paste.debian.net/1371364/ huh the critical action looks strange | 20:56 |
| Wizzup | what should the config path be again | 20:56 |
| Wizzup | /etc/UPower/Upower.conf right? | 20:56 |
| sicelo | yes | 20:56 |
| Wizzup | looks like my config is messed up somehow | 20:57 |
| Wizzup | it reboots again | 20:57 |
| Wizzup | I don't understand why upower does this, but clearly the config is wrong | 20:57 |
| sicelo | and definitely that upower -d says yours is wrong ... but you're on correct upower version at least | 20:57 |
| Wizzup | the config in /etc for me is empty | 20:57 |
| Wizzup | well, in /etc/UPower | 20:57 |
| Wizzup | but there is a leste-orig | 20:57 |
| sicelo | must be the packaging | 20:57 |
| Wizzup | will get to the bottom of this | 20:57 |
| Wizzup | ok, got it set to ignore now | 21:02 |
| Wizzup | phew | 21:02 |
| Wizzup | not sure what's up with leste config though, it looks ok | 21:03 |
| Wizzup | it does say my battery is at 0% again ,but at least no reboot | 21:03 |
| sicelo | i downloaded the binaries from maedevu: | 21:07 |
| sicelo | $ dpkg-deb -c leste-config-common_1.132-1+4m7_all.deb | grep UPower | 21:07 |
| sicelo | -rw-r--r-- root/root 3376 2025-04-18 12:09 ./etc/UPower.conf.leste | 21:07 |
| sicelo | dpkg-deb -c leste-config-common_1.132-1+4m7_all.deb | grep UPower | 21:07 |
| sicelo | -rw-r--r-- root/root 3376 2025-04-18 12:09 ./etc/UPower.conf.leste | 21:07 |
| sicelo | so the issue is in leste-config | 21:07 |
| sicelo | as for 0% ... is it reporting it while you're charging? i disabled percentage estimation while charging, because it over-estimates. when you disconnect charger, it should resume estimating | 21:11 |
| sicelo | as for 87% sounds like it was estimating indeed .. but should have been 82%, or it was 57%? it was clearly not directly from UPower, since your battery isn't reporting any percentage | 21:15 |
| Wizzup | sicelo: is that the latest leste-config? | 21:19 |
| Wizzup | did you pull from daedalus-devel? | 21:20 |
| sicelo | manual download from maedevu, because i wanted to be 100% sure that i'm not looking at any customization i might have locally | 21:24 |
| sicelo | and yes, according to changelog, 1.132 is the latest leste-config | 21:25 |
| Wizzup | I must have forgotten to commit something in git | 21:37 |
| Wizzup | yeah... | 21:37 |
| Wizzup | lol | 21:37 |
| Wizzup | I am off my game recently :) | 21:38 |
| Wizzup | building new leste-config | 21:39 |
| Wizzup | I did rename the file, but with 'mv' and not 'git mv' | 21:39 |
| sicelo | keep me updated :-) | 22:07 |
| Wizzup | it should be in the repo | 22:08 |
| sicelo | yes i saw. is your device fixed now? reporting is still 0%? | 22:16 |
| Wizzup | (work mtg, but will check) | 22:22 |
| Wizzup | yeah the percentage is ok | 22:22 |
| Wizzup | will update and reboot | 22:22 |
| sicelo | now an MR i've waited for years to be able to make, https://git.maemo.org/leste/leste-config/pulls/60 | 22:42 |
| arno11 | cool | 22:46 |
| sicelo | hah, was also in april, https://gitlab.freedesktop.org/upower/upower/-/merge_requests/40#note_1345120 | 23:19 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!