T42 | <elros34> @b100dian do you have some device with 5.0 sailfish affected by old notification with actions (screenshot notification) bug? On fxtec pro1x this is definitely still the case. I can't believe sony devices are not affected too | 12:37 |
---|---|---|
T42 | <b100dian> Hmm.. I do se 18%-20% from lipstick on tucana, which is gone after reboot, let me check sake too | 13:03 |
T42 | <elros34> on htop it's like 15% on fxtec until I got other notification without actions/reboot | 13:07 |
T42 | <b100dian> Not much on sake @elros34 | 13:07 |
T42 | <elros34> this is strange, thx | 13:08 |
T42 | <b100dian> ok, let me trigger a such notification with buttons, hang on | 13:08 |
T42 | <elros34> it must be notification with buttons (for example screenshot notification) | 13:09 |
T42 | <elros34> this doesn't afect regular | 13:09 |
T42 | <elros34> and it must disappear -> window must be hidden | 13:10 |
T42 | <elros34> notification window | 13:10 |
T42 | <b100dian> much smaller cpu raise, from 0.5% to 2.5% on sake, probably faster CPU? | 13:12 |
T42 | <b100dian> let me find the fix and try with it, I remember it was a simple line removed | 13:13 |
T42 | <elros34> hm 2.5 is quiet low. Do you see difference when you expand notification (there should be now cpu usage) vs when it's hidden | 13:14 |
T42 | <elros34> be no cpu usage* | 13:14 |
T42 | <b100dian> Are you talking about hiding it from top/transient view or from notification area? | 13:16 |
T42 | <elros34> transient view | 13:20 |
T42 | <b100dian> Yes, it's 0.1% when expanded in the top/transient area, and back to 2.something when dismissed. For reference, the jump in CPU usage on C2 is 0.1 to 1.4% but both devices (Sake and C2) consistently drop to almost 0% when notification is expanded and then back to 1 and 2 percent when dismissed. | 13:25 |
T42 | <b100dian> So I'd say it is still there on those devices too, it is just far less noticeable than 20% on tucana | 13:25 |
T42 | <elros34> maybe it's related to dynamic cpu freq. Better measue would be strace. strace -f -p <pid of lipstick> looks sane when notification shows and it's expanded but it's flooded of ppoll when you dismiss notification | 13:26 |
T42 | <elros34> need to update patch and description | 13:26 |
T42 | <b100dian> Well, for sake and C2 if you run strace -f -p `pidof lipstick` in Terminal and you interact with the phone (e.g. trigger screenshot) lipstick freezes.. Need to check at the computer... | 13:40 |
T42 | <b100dian> it's as if strace catches its own output interacting with lipstick and enters into oscillation / resonance :p | 13:45 |
T42 | <elros34> yeah on device this might be a bit unreliable | 13:52 |
T42 | <b100dian> @elros34 sorry for the long wait, but it does reproduce over telnet (and ssh respectively, I don't know why no telnet on C2) | 15:24 |
T42 | <b100dian> (it doesn't have init_disable_telnet) | 15:24 |
T42 | <elros34> it's not community build, not regular hybris boot and init-debug script | 15:25 |
T42 | <b100dian> It does have init-debug script:) | 15:41 |
T42 | <b100dian> Anyway, not the point here. Are you rebasing the patch or you want me to try it first | 15:42 |
T42 | <elros34> I have updated it for 5.0 | 15:43 |
T42 | <b100dian> confirmed working on tucana, thanks! | 15:50 |
mal | what kimd of patch? | 16:07 |
T42 | <elros34> old issue https://coderus.openrepos.net/pm2/project/sailfishos-notification-preview-high-cpu-usage-patch | 16:08 |
mal | hmm, need to discuss that internally at jolla next week | 16:09 |
T42 | <b100dian> Would it help to file it as a bug in https://github.com/sailfishos/issue-tracker/issues ? | 16:33 |
T42 | <Mister_Magister> bring back bugzilla.merproject.org | 16:35 |
T42 | <Mister_Magister> wait it was actually bugs.merproject.org | 16:36 |
*** phlixi is now known as Guest11499 | 16:51 | |
*** phlixi_ is now known as phlixi | 16:51 | |
mal | Mister_Magister: you probably haven't noticed this https://github.com/sailfishos/issue-tracker ? | 16:58 |
mal | ah, missed that it was mentioned already, sorry | 16:59 |
Mister_Magister | nope | 17:00 |
Mister_Magister | should make like 10 entries | 17:00 |
Mister_Magister | mal: got 5 minutes? | 17:00 |
mal | ? | 17:03 |
Mister_Magister | mal: I wanted you to ping me when you have time cause i want to describe to you issue with audio cause i need to get it out of my system and you're the one of few people who will even understand what i'm blabbing about :D | 17:04 |
mal | why don't you just explain it | 17:10 |
Mister_Magister | because i don't know if you have time to discuss it | 17:11 |
Mister_Magister | otherwise i would be shouting into the void xd | 17:11 |
mal | I have time now | 17:12 |
Mister_Magister | oki here i go | 17:12 |
Mister_Magister | Basically, i was trying to understand the audio config files and I remember that jusa was playing with mixer_paths on my x2, and I noticed there's mixer_paths and mixer_paths_overlay files, i stumbled upon amixer cset/cget and in the list of properties I was looking at what's available, the error was "no BE found for Tertiary MI2S Playback" and "no backend DAIs enabled for KONA Media1" (now i understand that BE stands for | 17:21 |
Mister_Magister | backend mostlikely) so in the cset stuff i was looking at stuff that would reference MI2S, Tetriary and Media1 and by accident or educated guess i did amixer cset name='TERT_TDM_RX_0 Audio Mixer MultiMedia1' on and that enabled the audio, but it was only mono and only in one speaker so i played further in settings but couldn't figure out any sensible combination, i tried your halium stuff but i didn't have the | 17:21 |
Mister_Magister | libsndcardparser.so so i stole it from internet hoping it will do something and it did but thats still an issue, but that only enabled 2 speakers in wrong orientation (more on that later) so that wasn't total fix either, then i looked into mixer_paths_overlay_static.xml and I noticed in speaker path it has the TERT_TDM_RX_0 Audio Mixer MultiMedia1 at which point i understood that mixer_paths literally just automatically do | 17:21 |
Mister_Magister | the cset and some of those don't have to be present, and that the mixer_paths_overlay_static.xml must not be in use because when i did alsactl monitor, it did not set it, so then i used strings and grepped for the mixer_paths_overlay_static.xml in vendor audio libraries and i didn't find it in 64bit version but i found it in 32bit version which was giant "aha" moment, and then i realised android binary is 32 bit so it loads | 17:21 |
Mister_Magister | 32 bit version which loads the correct file which correctly enables audio and halium stuff talks to the 32bit binary. Now since we have that, one mistery is solved but the issue still exists which i'll get to in next message | 17:21 |
Mister_Magister | basically when using halium stuff it enables top two speakers in stereo mode, now this tablet had 4 speakers and i think you realise where i'm going with this. by default halium only enables top two speakers in vertical orientation, and not only i have to figure out how to obtain libsndcardparser.so but also how to enable bottom speaker and third, i'll probably have to write mce plugin that changes audio config depending on | 17:23 |
Mister_Magister | the screen orientation because thats what android does because when its portrait you have two speakers top/bottom so they have to be stereo vertically and when its horizontal you gotta make them horizontal and I don't think sfos has system for that | 17:23 |
mal | didn't building halium module end up building libsndcardparser.so? is that nowhere on device? | 17:26 |
mal | how did you test it, just copying that module or packaging droid-hal after the build? | 17:27 |
Mister_Magister | i grepped entire out and it's not there and i grepped / of the device and its not there either | 17:27 |
Mister_Magister | i just copied the .so | 17:27 |
mal | you mean find not grep? | 17:27 |
Mister_Magister | yes | 17:27 |
mal | what error does it give when it loads? | 17:28 |
Mister_Magister | it complains about libsndcardparser.so | 17:28 |
mal | does it fail with that or just complain? | 17:29 |
T42 | <Mister_Magister> complain like a lot but does not fail | 17:29 |
T42 | <Mister_Magister> i can check right now | 17:30 |
Mister_Magister | basically i have it in /usr/libexec/droid-hybris/system/lib64/hw/ and is symlinked to the same path as audio.primary.kona.so | 17:30 |
Mister_Magister | now if i start pa manually | 17:31 |
Mister_Magister | oh wait i gotta start android service | 17:31 |
mal | like this https://github.com/mlehtima/droid-config-fp5/blob/master/sparse/usr/libexec/droid-hybris/system/lib64/hw/audio.primary.lahaina.so ? | 17:31 |
mal | and then that android service of course | 17:31 |
Mister_Magister | ah wait | 17:32 |
Mister_Magister | yeah | 17:32 |
Mister_Magister | but | 17:32 |
Mister_Magister | i'm getting error about it not being accessible from that namespace | 17:32 |
Mister_Magister | /usr/libexec/droid-hybris/system/lib64/hw/audio.primary.kona.so" needed or dlopened by "/usr/libexec/droid-hybris/system/lib64/libvndksupport.so" is not accessible for the namespace "(default)" | 17:32 |
Mister_Magister | so i used mount --bind instead | 17:32 |
mal | did you place audio.hidl_compat.default.so to /usr/libexec/droid-hybris/system/lib64/hw/ ? | 17:34 |
Mister_Magister | yes | 17:34 |
mal | do you start pulseaudio manually or via service? | 17:34 |
Mister_Magister | manually | 17:34 |
mal | did you remember the extra env var? | 17:34 |
Mister_Magister | anything else than HYBRIS_USE_VENDOR_NAMESPACE=yes pulseaudio --daemonize=no -n --file=/etc/pulse/arm_droid_default.pa? | 17:35 |
mal | that namespace thing was what I meant | 17:35 |
Mister_Magister | ok i've enabled the android service, rebooted, mount --bind and restarted pa | 17:36 |
mal | did you revert the audio xml changes? | 17:36 |
Mister_Magister | i did not do any audio xml changes | 17:37 |
Mister_Magister | and i get audio btw | 17:37 |
mal | ok | 17:37 |
Mister_Magister | idk it stopped complaining about libsndcardparser.so magically i think | 17:37 |
Mister_Magister | or its just not logged | 17:37 |
mal | maybe it was trying to do something different if android service was not running | 17:37 |
mal | are you sure the bind mount is needed? | 17:38 |
Mister_Magister | i can try jsut the symlink if you want | 17:38 |
Mister_Magister | yeap it doesn't start with symlink | 17:39 |
Mister_Magister | and yeah all 4 speakers are working, jsut not following orientation xd | 17:41 |
Mister_Magister | i don't think anyone ever in history of sfos had to rotate the audio channels but hey, i was also the first for front facing camera flash xd | 17:43 |
Mister_Magister | https://paste.opensuse.org/pastes/abc9dbba1e3e | 17:45 |
Mister_Magister | this could be it | 17:45 |
Mister_Magister | and the controls do exist | 17:45 |
Mister_Magister | but ye took me a lot of thinking to figure it out and i'm quite proud of it xd | 17:48 |
mal | yeah, probably no device so far has had that kind of setup | 17:51 |
mal | still odd why the symlink is not working | 17:52 |
mal | wait | 17:53 |
mal | I have one patch in system side | 17:53 |
mal | wondering how that can be handled in community style builds | 17:54 |
mal | I patched system/linkerconfig with https://pastebin.com/wTM8Wnxh but need to think where community style builds load that | 17:56 |
Mister_Magister | i could just grep for that | 17:57 |
Mister_Magister | but now, i have to leave you, thanks for reading my rant | 17:57 |
mal | try to see how you could make custom linkerconfig work | 17:57 |
Mister_Magister | tbh if we just locked orientation to horizontal one that would probalby do it | 17:59 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!