libera/#maemo-leste/ Tuesday, 2021-12-28

uvosdpkg: warning: downgrading mce-dev:armhf from 1.9.10 to 1.9.800:12
uvosuhh00:12
uvosoh00:12
uvosnvm user error00:12
uvosso new mce version now turns off cpu1 on mapphones while the display is off00:24
uvosas a hopfully temporary hack to reduce the ipis00:24
uvosnote that this is bad if your compiling stuff on it00:25
uvoshttps://github.com/maemo-leste/mce/blob/ad2ffa135137f01c3f1a009b7bd6b963d33f66d9/config/mce.ini#L33800:25
uvossee the above to turn it off00:25
uvosor unload the quriks-mapphone to disable all the other pm hacks as well00:26
Wizzupdid that actually help with pm for you btw?00:52
WizzupI think for my it actually made it use more power somehow00:58
Wizzupbut maybe I set it wrong00:58
Wizzup(when I did it manually I mean)00:59
uvosyeah01:01
uvosit did01:01
uvos15-20mW01:01
uvosaccording to the script anyhow01:07
uvosmaybe you can mesure it on psu01:07
Wizzupok01:17
Wizzupdid you also figure out the other difference between tmlind's droid4?01:18
Wizzupiirc in omapconf?01:18
Wizzupbtw we should sync stuff from devel to stable when the n900 X stuff is stable (and then we'll do a droid4-image)01:18
Wizzuperr omap-linux pkg01:18
uvosyeah so we have a device more active than tmlind01:20
uvosbut i have not done anything to see why01:20
uvosthis is also true for xt7501:21
uvos*xt87501:21
uvosso its seperate from the xt875 difference01:21
Wizzupok01:36
* Wizzup zz01:37
freemangordondo we know why omap8250_set_mctrl oops-es on poweroff/reboot?08:07
freemangordonon d408:07
freemangordonhmm, modesetting tearing with HW accel means that omapdrm has some issue08:55
freemangordontmlind: uvos: does anyone of you have access to d4 vendor kernel source code?09:16
freemangordonI need to know if vendor kernel uses PROGRAMMEDLINENUMBER_IRQ or TE_TRIGGER_IRQ09:22
tmlindfreemangordon09:37
tmlindomap8250_set_mctrl pm is broken09:37
tmlindfreemangordon: you can find the stock v3.0.8 kernel on github, i use github.com/NotKit/android_kernel_motorola_omap4-common as it's been patched for gcc v609:39
tmlindcan't find those defines there09:40
tmlindsounds like omap8250_set_mctrl needs serial8250_rpm_get() and put around it09:41
Wizzupuvos: looks like it worked indeed11:55
Wizzupin the morning my d4 still said 'two days left'11:55
Wizzup(I didn't charge it since the evening)11:55
Wizzuptook about 15% in ~8-9 hours11:56
Wizzupso that's like 54 hours11:56
Wizzup08:55 < freemangordon> hmm, modesetting tearing with HW accel means that omapdrm has some issue11:57
Wizzupjust to be clear, why is that?11:57
Wizzupmodesetting also has tearing problems does it not?11:57
Wizzupit has on my amdgpu11:57
Wizzupwell actually that is not modesetting I think11:57
Wizzupin any case even my amdgpu needs: randr --output HDMI-A-0 --set TearFree on11:57
uvosbtw the device we have on more than tmlind is the audio codec11:59
uvosbut i have really no idea why11:59
uvosbooting w/o pulse dosent help, messing with the mixer dosent help and blacklisting the faling hdmi audio module dosent help either12:00
Wizzupuvos: maybe it's some modem config?12:04
uvosits the omap4 audio codec12:04
Wizzuprafael2k: will try to do all the pkgs today, had christmas/family stuff12:04
uvosthe modem is not involved12:04
Wizzupuvos: hm12:04
Wizzupuvos: kernel config diff?12:04
Wizzuphdmi related?12:04
rafael2kWizzup: hey!12:04
uvosmaybe12:04
uvosi balcklisted hdmi audio12:04
rafael2kwhat up!12:05
rafael2kWizzup: which packages? just got sick as a dog the last days... after covid I got some virus12:05
Wizzupofono, kernel, etc12:05
Wizzupthe work you did12:05
WizzupI'll get it packaged12:05
rafael2kthey worked well12:05
WizzupI need to go out and pick up some friends from the station12:05
uvosWizzup: sure maybe something keeps alsa busy even when pulse is not there12:05
rafael2kI'm using for days12:05
Wizzupyeah but we need them in the CI etc12:06
uvosvia alsa12:06
Wizzuprafael2k: great12:06
uvoscould be mis or soemthing12:06
Wizzupuvos: I am not sure if I have hdmi blacklisted, maybe some probe doesn't complete12:06
uvosWizzup: the probe of hdmi audio errors out yes12:06
uvossince 5.9 irrc12:06
rafael2kjust the kernel postinst needs some care, like copying dtbs are copy kernel and initrd to the proper place to make uboot pick them by default12:06
rafael2kalso, is it normal that I get "Calibration needed" for battery applet?12:07
uvosrafael2k: btw could you implement the emergency shell on pp via the volume buttons?12:07
rafael2ksound setup need some care with newer kernel need some care too - this is my next goal12:08
uvos"Calibration needed" is normal if the battery dosent report its current full cap.12:09
rafael2kuvos: could be, through u-boot right? Any hint what command args to pass to kernel?12:09
Wizzupuvos: maybe this is not the case for tmlind12:09
rafael2kuvos: I realized no matter what, battery always shows changing, even after a long time charging...12:09
uvosrafael2k: https://github.com/maemo-leste/mapphone-kexecboot-config/blob/cba6c5acef6a03f36d4a8e62c1aea6b9ee243302/boot.cfg.bionic#L1212:09
uvossee that12:09
Wizzuprafael2k: this is upower state bugs12:09
Wizzupafk12:10
uvosyes via uboot, would be cool if one of the vol buttions woudl cause the device to boot that12:10
uvosWizzup: tmlind reported the hdmi autio bug too i think but <---- tmlind12:10
rafael2ksure, that would be easy, and indeed, with three options (vol up, vol down, and no vol pressed) it is good for the common user (one for eMMC, one for emergecy shell, default for maemo normal boot)12:11
rafael2kWizzup: ok!12:11
uvosrafael2k: youll need fbkeyboard installed12:11
rafael2kuvos: ok12:11
rafael2kone feature I did not see in other pinephone distros is the call recording VoiceCall profile... I'd like to implement this too. How cool wouldnt it be to record a real CS voice call!12:15
rafael2kor just plug any audio recorder in the pa source and voila!12:15
rafael2kjust on very new Android phones this is available, and I don't think there is even official API for that12:16
uvosnot really12:16
uvosits been avaiable on random android phones for a long time12:17
uvosmostly beacuse of regulatory reasons12:17
rafael2kanyway, it does not matter, we can do it12:17
uvosyes12:17
rafael2k; )12:17
uvoswould be great to have sphone support this via module12:18
uvosxt875/xt894/xt860 can also do this12:18
rafael2kthat is what I'm thinking12:18
rafael2kbtw, this battery issue that upower does not report its current full cap can fuck up the battery?12:20
uvosno12:20
rafael2ktks12:21
rafael2kcan we hardcode this somehow to make it nice in the UI?12:22
uvosim sure the pp's hw setup can report battery capcaity if callibrated12:22
uvosi just dont know how beacuse i dont have one12:22
rafael2kright12:22
uvoson many devices you have to charge fully and discharge the device fully once in a single boot12:23
uvosfor it to mesure the capacity12:23
rafael2kright, I can do this and see what happens, tks!12:23
freemangordonWizzup: modesetting tears *without* HW accel, typo12:27
freemangordonit should not, as it does page flip12:27
freemangordonpage flip happens on vsync (in theory) so it should not tear12:27
freemangordonall this with compositing ofc12:28
freemangordonso, compositing + page flip should not tear, IIUC12:28
freemangordonbut it does12:28
freemangordonwhich means page flip is somehow broken, thus omapdrm has issues12:28
freemangordondoes that make sense to you?12:29
Wizzupwell I know that modesetting is not considered tear free on I think any device12:35
freemangordonWizzup: but, why should it tear with compositing WM, like h-d?12:43
freemangordonI mean, rendering happens like (with 2-buffer):12:44
freemangordon1. back buffer is being rendered12:44
freemangordon2. page flip is queued for that back buffer12:44
freemangordon3. page flip happens12:44
freemangordon4. old back buffer is now front, old front is now back and is given to GPU to render the next frame12:44
freemangordonthere is simply no way front buffer to be rendered to12:45
freemangordonWizzup: also, on PP there is absolutely no tearing12:45
freemangordonmodesetting/glamor12:46
uvoscompositing window managers sould be able to prevent tearing on modesetting without tear free, yes12:46
uvosunless you have multiple monitors12:46
uvosor your talking about x draw call tearing12:46
freemangordonit seems Wizzup is talking about that,yes12:47
freemangordonthere I agree we cannot prevent tear12:47
freemangordonbut with h-d there should be none12:47
freemangordonand at least on PP this is what happens12:47
Wizzupfreemangordon: hm ok wrt compositing12:49
freemangordonuvos: tmlind: do you know why te_support is false in droid4 panel data?12:49
Wizzupfreemangordon: hm @ pp and tearing, I need to check12:49
freemangordonI already did, but yes, please do12:49
freemangordonwith latest fixes that is (clutter)12:50
uvosfreemangordon: no12:50
Wizzupok12:50
freemangordonuvos: so, panel driver in vendor kernel should be panel-mapphone.c, right?13:00
uvosfreemangordon: i would say so, but cant confirm since uvos.xyz is currently down and i cant find where i have the dmesg from stock kernel locally13:05
freemangordonok. looke tear-free is enabled there13:06
freemangordon*looks13:06
freemangordongoing to enable it in our kernel and test13:07
freemangordondoesn;t look good13:17
freemangordon"DSI: omapdss DSI error: TE not received for 250ms!"13:17
rafael2kI come back later13:41
lelIMbackK closed an issue: https://github.com/maemo-leste/bugtracker/issues/547 (leste-config: configure mce battery-guard module (needs new mce as well))14:02
uvosso i managed to make cpcap-charger charge reliably even on the most crapy of cables17:01
uvosby having it ramp up current slowly instead of instantly switching on charging17:01
Wizzupmakes sense17:01
Wizzupthat's neat17:01
freemangordonhmm, tear-free support in omapdss is FUBAR :(17:09
uvoscomplain on linux-omap i gues17:09
freemangordonI don;t really get why it is so hard to keep what is already working when new things are implemented17:09
freemangordonyeah17:09
uvosthere was some pretty big refactoring to dss in 5.10 or so17:10
freemangordonI would say it has never been implemented17:10
uvosmaybe check if worked before then (if its easy)17:10
uvosfreemangordon: ok17:10
freemangordonI think it was working in omapfb17:10
freemangordonbut not in omapdrm17:10
freemangordonactually, there seems to be some TE support17:11
freemangordonbut, it seems it is supported only the gpio variant17:11
freemangordonbut not fully17:12
freemangordonbecause, for example n950 has te-gpios in DTS17:12
freemangordonbut panel data says te_enabled=false17:12
uvosappearntly the CPCAP_REG_CRM_CHRG_LED_EN register bit is load bearing17:33
uvos:P17:33
uvos(charging fails entirely if you try and use it to disable the light)17:34
uvoswierd17:34
tmlinduvos: that's weirde19:27
freemangordontmlind: there is a note in omapdrm framedone handler, that we have framedone irq when "DISPC has finished sending pixels"19:40
freemangordonbut: "However, DSI still has the pixels in its buffers"19:41
freemangordonso, if I read that correctly, we cannot use framedone as vsync19:41
freemangordonOTOH tear-free is enabled in vendor kernel19:42
freemangordonI tried to enable it in our, but it leads to lots of errors19:42
freemangordonobviously I am missing something19:42
tmlindfreemangordon: ok no idea about that one19:47
_inkyWizzup: i updated pinephone from devel, and wow, you made a lot of progress. there are still problems, but it's now much much better.21:29
Wizzupyes the other problem is a lima bug21:30
_inkymahjong is playable, keyboard almost always works, but sometimes some buttons "stuck", like on real keyboard, when it went down but doesn't come back, so in our case keyboard's key can remain in blue color, instead of getting the default color. also after some time it can again become blue.21:30
_inkythe other problem that i see is with my lightmeter application: when you press on a popup-menu with list of aperture or shutter settings21:31
_inkyit often reappears again when not pressed, just old memory.21:31
_inkyso the video driver is haunted by old memories. (:21:31
Wizzupyes, swap buffer problem21:32
_inkyis that the same as lima bug?21:32
Wizzupyes21:32
_inkyis lima driver open, does anyone work on it? did you try latest version? i remember sxmo was perfect though it was x.21:32
_inkysorry i know you put a lot of work.21:33
_inkybut i came to express the gratitude.21:33
_inkyso thank you this changes many many many things.21:33
_inkyi'll try now to use it on a daily basis under maemo.21:33
_inkyit's a pity, megapixels (camera) app uses gtk4, and megapixels-compat (just tried it, i have a build in my home) is first of all, rotated21:34
Wizzupyeah21:34
_inkyand secondly it updates the screen slowly.21:34
_inkybut i guess it is possible to make a photo with some effort. i'll experiment further.21:34
_inkyi meant by the question, did you reach the deadend already, or there is hope for further improvement?21:35
WizzupI need to file a bug with lima21:35
_inkygreat.21:35
WizzupI have the trace, I just need to read back what fmg wrote to me21:35
uvossxmo likely works becaue it dosent comopse via gl21:35
_inkyyes, true.21:35
_inkybut this is much much much much better already.21:36
uvosyou can suspend composistion on windows that give you trouble as a workaround until upstream hopefully fixes the issue21:36
_inkyi'll try frozen-bubble as well, it was a disaster before.21:36
uvosnot sure how you play forzen-bubble on pp. otg keyboard?21:37
_inkyyes, with bt kbd.21:37
_inkyi wanted to make a video comparing frozen-bubble on droid4 and pp.21:37
_inkydino xmpp client has a 'handy' branch with small screen optimized interface, plus they added voice and video calls.21:38
_inkyi'll build it, if it turns out well, will add it to maemo repos.21:38
uvosyeah would be great to get libhandy applicaitons to work on maemo21:38
uvosbut they need newer gtk3 etc21:38
_inkyhmmmm. didn't think of it. didn't think dino uses libhandy.21:38
_inkyi think it doesn't.21:38
uvosand also plamo applicaitons21:38
_inkynot sure. will find out.21:38
_inkyplamo?21:39
_inkyplasma?21:39
uvosplasma mobile21:39
_inkyah.21:39
_inkymaemo feels (even with the breaks it has currently) more lightweight than phosh.21:40
_inkythat is also why it is more attractive to me. also i don't like phosh' interface: the minimized windows on the top need to be scrolled through to choose a window to push out.21:41
uvosphosh was terribly laggy on xt1602 when i tried it21:41
_inkyin maemo all minimized programs are visible in one screen and it is very cool.21:41
uvosbut plamo was fine21:41
_inkyfrom what i see (didn't install plamo myself, but my two friends have pinephone) phosh is much faster than plasma.21:42
_inkyi think i've heard that's because plamo has a lot of js running.21:42
_inkyon the other hand, one of my friends uses sailfish on pinephone, and it is pretty smooth and feels fast.21:42
uvosyeah but the fact that hildon is almost terrible in portrait is a problem imo since you cant get to the buttons at the top one handed (unless you have the hw buttons like d4)21:43
uvosi still want to fix that21:43
_inkynothing is ever as slow as with phosh.21:43
uvosotherwise hildon is pretty good ui wise21:43
_inkyyes, right?21:43
_inky(:21:43
_inkybtw, on ui, dino mobile put the 'back' button to the left upper corner of the ... of itself.21:44
_inkythat's the most far place for the fingers to reach.21:44
uvosios dose this too21:44
_inkyi noticed today maemo places the similar button at the right corner.21:44
uvosits idiotic21:44
_inkynever used ios, but wow, i thought it is well designed.21:45
uvoson modern phones this means the back button is like a km awayy21:45
uvoswell to be fair they have gesture navigation now21:45
uvosso its mitigated21:45
_inkytoday i received a battery for my wife's google nexus s, i know postmarket supports it, probably will install sxmo on it to test how is it. it was also supported by SHR distro, i remember, but it's very very very old now.21:46
_inkySHR was for openmoko phones.21:47
_inkyand is currently dead. they still have the site i guess.21:47
_inkyi hope.21:47
_inkyit used enlightenment toolkit.21:47
uvosi dont think messing with shr is usefully anymore21:47
_inkyme too, i'll try sxmo. but i'm afraid it'll be anyway too heavy for that device.21:48
_inkyit is comparable with droid4, i guess.21:48
uvosno21:48
uvosgalaxy nexus is comparable to d421:48
uvosnexus s is more d221:48
_inkyoh. thanks!21:48
uvosnexus s is 1 core cortex-a821:49
_inkyi also guess, pinephone under maemo doesn't go to sleep, that's why it will eats more battery power. but not sure, now after upgrade how is it.21:49
uvosie /4 cpu performance of d4 just about21:49
_inkywow!21:49
_inkyi have a fibbonacci benchmark that works on one cpu21:50
_inkyi'll try. surprisingly pinephone gave me about same performance (almost) like xperia xa2 with sailfish.21:50
_inkyand how would you compare pinephone? i don't understand its hardware.21:50
_inkyi understand more or less quallcom chipsets as reference point.21:51
_inkyis pinephone a 4cpu version of droid4? in terms of cpu power?21:51
uvosno idea what soc dose it have?21:51
_inkyminute21:51
_inkyit doesn't list soc, it lists several cpus, a something21:52
uvosits lots faster than d4 for sure21:52
_inkylet me find21:52
siceloi love enlightenment stuff, and can't wait till they're fully wayland)21:52
_inkyseriously, huh? i don't understand this wayland anyway. is it really that better than xorg? i love maemo for being able to use ssh forwarding with it.21:54
uvosso pp soc is really close in perf to MSM891621:55
uvosif that tells you anything21:55
uvossince both are 4x Cortex-A5321:55
uvosso same as xt160221:56
_inkyuvos: Quad-Core Allwinner A64 @ 1.152 GHz21:56
_inkyyes, you found it.21:56
uvoswayland is better than x1121:56
_inkyso it's like snapdragon 41021:57
uvoswhile the wayland display servers/wms are not really better than xorg and often worse21:57
uvosyeah snapdragon 410 is marketing speak for MSM891621:57
_inkyi think i can tolerate it on xorg. well, i patched that squeeekboard with armenian layouts, it is really very useful.21:58
_inkyand the fact that vkbd opens a separate window to type freaks out my friends from maemo.21:58
uvoswith good reason the vkb is bad21:58
_inky"i think i can tolerate it on xorg" i meant "i think i can toleraty wayland on phone"21:59
_inkyit is so discouraging that programs written for sailfish is very hard to port to other linuxes, authors tell me they need to rewrite ui from scratch.21:59
uvosthis is a problem that fremantle/leste applications also have sadly22:00
uvosless, but still22:00
_inkyand though sailfish is really nice, and lightweight, i don't feel like putting efforts in to it.22:00
_inkybut with fremantle/leste it is much better.22:00
uvosthe mobile toolkit space on linux suffers from crippling fragmentation22:00
_inkyi write on my computer, then build on device and it works.22:00
_inkybut i agree, i understand what you usually say very well.22:01
_inkyi also would like hildon to be something like gnome or phosh, installable on any distro.22:01
siceloit can, if someone's willing to package it there ... pmOS packaged it at some point22:03
siceloand it was in official debian repos even in fremantle days22:04
_inkyi remember about pmos, i think it rings a bell about debian, but wasn't it more like diablo days? whatever, yes, that would be cool.22:04
sicelomaybe. i know there used to be hildon packages in debian, although i never used them myself22:08
Wizzup21:49 < _inky> i also guess, pinephone under maemo doesn't go to sleep, that's why it will eats more battery power. but not sure, now after upgrade how is it.22:09
Wizzupunchanged22:09
Wizzupwe need to think about how/if we want to integrate those sleep hacks22:09
uvosprobubly a button in power menu: sleep22:09
Wizzupor do it automatically when locked22:10
uvosand a quirks-pinephone mce module that makes it happen22:10
uvoshmm what about listening to music or downloading files compileing etc?22:10
uvosdosent pinephone need to suspend?22:10
Wizzupuvos: yes, so none of that will work22:12
Wizzupor it needs like clever hacks22:12
Wizzupbut it can never work well22:12
WizzupI don't want to add wake locks or anything like that atm22:12
uvosso "or do it automatically when locked isent an option"22:12
uvosand failing android style wake locks  only a sleep button remains22:13
siceloisn't a wake lock a mechanism to *prevent* sleeping/idling?22:16
uvosno22:16
uvoswell yes but no22:16
uvos:P22:16
uvosthe wake locks on android tell the os that it can suspend to ram, (if the hw needs it for pm its not used on d4 for instance or on xt1602) if none are held22:17
uvosone could for instance add a .desktop prop that tells hildon it can infrom mce that we cant suspend22:19
uvosand suspend in all cases where the display is off and no applicaiton with sutch a .desktop file is open22:19
uvoseffectively creating android style wakelocks on lesste22:19
uvosthis would work pretty well (not nearly as good as android but still)22:20
_inkyso do you know how maemo fremantle or sailfish work, they receive jabber messages immediately, they see that the contact changed status immediately.22:55
_inkyand still the battery works very well.22:55
_inkypinephone sleeps under pmos, so when i start using it, dino reconnects, and syncs messages.22:56
_inkyi guess sailfish and maemo fremantle don't sleep.22:56
_inkyyou said the os can put some apps to sleep depending on .desktop file.22:57
_inkybut whole os doesn't sleep right?22:58
Wizzupon pp it has to to save battery23:05
sicelo_inky: fremantle doesn't suspend-to-(ram/anything else) in the way pmOS and others do for pp (hence Wizzup's comments earlier). instead, runtime power management is used. pmOS recently started documenting something along those lines here, https://wiki.postmarketos.org/wiki/Runtime_Power_Management23:06
_inkyhmmm. but i don't understand what does it mean it needs to suspend?23:12
_inkyjust curious.23:12
_inkywhy it needs to? because of the some features of chipset?23:12
_inkyok let me see the wiki.23:13
siceloyes soc must support it in mainline. seems TI OMAP have some of the best support for that kind of power management, as far as linux kernel is concerned. even there making it work reliably isn't a walk in the park (e.g. for N900 it's currently not very trivial)23:17
_inkythe device lacks untime power management with mainline kernel.23:17
_inkydidn't know that, now it is more clear.23:18
sicelo:-)23:18

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!