freemangordon | Wizzup: Linux devuan-droid4 5.15.2 #1 SMP PREEMPT Wed Dec 15 00:29:28 UTC 2021 armv7l GNU/Linux | 00:03 |
---|---|---|
freemangordon | hangs as soon as pvr is inited | 00:03 |
uvos | buZz: whats wrong with the osk? | 00:06 |
buZz | oh eh, i was typing on the hw keyboard, it still appeared :) | 00:07 |
uvos | it should not | 00:08 |
uvos | did you build it yourself? | 00:08 |
freemangordon | buZz: what device is that? | 00:08 |
uvos | freemangordon: this is not him | 00:09 |
uvos | freemangordon: the application has a built in keyboard | 00:09 |
freemangordon | ah | 00:09 |
uvos | but i ported it to ask mce | 00:09 |
uvos | if its neede | 00:09 |
uvos | d | 00:09 |
freemangordon | ok | 00:09 |
freemangordon | uvos: btw, we made some progress with lima | 00:09 |
uvos | ok | 00:09 |
uvos | ill read backscroll later | 00:09 |
freemangordon | ok | 00:10 |
buZz | freemangordon: droid4 | 00:10 |
uvos | oh btw are you on devel? | 00:10 |
buZz | uvos: eh , no i just installed it from appstore | 00:10 |
uvos | devel or stable | 00:10 |
buZz | i might be, i keep forgetting to switch back , lemme check | 00:10 |
buZz | on stable | 00:11 |
uvos | ah ok | 00:11 |
freemangordon | night guys! | 00:11 |
buZz | nn freemangordon | 00:11 |
uvos | your mce it too old | 00:11 |
uvos | gn8 | 00:11 |
buZz | well, cool stuff :) its a pretty fast browser | 00:12 |
buZz | sadly webgl wasnt working :P | 00:12 |
uvos | its not accelreated | 00:12 |
uvos | at all | 00:12 |
uvos | is all sw rendering | 00:12 |
buZz | isnt it just chrome's engine underneath? | 00:13 |
uvos | a fork of it yeah qwebengine | 00:13 |
uvos | but its broken on pvr | 00:13 |
uvos | it fails to compile a shader | 00:13 |
buZz | ah hm | 00:13 |
buZz | i -think- i had functional webgl on droid4 at some point, slow, but functional | 00:14 |
uvos | firefox works fine | 00:14 |
uvos | and is accelerated | 00:14 |
uvos | dunno of webgl works | 00:14 |
uvos | it should really | 00:14 |
buZz | i'll try :) | 00:14 |
buZz | ooo, maybe troubled because of having gl4es installed :D | 00:29 |
mighty17[m] | Any clues why this happens, when I cold boot my mmc5 node (ie sdio for wifi) doesn't show in dmesg, and wifi doesn't work as well, but if I boot from twrp into mainline (ie reboot to system option) mmc5 and wifi works | 06:07 |
tmlind | Wizzup, freemangordon: not sure if i understand the lima thing.. but noticed this commit is referenced also in a wlroots bug lima: fix xserver page flip fail for full screen client | 07:51 |
tmlind | mentioned in this bug: https://github.com/swaywm/wlroots/issues/2795 | 07:52 |
freemangordon | tmlind: clutter was not updateing damaged areas correctly | 07:53 |
freemangordon | not related to xserver or lima | 07:53 |
freemangordon | but that's only part of the story | 07:53 |
freemangordon | lima still has issues | 07:53 |
tmlind | ok | 07:54 |
freemangordon | in the meanwhile something got broken on my d4 :( | 07:54 |
freemangordon | as soon as I try to init pvr device hangs | 07:54 |
tmlind | heh ok | 07:54 |
tmlind | hmm i don't think we have EGL_KHR_partial_update, do we? or is there something similar | 07:55 |
tmlind | freemangordon: maybe you're trying to init pvr for wrong hardware? check the pvr env variables | 07:56 |
freemangordon | tmlind: no matter, it should not hang the device | 07:57 |
freemangordon | and it was working since yesterday | 07:58 |
freemangordon | when I did dist-upgrade | 07:58 |
freemangordon | oh, maybe libs are broken? | 07:58 |
freemangordon | hmm, why such a big diff in dtb size between what I have here and repo kernel? | 07:59 |
freemangordon | this is the last resort, copying kernel and modules that I used didn;t help | 07:59 |
tmlind | you can decompile the dtb to check with dtc -I dtb -O dts your.dtb | 08:00 |
freemangordon | mhm | 08:00 |
freemangordon | but lets see first if using my dtb helps | 08:01 |
freemangordon | no! | 08:01 |
freemangordon | wtf? | 08:01 |
freemangordon | ok, sgx user libs then | 08:01 |
tmlind | hmm looks like EGL_KHR_partial_update is listed for series6 only in PowerVR.Supported+Extensions.OpenGL+ES.EGL.pdf | 08:02 |
freemangordon | why do you look at that? | 08:03 |
tmlind | just wondering if the wlroots bug above is related to the black rectangles | 08:05 |
tmlind | the missing flush bug | 08:05 |
freemangordon | yep, seems ddk blobs in -experimental are broken | 08:05 |
tmlind | wrong commit or wrong env? | 08:06 |
freemangordon | no idea | 08:06 |
freemangordon | but if I LD_LIBRARY_PATH to the local blobs dir I keep on device, init no longer hangs | 08:07 |
tmlind | ok | 08:09 |
freemangordon | ugh | 08:11 |
freemangordon | this is for wrong device | 08:11 |
freemangordon | -rw-rw-r-- 1 root root 99776 Oct 26 16:16 libsrv_init.so.1.17.4948957 is in my blobs dir | 08:11 |
freemangordon | -rw-r--r-- 1 root root 103736 Dec 15 01:36 /usr/lib/arm-linux-gnueabihf/libsrv_init.so.1.17.4948957 is installed with sgx-ddk-um-ti343x | 08:12 |
freemangordon | wait, what?!? | 08:12 |
freemangordon | sgx-ddk-um-ti343x | 08:12 |
tmlind | heh | 08:12 |
freemangordon | why is this installed on d4?!? | 08:12 |
freemangordon | Wizzup: xserver-xorg-video-omap most-probably is being held because of a wrong dependency to sgx-ddk-um-ti343x | 08:13 |
freemangordon | https://pastebin.com/fvwZJTaR | 08:15 |
tmlind | freemangordon: so was the solution to the clutter issue to update clutter or some commit? just wondering.. | 08:22 |
freemangordon | update clutter | 08:22 |
tmlind | ok nice | 08:23 |
freemangordon | update, like enable support for double-buffering | 08:23 |
tmlind | ok | 08:23 |
freemangordon | https://github.com/maemo-leste-upstream-forks/clutter-0.8/commit/13903d341009266d0bfa19806e74625a16ab552a | 08:23 |
freemangordon | the code was there but disabled by Nokia, with a big fat warning to re-enable it when needed :) | 08:24 |
tmlind | heh ok :) | 08:24 |
tmlind | so did that fix the missing flush issue? | 08:24 |
freemangordon | umm... what issue? | 08:25 |
tmlind | the black squares on some frames issue | 08:25 |
freemangordon | this was fixed in ddx | 08:25 |
tmlind | oh, i've missed that, so what change fixed that? | 08:26 |
freemangordon | waiting sgx to finish rendering before doing flips | 08:26 |
freemangordon | lemme find the commit | 08:26 |
tmlind | ok thanks nice if it's fixed finally | 08:26 |
tmlind | yeah commit would be nice to see for longer term memory to revolve the issue :) | 08:27 |
tmlind | to resolve the issue | 08:27 |
tmlind | trauma therapy kind of thing.. | 08:27 |
freemangordon | this https://github.com/maemo-leste/xf86-video-omap/commit/12836b9d1b03e5a9de0b89cba727c801f74b4184 | 08:27 |
freemangordon | magic happens in waitForBlitsCompleteOnDeviceMem | 08:28 |
freemangordon | the only remaining issues are correct 3-buffer implmentation and tearing | 08:30 |
tmlind | hmm ok yeah i recall seeing this but never figured out where exactly it ends up waiting.. | 08:30 |
freemangordon | https://github.com/maemo-leste/xf86-video-omap/blob/maemo/beowulf-experimental/src/omap_exa_pvr.c#L159 | 08:31 |
tmlind | ok thanks, yup you already explained things for me earlier from there | 08:35 |
tmlind | looks like simple-dmabuf-egl.c in wait_for_buffer_release_fence() uses EGL_KHR_wait_sync, i guess that's how it ensures updates.. | 08:48 |
freemangordon | yes | 08:49 |
freemangordon | well, I guess yes | 08:49 |
tmlind | oh but EGL_KHR_wait_sync is listed series6 only? | 08:49 |
freemangordon | no, why? | 08:49 |
freemangordon | our driver reports it supported, iirc | 08:50 |
freemangordon | lemme check | 08:50 |
freemangordon | (d4 boots...) | 08:50 |
tmlind | hmm so do we get "Warning: EGL_KHR_wait_sync not supported" with weston-simpel-dmabuf, let me check.. | 08:50 |
tmlind | i'm only seeing explicit_synchronization warnings | 08:51 |
freemangordon | I see EGL_KHR_wait_sync as suported extension | 08:54 |
freemangordon | *supported | 08:54 |
freemangordon | https://pastebin.com/SLD8wvd0 | 08:54 |
tmlind | ok | 08:55 |
tmlind | hmm so is the conclusion that the client's responsibility to call EGL_KHR_wait_sync, basically the topmost app (or xserver related stuff) with black squares issue? | 08:59 |
tmlind | hmm i'm seeing also a maximized weston-terminal suffer from some areas not being updated when switching to it, so maybe there still is some driver issue too | 09:02 |
tmlind | not sure if i ever saw this issue with weston, not that i did much testing either | 09:04 |
freemangordon | tmlind: I am not sure I get it - do you still see issues with xserver applications? | 09:07 |
* freemangordon wonders what is the difference between EGL_NOKIA_texture_from_pixmap and EGL_NOK_texture_from_pixmap | 09:08 | |
freemangordon | TBH I am not sure how usable EGL_KHR_wait_sync is | 09:09 |
tmlind | freemangordon: so my test case is slow down pvr, then flip sway desktops with termite running.. but maybe that uses xwayland and i really should test that with weston-terminal for example. sounds like you got it fixed for xorg | 09:18 |
tmlind | freemangordon: so i presume things work now fine for you also with sgx clock slowed down with no black squares? | 09:20 |
freemangordon | mhm | 09:21 |
tmlind | ok great | 09:21 |
tmlind | hmm so for my test case, maybe "lima: fix xserver page flip fail for full screen client" is really the right fix if termite uses xwayland, or if switching from any xwayland app to let's say weston-terminal | 09:26 |
freemangordon | why lima? | 09:26 |
freemangordon | I mean - that should not affect pvr, IIUC | 09:27 |
tmlind | no idea.. but it sounds like it's the previous app that needs to wait for gpu render to complete before the switch to any other app happens | 09:27 |
freemangordon | sure | 09:27 |
freemangordon | well, actually it should wait before the flip | 09:27 |
tmlind | so i tried flipping between two desktops with a weston-terminal only in each desktop and so far have not been able to trigger the artifacts | 09:28 |
freemangordon | flip == "drm page flip" | 09:28 |
* freemangordon tries EGL_NOK_texture_from_pixmap | 09:28 | |
tmlind | ok so the wait before the flip may only be missing for my test case when switching from some xwayland app to some other app.. well at least i need to test that | 09:31 |
tmlind | hmm but then why do i see artifacts when switching from weston-terminal to termite? if it's up to weston-terminal to wait for gpu to complete before flip, and assuming it's doing the right thing.. | 09:33 |
tmlind | sounds like i need to test with xwayland disabled at least | 09:33 |
tmlind | seeing artifacts also between two weston-terminal windows as long as one is maximized so they look different, the artifacts seem very small though just maybe 10x20 pixels at the top left corner in portrait mode | 09:38 |
tmlind | mode like p shaped small artifact with about 10x10 pixels top part and 5x10 part bottom.. | 09:39 |
tmlind | oh well so much for that theory | 09:40 |
freemangordon | ugh, we have a glmark fps regression | 09:42 |
freemangordon | with glmark2-es2-drm that is | 09:42 |
freemangordon | tmlind: how to check the clocks? | 09:43 |
freemangordon | glmark2 Score: 37 | 09:46 |
freemangordon | on d4 | 09:46 |
freemangordon | wtf? | 09:46 |
tmlind | freemangordon: the d4 sgx clock slow down trick is rwmem 0x4a008164=0x1f, let me check the value i have by default | 09:47 |
freemangordon | 225 | 09:47 |
freemangordon | it is ok | 09:47 |
freemangordon | something else slows it down | 09:47 |
freemangordon | maybe mesa | 09:48 |
tmlind | yup 0x225 is the default value | 09:48 |
freemangordon | lemme try with my kernel first | 09:49 |
sicelo | tmlind: i *think* termite is not xwayland, nor is weston-terminal | 09:50 |
sicelo | yes, my d4 3d accel has been slow since last update (as well as that ti343x issue) | 09:51 |
tmlind | sicelo: right, but it occured to me that if termite was using xwayland then the issue should not happen with weston-terminal.. but the issue is there for sure too with weston-terminal | 09:51 |
tmlind | i think termite can use both xorg or wayland depending on the env | 09:52 |
sicelo | possible. when i used sway and weston on N900 (where i used termite), i was specifically disabling xwayland support ... | 09:54 |
sicelo | (i thought xwayland would add overhead for N900) ;-) | 09:55 |
freemangordon | ok, something in the kernel | 10:02 |
freemangordon | with 'my' kernel fps are back to normal | 10:03 |
freemangordon | it is kernel config though, not kernel as such | 10:03 |
sicelo | what's the offending config? | 10:11 |
freemangordon | no idea | 10:12 |
freemangordon | it is just that kernel in repos is build with different config | 10:12 |
tmlind | some debug option? | 10:22 |
freemangordon | no idea | 10:23 |
freemangordon | lemme compare | 10:23 |
tmlind | make savedefconfig for both, then diff -u | 10:23 |
freemangordon | umm... I will diff /proc/config.gz stuff | 10:24 |
freemangordon | https://pastebin.com/A29jUqtK | 10:26 |
freemangordon | I would bet on CONFIG_COMPACTION=y | 10:29 |
freemangordon | missing in repo kernel | 10:29 |
freemangordon | so device idles faster | 10:29 |
freemangordon | tmlind: what do you think^^^? | 10:29 |
freemangordon | hmm, this is crazy | 10:35 |
freemangordon | after a reboot fps is back to normal | 10:35 |
freemangordon | sicelo: how do you know 3d is slower for you? | 10:36 |
sicelo | e.g. switching between desktop is noticeably laggier than before | 10:38 |
sicelo | *desktops | 10:38 |
freemangordon | what desktops? | 10:38 |
sicelo | desktop 1, desktop 2, etc. not sure how to describe them :-) | 10:39 |
freemangordon | in hildon-desktop or where? | 10:40 |
sicelo | yes, h-d | 10:40 |
freemangordon | how do you "switch"? | 10:40 |
freemangordon | scroll? | 10:40 |
sicelo | desktop = where desktop widgets stay in h-d | 10:40 |
freemangordon | ok, but how do you "switch"? | 10:41 |
sicelo | so you switch by swiping desktop left or right | 10:41 |
freemangordon | ok, so you "scroll" | 10:41 |
freemangordon | :) | 10:41 |
sicelo | hehe, | 10:41 |
freemangordon | hmm, this is weird, this should render with 54 fps | 10:41 |
freemangordon | do you want to measure the fps? | 10:42 |
sicelo | i can test in a proper way, if you tell me how to | 10:42 |
freemangordon | ok | 10:42 |
sicelo | otherwise nearly every interaction definitely lags for me | 10:42 |
freemangordon | /usr/sbin/dsmetool -k /usr/bin/hildon-desktop | 10:42 |
* sicelo boots leste | 10:42 | |
freemangordon | CLUTTER_SHOW_FPS=1 maemo-summoner hildon-desktop.launch | 10:43 |
freemangordon | but before that, which xorg driver is in use? | 10:43 |
freemangordon | you said you had issues with pvr blobs package | 10:44 |
freemangordon | how did you solve that? | 10:44 |
sicelo | mmm, xorg not starting after my last tries :-P | 10:45 |
freemangordon | sicelo: sudo apt-get install xserver-xorg-video-omap=0.5.0+2m7.4 | 10:47 |
freemangordon | ugh | 10:48 |
freemangordon | the one in -devel is broken | 10:48 |
freemangordon | lemme fix it | 10:48 |
tmlind | freemangordon: weird so if CONFIG_COMPACTION=y makes the fps go back to normal, we need to add some pmqos limit somewhere | 10:54 |
freemangordon | tmlind: no, it was a reboot that made them go back to normal | 10:55 |
freemangordon | no idea what was it | 10:55 |
freemangordon | the same kernel now gives glmark 62 | 10:55 |
tmlind | oh ok even stranger | 10:56 |
freemangordon | yeah | 10:56 |
Wizzup | freemangordon: it should not pull ti343x for sure | 10:59 |
sicelo | freemangordon: heh, fps between 2 and 5. i guess my accel is completely broken then? | 11:00 |
freemangordon | yeah, the issue was that -experimental was not pulled in -devel | 11:00 |
freemangordon | sicelo: sure | 11:00 |
freemangordon | wait 5 minutes | 11:01 |
freemangordon | sicelo: CI is building it | 11:01 |
Wizzup | freemangordon: wait you need CONFIG_COMPACTION for higher fps?? | 11:01 |
freemangordon | no | 11:02 |
freemangordon | I was thinking that was the case, but after a reboot fps is back to normal | 11:02 |
freemangordon | NFC what was that | 11:02 |
freemangordon | Wizzup: so, I got low fps with kernel in the repo | 11:02 |
freemangordon | then I boot my zImage which differs by only config options with the repo one | 11:03 |
freemangordon | and it performed just fine | 11:03 |
freemangordon | then I compared configs and the only suspicious one was CONFIG_COMPACTION | 11:03 |
freemangordon | then I boot repo kernel again, but this time fps were ok | 11:03 |
freemangordon | so false alarm | 11:04 |
Wizzup | freemangordon: maybe it's me enabling cma | 11:05 |
Wizzup | ok | 11:06 |
freemangordon | no, it works fine ever since | 11:06 |
Wizzup | ok | 11:06 |
Wizzup | freemangordon: do you have leste-config-droid4 and hildon-meta-droid4 | 11:06 |
freemangordon | I think yes | 11:06 |
Wizzup | that is maybe why you got ti343x | 11:06 |
Wizzup | ok | 11:06 |
Wizzup | (if you don't have it) | 11:06 |
freemangordon | no, I get it because of omap xorg driver | 11:06 |
Wizzup | ok | 11:07 |
freemangordon | the version in -devel is higher than the version in -experimental | 11:07 |
freemangordon | but code-wise it is older | 11:07 |
Wizzup | I'll check it out, I'm trying to get rid of a headache this morning :D | 11:07 |
Wizzup | freemangordon: oh so I forgot to bump | 11:07 |
freemangordon | no need (to check out) I am already on it | 11:07 |
freemangordon | Wizzup: no, you forgot to pull | 11:07 |
Wizzup | ok | 11:07 |
Wizzup | weird | 11:07 |
freemangordon | experimental in devel | 11:08 |
Wizzup | the branches here all point to the same commit | 11:08 |
Wizzup | [maemo/beowulf-devel] [maemo/beowulf-experimental] {origin/maemo/beowulf-devel} {origin/maemo/beowulf-experimental} <0.5.0> PVREXA: add some more framebuffer flush points | 11:08 |
freemangordon | no idea, bu on github devel was 2 commits behind experimental | 11:08 |
Wizzup | hm | 11:08 |
freemangordon | maybe you didn;t push | 11:08 |
freemangordon | Wizzup: no, you didn;t pull | 11:09 |
freemangordon | your code lacks 2 commits | 11:09 |
freemangordon | one of them being https://github.com/maemo-leste/xf86-video-omap/commit/6a1eb3bf227ff69326895d11fdcf44ccb6dbca7c | 11:09 |
Wizzup | ok | 11:09 |
Wizzup | pls fix :) | 11:09 |
freemangordon | already did | 11:10 |
freemangordon | building atm | 11:10 |
freemangordon | sicelo: do you have -experimental enabled? | 11:15 |
sicelo | yes | 11:15 |
freemangordon | apt update/upgrade then | 11:15 |
sicelo | on it | 11:16 |
sicelo | done. still getting 5fps. let me reboot perhaps | 11:17 |
freemangordon | sicelo: also, make sure you have meta installed | 11:18 |
freemangordon | leste-config-droid4 and hildon-meta-droid4 that is | 11:18 |
sicelo | right. hildon-meta-droid4 was missing | 11:19 |
freemangordon | sicelo: I would guess your xorg.conf is wrong or broken too, but lets see | 11:20 |
freemangordon | most-probably you use modesetting without glamor | 11:20 |
freemangordon | to have 2-5 fps | 11:21 |
sicelo | now seems snappy again :-) | 11:23 |
sicelo | let me try the fps readout | 11:23 |
freemangordon | :) | 11:23 |
freemangordon | should be 52-53 | 11:23 |
sicelo | what's the current xorg.conf file? i recall there were some changes in location, etc. | 11:26 |
freemangordon | /usr/share/X11/xorg.conf.d | 11:26 |
freemangordon | 99-omap | 11:26 |
freemangordon | why do you need that? | 11:26 |
freemangordon | there is nothing in there realyl | 11:26 |
sicelo | wanted to check because you said it's possible my xorg.conf is wrong or broken wrt modesetting/glamor | 11:28 |
freemangordon | no need | 11:28 |
freemangordon | if it is snappy, then it is using the correct xorg.conf file | 11:28 |
sicelo | i got 67 once, but mostly i'm in the 40s. i guess this is correct | 11:30 |
sicelo | thanks for the help. | 11:30 |
Wizzup | freemangordon: is the tearing fixed? | 11:31 |
freemangordon | no :( | 11:32 |
freemangordon | sicelo: hmm, no | 11:32 |
freemangordon | not correct | 11:32 |
freemangordon | you should get > 50 | 11:32 |
freemangordon | what version is xserver-xorg-video-omap ? | 11:33 |
sicelo | xserver-xorg-video-omap 0.5.1+2m7 | 11:33 |
Wizzup | freemangordon: so regarding lima, what are the next steps, I see you want me to query for the swap behaviour, and I should file a bug report with the trace and video | 11:36 |
Wizzup | what else, should we try the partial update khr extensions? | 11:36 |
Wizzup | enunes hinted at that | 11:36 |
Wizzup | also: | 11:37 |
Wizzup | If your UI application is compositing multiple uncompressed layers which make heavy use of transparencies, then using EGL_BUFFER_PRESERVED is probably a sensible thing to do. | 11:37 |
Wizzup | (from that mali blob) | 11:38 |
Wizzup | blog* | 11:38 |
Wizzup | it might make sense to use EGL_KHR_partial_update and EGL_KHR_swap_buffers_with_damage | 11:39 |
sicelo | i see it's offering me another update now, from devel, not experimental, https://paste.debian.net/1224713/ | 11:39 |
Wizzup | yeah see 11:07 < freemangordon> Wizzup: no, you forgot to pull | 11:40 |
freemangordon | sicelo: yeah, I built for both | 11:43 |
freemangordon | Wizzup: I don;t really see what will we gain from using those extension | 11:44 |
freemangordon | lima guys should just fix their driver | 11:44 |
freemangordon | so, what I think are the next steps - capture another trace and a video of it and open an issue attaching those | 11:45 |
freemangordon | this is without setting EGL_BUFFER_PRESERVED | 11:45 |
freemangordon | not sure about TFP, either should work | 11:46 |
freemangordon | ah, you said with TFP issue is not visible while capturing a trace | 11:46 |
freemangordon | so no TFP then | 11:46 |
freemangordon | sicelo: if you do left-right-left-... swipe really fast (on same desktop) you should hit 53 fps | 11:47 |
freemangordon | unless somehow 3-buffer is disabled in xorg.conf | 11:48 |
uvos | a transition is still the better test tho | 11:49 |
uvos | since the swipe fps depens on how fast you swipe | 11:49 |
uvos | and how many icons you have | 11:49 |
uvos | and on ts event rate | 11:49 |
freemangordon | uvos: sure, but it requires ini changes | 11:49 |
freemangordon | otherwise I agree | 11:49 |
uvos | right | 11:49 |
uvos | sicelo's 40 fps might be entirely reasonable depending on the above factors | 11:50 |
freemangordon | hmm, ok | 11:50 |
uvos | 40 ish is about what it get on a full desktop | 11:50 |
freemangordon | I would expect more from pvr540 :) | 11:50 |
uvos | icons wise | 11:50 |
uvos | me too | 11:50 |
uvos | android is locked 60 | 11:50 |
freemangordon | actually fps should not depend on the number of icons | 11:51 |
uvos | it dose tho | 11:51 |
freemangordon | or it does? | 11:51 |
freemangordon | ok | 11:51 |
sicelo | i think it's alright now. i just got 70 at some point | 11:51 |
freemangordon | that's weird as well | 11:51 |
sicelo | heh, what? the 70? | 11:52 |
freemangordon | mhm | 11:52 |
uvos | i get that too | 11:52 |
uvos | sometimes | 11:52 |
freemangordon | but how? | 11:52 |
uvos | no idea seams random | 11:52 |
freemangordon | I mean - it seems sometimes drm flip takes an instant | 11:52 |
freemangordon | instead of 6 ms | 11:53 |
uvos | right | 11:53 |
freemangordon | anyway | 11:53 |
sicelo | was scrolling between widget-less desktps. no idea | 11:53 |
uvos | but the scrolling dosent look 70 fps either way | 11:53 |
freemangordon | yes | 11:53 |
uvos | somewhere frames must be being discarded | 11:53 |
freemangordon | Wizzup: definitely there is some issue in lima and it has to be fixed, all devices I replay latest traces on show no issues | 11:55 |
freemangordon | uvos: mhm | 11:55 |
freemangordon | it is not smooth | 11:55 |
freemangordon | but I'll get back to it when we have the other issies fixed | 11:55 |
freemangordon | I would say it is related to tearing | 11:55 |
freemangordon | Wizzup: well, all devices are NV and Intel, but still | 11:56 |
tmlind | so is devel about to get updated then or a there still pending issues? | 12:01 |
tmlind | s/a/are/ | 12:01 |
sicelo | it's updated. at least i have the xserver-xorg-video-omap from devel now, and it's running correctly | 12:02 |
tmlind | oh ok nice, i'll try to find my test device and update it then :) | 12:02 |
* tmlind goes poking in a pile of cardboard boxes | 12:03 | |
Wizzup | freemangordon: ok, will go, bbiab | 12:06 |
freemangordon | tmlind: make sure you have meta packages installed if you played with it a lot | 12:13 |
tmlind | heh bionic just shuts down now with no screens found after upgrade :) | 12:17 |
freemangordon | so, do you have leste-config-droid4 and hildon-meta-droid4 installed? | 12:18 |
freemangordon | or this is diifferent device? | 12:18 |
tmlind | need to check | 12:19 |
tmlind | trying to boot to m-l emergency shell | 12:20 |
tmlind | uvos: that worked nicely :) | 12:21 |
freemangordon | not for me, it should not try to start powervr service ;) | 12:23 |
tmlind | freemangordon: i have leste-config-bionic instead of leste-config-droid4 | 12:25 |
freemangordon | I guess that's fine | 12:25 |
freemangordon | what about hildon-meta-droid4? | 12:25 |
freemangordon | or it should be hildon-meta-bionic? | 12:25 |
tmlind | yup i have hildon-meta-bionic | 12:26 |
freemangordon | that might be a problem | 12:26 |
freemangordon | uvos: Wizzup: is this updated? | 12:26 |
freemangordon | tmlind: so, Xorg does not start? | 12:28 |
freemangordon | may I have the log file | 12:28 |
freemangordon | or at least - what is the error there? | 12:28 |
tmlind | freemangordon: just a sec and i'll grep for EE | 12:28 |
tmlind | /var/log/Xorg.0.log or some other one? | 12:29 |
freemangordon | try to start Xorg | 12:29 |
freemangordon | and then provide /var/log/Xorg.0.log | 12:29 |
tmlind | just run Xorg manually in the emergency shell or some other way? | 12:30 |
freemangordon | yes, just run it | 12:30 |
freemangordon | tmlind: hmm, wait | 12:30 |
freemangordon | what is mesa version you have installed? | 12:30 |
freemangordon | d4 wants libegl-mesa0 (>= 21.2.5) | 12:31 |
freemangordon | and sgx-ddk-um-libs | 12:32 |
tmlind | dpkg -l | grep libgl only returns libegl1-sgx-omap4 | 12:33 |
freemangordon | yeah, that's the issue | 12:33 |
freemangordon | tmlind: please install libegl-mesa0 | 12:34 |
freemangordon | and sgx-ddk-um-libs | 12:34 |
tmlind | ok will do, need to start wpa_supplicant and wlan0 first | 12:34 |
Wizzup | tmlind: I did not test bionic yet with 5.15 | 12:34 |
freemangordon | and libegl1 libgles1 libgles2 | 12:35 |
Wizzup | maybe the pkg doesn't pull the sgx ddk stuff | 12:35 |
freemangordon | :nod: | 12:35 |
Wizzup | I kind of assumehoped that uvos tested it | 12:35 |
Wizzup | droid3 works with droid4 meta installed | 12:35 |
Wizzup | ... :) | 12:35 |
freemangordon | meta does not pull upgraded mesa/sgx libs | 12:35 |
Wizzup | yeah I installed the droid4 meta | 12:35 |
Wizzup | I can probably clean this up in ~20 mins or so | 12:35 |
freemangordon | please do when you have time, I am going on a walk | 12:36 |
freemangordon | ttyl | 12:36 |
tmlind | Wizzup: ok, any easy way to start wlan0 in emergency shell or just manually start wpa_supplicant and dhclient? | 12:36 |
freemangordon | maybe start icd2 | 12:37 |
tmlind | ok | 12:37 |
freemangordon | if you have autoconnect enabled | 12:37 |
freemangordon | not sure it will start though | 12:37 |
tmlind | yup | 12:37 |
freemangordon | but you can try | 12:38 |
freemangordon | I think you need those https://pastebin.com/p2zqd6JY | 12:38 |
tmlind | heh tries to start xorg and shuts down | 12:38 |
freemangordon | :( | 12:38 |
sicelo | i was starting wpa_supplicant by hand, then using wpa_cli | 12:39 |
freemangordon | I do that too when I have to recover | 12:39 |
freemangordon | but, I have conf file with my AP and password :) | 12:39 |
freemangordon | so the only thing I need is ctrl-R :p | 12:39 |
sicelo | ;-) | 12:39 |
sicelo | iwd people will say switch | 12:40 |
freemangordon | what to? iwd? | 12:41 |
sicelo | yeah :p | 12:41 |
freemangordon | never used, is it better? | 12:41 |
freemangordon | or some NIH syndrome? | 12:41 |
sicelo | ymmv. i'm using it now on my laptop. its cli is definitely easier/friendlier than wpa_cli. other than that, personally don't see much benefit | 12:42 |
sicelo | but they say power savings, making it more suitable for mobiles | 12:42 |
freemangordon | I see | 12:44 |
freemangordon | anyway, I am going on a walk, bbl | 12:44 |
Wizzup | sicelo: what power savings | 12:44 |
Wizzup | tmlind: fwi with -D you can start services without starting dependecies | 12:44 |
Wizzup | tmlind: so /etc/init.d/dsme start | 12:44 |
Wizzup | and then | 12:44 |
Wizzup | /etc/init.d/icd2 start -D | 12:44 |
Wizzup | (or just icd2 -l0 &) | 12:45 |
tmlind | ok | 12:46 |
tmlind | icd2 fails to starat with /tmp/session_bus?address.user not there | 12:49 |
sicelo | Wizzup: https://wiki.gentoo.org/wiki/Iwd (fast network discovery, and better roaming definitely NOT what i've experienced so far ... unless it's debian or hw issues that affect me) | 12:50 |
sicelo | as for the battery/power savings, "battery life (kernel crypto framework is more efficient than wpa_supplicant)" from https://gitlab.com/postmarketOS/postmarketos/-/issues/45 | 12:51 |
Wizzup | tmlind: /etc/init.d/dbus-user start | 12:52 |
tmlind | ok | 12:52 |
Wizzup | sicelo: seems mostly not relevant, that's only valid when actually doing lots of rx/tx | 12:52 |
Wizzup | tmlind: btw I assume your icd2 is set to automatically connect? | 12:52 |
Wizzup | tmlind: sorry for the pain btw, I'll try to fix the droid3 and bionic meta | 12:52 |
Wizzup | hm actually it looks like I did have the bionic3 already with some of these packages | 12:53 |
tmlind | no problem, just a pain to try to fix using the emergency shell | 12:56 |
tmlind | bad battery too, so uart is only somewhat usable.. | 12:56 |
tmlind | i'll let the battery charge up, then retry | 12:57 |
tmlind | Wizzup: now /etc/init.d/icd2 start -D produces openrc error bad variable name | 13:00 |
Wizzup | huh | 13:00 |
Wizzup | that sounds like something is corrupted almost | 13:00 |
tmlind | 1: export: directory:: bad variable name | 13:01 |
Wizzup | huh yeah that's weird.. | 13:02 |
tmlind | oh well i'll start it manually when i have a serial console and don't need to deal with the vkb | 13:04 |
Wizzup | mhm | 13:05 |
tmlind | uhh this xorg not working and shutdown when things go wrong is just annoying :( | 13:16 |
tmlind | can't even install packages without system shutting down automatically | 13:17 |
Wizzup | tmlind: can you check what sgx-ddk-um-ti is installed | 13:19 |
Wizzup | dpkg -l | grep sgx-ddk-um- | 13:19 |
Wizzup | tmlind: if you don't want it to reboot, just touch this file | 13:19 |
Wizzup | touch /etc/no_lg_reboots | 13:19 |
Wizzup | tmlind: ^^ | 13:19 |
tmlind | ok | 13:19 |
Wizzup | I guess for 'end users' it is probably better than phone reboots if X crashes | 13:20 |
Wizzup | for developers it's mostly annoying :) | 13:20 |
Wizzup | sicelo: did you see there are N810 patches for pwm and led on linux-omap? fun | 13:21 |
sicelo | yeah, i love that work :-) | 13:21 |
sicelo | interesting - when my xorg wasn't starting, i didn't get system shutting down. instead it just dropped me to a shell after a short while | 13:24 |
sicelo | ah, i do have no_lg_reboots, although i don't remember when i set it :p | 13:24 |
tmlind | heh started getting battery low warnings while using serial console but got the charger in in time :) | 13:33 |
Wizzup | in dmesg or from ui? | 13:33 |
tmlind | dmesg | 13:33 |
Wizzup | mhm | 13:33 |
tmlind | still in emergency shell and installing the needed packages now | 13:33 |
Wizzup | which ones were missing for you | 13:35 |
tmlind | it's pretty messed up right now, can't configure hildon-desktop because xorg still won't start, need to continue later, gotta go now ttyl | 13:48 |
Wizzup | tmlind: I'll check bionic, I just upgraded my droid3 and added the right meta and it works for me | 14:04 |
Wizzup | it seems to get stuck on reboot sometimes though (when shutting down) | 14:08 |
Wizzup | I bet this is related to the serial panics | 14:08 |
Wizzup | s/panics/oopses/ | 14:12 |
freemangordon | Wizzup: BTW, do we already support those tablets you gave me? | 14:54 |
freemangordon | what the? PP battery discharged while being out of the device | 15:02 |
freemangordon | same black screen as yesterday, it turned out GPU cooling fan had been stuck :) | 15:56 |
freemangordon | hmm, pinephone image does not boot | 16:01 |
freemangordon | Wizzup: ^^^ | 16:01 |
freemangordon | well, I get to login prompt, but can;t do much there | 16:03 |
freemangordon | hmmm | 16:19 |
freemangordon | false alarm | 16:19 |
freemangordon | I wrote the image for the second time and now it boots | 16:19 |
mighty17[m] | <mighty17[m]> "Any clues why this happens, when..." <- Bump :( | 16:23 |
freemangordon | hmm, very strange behaviour om PP with kernel from -devel | 16:39 |
freemangordon | brightness constantly changes | 16:40 |
freemangordon | and device seems to become very hot | 16:40 |
freemangordon | this is some 5.14-rc1 | 16:40 |
Wizzup | freemangordon: rafael2k is working on new kernel | 17:05 |
Wizzup | freemangordon: no, those tablets do not have working panel (bridge) yet | 17:06 |
Wizzup | freemangordon: the kernel is ok for me | 17:06 |
Wizzup | idk about als/brightness | 17:06 |
freemangordon | Wizzup: it is not als, it does it even before h-d has started | 18:19 |
freemangordon | at least that's my understanding | 18:19 |
Wizzup | freemangordon: als/mce runs before hd fwiw | 18:51 |
Wizzup | but yeah if it is before mce... | 18:51 |
freemangordon | I stopped mce and it continues | 19:19 |
freemangordon | blinking that is | 19:19 |
freemangordon | kernel issue, I downgraded to 5.10.0-1+2m7.1 in it is gone | 20:03 |
Wizzup | freemangordon: ok | 20:16 |
rafael2k | ow yeah, kernel packages ready | 20:41 |
rafael2k | the repo should be fine | 20:41 |
rafael2k | I'll finish up ofono repo soon, I will just wrap a missing glib function and it will be ready for maemo upstreaming | 20:42 |
tmlind | Wizzup: getting closer to a working system, now missing XDG_RUNTIME_DIR /run/user/1000 | 20:42 |
tmlind | rebooted, let's see, looks like desktop starts at least | 20:49 |
tmlind | ok updated, bbl | 20:52 |
rafael2k | btw, I realize the kernel is installed correctly, but the default initrd and Image.gz are not updated automagically - is this done somewhere else? | 20:55 |
freemangordon | "maemo upstreaming" yay! | 20:56 |
rafael2k | ; )) | 20:59 |
rafael2k | yaaay! | 22:15 |
rafael2k | I have my battery level back! | 22:15 |
rafael2k | ; )) | 22:15 |
rafael2k | for the first time since I don't remember when | 22:15 |
rafael2k | http://173.255.215.196/pinephone/maemo.jpg | 22:22 |
rafael2k | .dtb are not being copyed too... I need to check the debian/ ... I had to do some tweaks as default maemo pp debian dir did not work out-of-the-box | 22:23 |
rafael2k | Linux devuan-pinephone 5.15.10 #1 SMP Sat Dec 25 10:36:14 MSK 2021 aarch64 GNU/Linux | 22:24 |
rafael2k | now rebooting to see how X with accell is going on | 22:25 |
rafael2k | so this works: https://github.com/rafael2k/pine64-kernel/tree/maemo/beowulf-devel with this kernel https://github.com/rafael2k/sunxi64-linux/tree/mobian-5.15 | 22:27 |
rafael2k | just need to sort out setting the default kernel, initrd and dtbs | 22:28 |
rafael2k | https://gitlab.com/mobian1/devices/eg25-manager/-/merge_requests/46 | 22:57 |
rafael2k | My eg25-manager patch to make it compile without mm also went upstream | 22:58 |
rafael2k | X is better, not perfect, but way much better... thanks for the all the work - | 23:15 |
Wizzup | rafael2k: we-ll fix the other bug soon | 23:22 |
Wizzup | :) | 23:22 |
rafael2k | there is some work to do kernel side in pp, especially some firmware (like https://github.com/dragan-simic/ov5640-autofocus-firmware/ for camera), audio needs other tweaks (now there is hdmi also and new stuff), but all in all, things are getting better | 23:23 |
rafael2k | Wizzup: yay! | 23:23 |
rafael2k | alsa is working great through mplayer specifying the hw:0, but all ucm2 config is trashed in the new kernel, need to re-do the audio configs to match the new kernel audio feats | 23:57 |
rafael2k | not only ucm2, but standard alsa configs are not ok in the new kernel (it was not totally ok before... now they just do not work out-of-the-box) | 23:59 |
rafael2k | *with the new kernel | 23:59 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!