libera/#maemo-leste/ Friday, 2022-02-18

tmlindsicelo: only one wlroots patch needed after the dbm patch, the one from jonathan bakker05:08
freemangordonWizzup: https://github.com/maemo-leste/osso-abook/commit/72ed1ff38e66f6542cac99da2d6258bec01377bf :)08:16
sicelotmlind: nice :-)09:05
Wizzupfreemangordon: \o/10:07
WizzupPali: planning to look at u-boot now, any other patches I should carry for the DM usb stuff?10:51
PaliWizzup: Ok! IIRC all relevant patches are now merged in master branch.10:53
Wizzupok10:54
freemangordonWizzup: could you push ddx-um from devel to stable? I want to promote the REed dbm in devel11:02
Wizzupfreemangordon: what is in devel that is not in stable atm?11:04
freemangordonES1 fix11:04
Wizzupso that is 76e87709c19acfe0bf3d1bc7b1b1a23252aa29c2 - you want that in stable?11:04
freemangordonyes11:05
freemangordonit allows REed dbm to be build, but it is not actually installed - blob libdbm is used still11:06
Wizzupah, ok!11:06
Wizzupand this is tested in devel alreadu?11:06
freemangordonit is there since I fixed it11:06
Wizzupok11:06
Wizzupok, will build for stable now11:06
freemangordonwhether it is tested - dunno, but the only actual change is the ES1 fix from uvos11:07
Wizzupbuilding11:07
freemangordonmy patches don;t actually affect the distributed binaries11:07
freemangordonok11:07
freemangordontmlind: please test if  https://github.com/maemo-leste/sgx-ddk-um/tree/render_node still works for you11:22
freemangordonalso, please have I look if https://github.com/maemo-leste/sgx-ddk-um/commit/0ef13219f473671b66028102697727114ca8b591 looks ok11:22
tmlindfreemangordon: looks good and works for me thanks12:04
tmlindfreemangordon: so i guess pretty much the only remaining weirdness is why stellarium renders at 7fps vs 11fps ish compared to the original blobs12:06
tmlindfreemangordon: oh and the mesa EGL_EXT_image_dma_buf_import_modifiers issue12:08
tmlindhmm maybe the objectcache stuff causes the performance difference?12:15
freemangordontmlind: WDYM by "objectcache"?12:21
freemangordonpvr_dri is 100% the same as in the blobs12:21
tmlindweird, i wonder what the difference could be? the blobs also was crashing with glmark2 so something also fixed that somehow12:26
freemangordontmlind: umm, no, blobs were not crashing, it was chromeos dri that was crashing12:27
tmlindoh ok12:27
freemangordonbut blobs does not support xorg, so I wonder there do you get 11fps from, chromeos dri?12:27
tmlindyeah i'm pretty sure swapping some earlier dri makes stellarium much faster and glmark2 to crash12:28
tmlindyeah probably it was the chromeos dri then, not the old binary dri12:28
freemangordonyeah12:29
tmlindif it was the chromeos dri, then 87790e2b9ce7926f6cc492fce01ca9d2f5589107 adds the object cache stuff which could explain12:29
freemangordonis it possible to run stellarium on wayland?12:29
tmlindyeah that's what i use it for12:29
freemangordonwith blobs that is12:29
freemangordonand see what fps we get there12:29
tmlindyeah that's what i was testing with over past few years12:29
freemangordonit could be some mesa change that lowered the fps12:30
tmlindsomething lowered the fps and fixed the glmark2 hang12:30
tmlindmay have been two separate changes though12:30
freemangordondoes 87790e2b9ce7926f6cc492fce01ca9d2f5589107 fixes fps?12:31
tmlindno, that still can't be done with wlroots, i've had that always disabled for wlroots12:32
tmlindso i'm pretty sure bakker's original mesa (with EXT_image_dma_buf_import_modifiers disabled) is faster but hangs with glmark212:34
freemangordonyeah12:34
tmlindso the object cache seems like the biggest difference there to me12:34
freemangordonmhm12:34
freemangordonbut, can you check the fps with blobs12:35
tmlindwhich blobs?12:35
tmlindthe original ti blobs?12:35
tmlindi'll try disbling the object cache somehow12:36
freemangordonyes, original blobs12:39
freemangordonif they render with 7fps, then obviously that's all we can get12:40
tmlindhmm not sure i even remember what all hacks wlroots needs to use the original blobs..12:42
tmlindlet's see if this works: git diff 87790e2b9ce7926f6cc492fce01ca9d2f5589107.. src/mesa/drivers/dri/pvr/pvrdrawable.c | patch -p1 -R12:42
tmlindonly steps to go12:43
tmlindonly 100 steps12:43
sicelofor original blobs, iirc we only needed 0001-render-Don-t-use-GL_EXT_unpack_subimage-when-not-ava.patch12:47
sicelofor wlroots, that is12:47
tmlindyeah maybe for wlroots-0.1412:48
tmlindi'll test with mesa commit at 8206053e95d326d4833f70c65d26527b3a8f160712:54
freemangordonthis is chromeos I would say12:55
tmlindyeah, that commit does stellarium at close to 13fps12:58
tmlind12.8 to 12.9 fps looks like12:59
tmlindlet's see if glmark2 hangs now..13:00
tmlindyup, glmark2 causes some pvr dump13:01
tmlindPVR_K:(Error): SGXOSTimer() detected SGX lockup (0x14 tasks)13:01
tmlindand so on13:01
tmlindok need to continue at some point later on, ttyl13:02
freemangordonWizzup: did you forget about libsdl?13:53
uvosWizzup: did you try kicking neverball again?13:53
Wizzupfreemangordon: did not forget14:13
Wizzupuvos: see other channel14:14
Wizzupfreemangordon: I've had to deal with some stuff, I'll do it today14:14
freemangordonok14:21
freemangordonuvos: is it possible to turn display on *after* vtklock confirms it is shown?15:47
uvosfreemangordon: if thats not how it works allready then you can by changing the module load order, but i cant recommend doing that willy nilly, it causes a lot of subtle bugs15:56
freemangordonok15:57
freemangordonI am thinking of how to deal with vtklock still being drawn when display turns on15:57
uvosit signals mce its finished to early maybe15:58
freemangordonbut maybe this is some gtk issue, or vtklock issue itself15:58
freemangordonhmm, ueah15:58
uvosand this some real issue15:58
freemangordon*yeah15:58
freemangordonwill check it15:58
uvosso tklock.c waits untill it gets the finished signal via dbus15:59
uvosif thats before or after the when the display gets turned on depends on module load oder15:59
uvosbut im pretty sure its before15:59
Wizzupcould it be a window resize problem?16:00
Wizzupinitially it has some default size and then resizes16:00
uvosright16:00
uvosyeah thats very possible16:00
uvosWizzup: btw dose openlara show for you now?16:01
uvosin ham16:01
Wizzupyes16:02
Wizzupwhen I filter by typing 'lara' it did not show16:02
WizzupI am not sure why, but I can see it in the games category16:02
uvoswierd16:02
uvosok but thats unlikely to be the packages fault16:02
Wizzupdepends on what it filters on16:02
Wizzupif I search for 'open' I can find it16:03
Wizzupbut not for 'lara'16:03
uvoshmm16:03
uvosi would expect it to filter on the x-maemo-name thing16:03
uvosand/or the pacakge name16:03
uvosdose it find Lara?16:03
Wizzupchecking16:08
Wizzupno16:09
Wizzupmaybe it just matches the start of every word16:09
Wizzuplooks like that's it16:09
uvoshm that could be improved16:10
Wizzupfirst I think want a separate category for debug symbols16:25
uvosjust hide everything that maches dbgsym16:36
uvosits not like users that want to use this gui for pacakge management but cant use apt are going to have a need for debug symbols16:36
Wizzupagreed16:36
freemangordonyeah, sounds right (hide everything dbgsym)17:43
Wizzupprobably by category but yeah17:54
lelMerlijnWajer opened an issue: https://github.com/maemo-leste/bugtracker/issues/609 (hildon application manager: hide debug symbol packages)17:55
freemangordonyeah, maybe18:03
WizzupI mean the dbgsym have a specific section: in debian18:05
freemangordonmhm, got it18:10
freemangordonugh, I have to fix DDX fd limits issue19:47
freemangordonyeah, a task for the weekend19:47
freemangordonuvos: hmm, openlara fails to start19:48
freemangordondo I miss something?19:48
freemangordongives BadDrawable on startup19:48
uvosfreemangordon: its a engine reimplementation19:53
uvosfreemangordon: for a comertial game, you need to copy the files of the original disk into ~/.openlara19:53
uvosfreemangordon: yes i need to document this somehere19:53
uvosi reported the BadDrawable upstream19:53
freemangordonah, ok19:53
uvosbut its not a big deal it just happens on exit19:54
freemangordonah, so it exits because data files are missing19:54
freemangordonI see19:54
uvosright19:54
freemangordonyeah, lemme see if I can find the data files19:54
uvosi can give them to you19:54
uvosits just 64mb or so19:54
freemangordonplease do19:54
freemangordonTBH I was never a fan of thomb raider but I am curious to see how this is going to run19:55
uvos45 fps ish, works pretty well gameplay wise because its a game played only with kbd19:56
freemangordonok, still it is good to have it running19:56
freemangordonI may use it as a benchmark as well19:57
freemangordonuvos: is this https://github.com/andrewsyc/Tomb-Raider-1-2-3-4-5-Map-viewer-and-levels what I need?20:04
uvosfreemangordon: yes but you need the file/folder structure of the orignal game, also the audio files look to missing there20:08
uvosalso lol @ hosting copyrighted files on gh20:08
freemangordonyeah :)20:08
uvossent you an email20:08
freemangordonbut I doubt anybody cares for a 20yo(or even more) game20:09
freemangordonthanks!20:09
uvosprobubly not yeah20:09
uvos(you need to extract the archive into ~/.openlara)20:10
freemangordonmhm20:10
freemangordonuvos: downloaded20:20
freemangordonuvos: if you fix neverball I can respin th ebuild20:28
freemangordonor, do you want me to fix it?20:28
freemangordonit hangs SGX :(20:49
freemangordon(neverball(20:49
rafael2k< typing from maemo leste in the pinephone, with keyboard and bt mouse20:49
rafael2k: )20:49
freemangordon:)20:49
rafael2kI just paired the mouse with bluetoothctl and mouse worked20:50
freemangordonno reason not to20:50
freemangordonh-d has support for mice (it should show a cursor)20:51
rafael2kand a very nice black cursor showed up20:51
freemangordonmhm20:51
freemangordonthats how we test in amd64 VM ;)20:51
rafael2kcool20:51
rafael2kit works pretty well indeed20:52
uvos hmm localy built neverball works perfectly fine here21:43
freemangordonweird21:45
freemangordonit hangs SGX in 10 seconds or so21:45
freemangordonas sitting in the main menu21:45
uvoswierd21:46
freemangordonmhm21:46
uvosi have played several levels21:46
freemangordonlemme rebuild it agains old sdl21:46
uvosyour device is fine with loading sgx otherwise right?21:46
freemangordonyep21:46
freemangordonno issues whatsoever21:46
uvosie it dosent crash if you have it render glmark or openlara or so21:46
uvosok21:47
freemangordonit could be that I built it agains lates libsdl2, trying the one in the repos now, lets see21:49
freemangordonstill hangs :(21:52
freemangordonuvos: do you have something in the settings?21:53
freemangordonlike fullscreen, or... dunno21:54
uvosdident change anything21:54
uvosfullscreen dosent work21:54
uvosit just gets nocked back to windowed21:54
freemangordonmhm21:54
freemangordonI even run it with h-d stopped21:54
freemangordonit plays video for several seconds and hangs21:55
uvosi have had it in menu for >1min now21:56
uvosnothing bad happens21:56
uvoswell asside from https://github.com/Neverball/neverball/issues/28021:56
freemangordonok21:56
freemangordonhmm, it does not look like that here21:57
uvosit allso has the bad drawable bug21:58
uvoson exit21:58
uvosinteresting21:58
uvosmust be an issue with sdl221:58
freemangordonstill, I wonder why I get SGX HW recovery triggered21:59
freemangordonhmm: GL version: OpenGL ES-CM 1.122:00
freemangordoncould that be the issue?22:00
uvosGL renderer: PowerVR SGX 54022:00
uvosGL version: OpenGL ES-CM 1.122:00
freemangordonyeah, same here22:00
freemangordon:(22:01
freemangordondid you upgrade sgx-ddk-um-ti443x?22:01
freemangordonto 1.17.4948957+leste14-1+2m722:02
uvossha1sum /usr/lib/arm-linux-gnueabihf/libGLESv1_PVR_MESA.so.1.17.494895722:02
uvos742b822fc7a6e5b986abf43b1e743e1d9e54de3122:02
freemangordonsame here22:02
uvosi dident upgrade22:02
uvosbut i replaced by hand22:02
freemangordonsha1sum /usr/lib/arm-linux-gnueabihf/libdbm.so.1.17.494895722:02
freemangordon48a195956f82f402e9ec5bb746fbf8323e58aa8f  /usr/lib/arm-linux-gnueabihf/libdbm.so.1.17.494895722:02
uvosc53d1a7683f5f5063fdb5aee47cef0f3403045aa22:03
uvoshmm22:03
freemangordonok, you are on the old version22:03
freemangordonlemme revert to see if it will make any difference (I doubt)22:03
uvosyeah as mentioned i just replaced GLESv122:03
freemangordonmhm22:03
freemangordonuvos: it does not crash with SDL_RENDER_DRIVER=opengles22:18
freemangordonfor some reason I have SDL_RENDER_DRIVER=opengles2 defined22:19
freemangordonany clue where this comes from?22:20
uvosi have that too22:22
uvosits from leste-config22:22
uvosto make sdl prefer gles22:22
uvosdosent harm my neverball22:22
freemangordonbut, afaik you use locally compiled mesa, no?22:22
uvosuser@maserati:~/programing/neverball$ printenv | grep SDL22:22
uvosSDL_VIDEO_X11_FORCE_EGL=122:22
uvosSDL_RENDER_DRIVER=opengles222:22
uvosSDL_VIDEO_X11_XRANDR=122:22
uvosfreemangordon: no22:22
uvosfreemangordon: not anymore22:22
uvosi also think i use repo sdl22:23
uvoslet me upgrade22:23
freemangordonhmm, I don;t have SDL_VIDEO_X11_XRANDR=122:23
uvosi might have added that22:23
uvosmay try it22:23
freemangordonbut I doubt this is the issue22:23
uvosunlikely yeah22:23
uvosSource: libsdl222:25
uvosVersion: 2.0.14+dfsg2-322:25
freemangordonwtf? now it hung with SDL_RENDER_DRIVER=opengles22:25
uvosupgrade wants to update sdl22:25
freemangordonno idea why22:25
freemangordonok, if you go to options very fast, it does not hang22:27
uvoswell it stops rendering 3d then22:27
uvosdosent hang with newer sdl22:27
uvoslet me rebuild the thing22:27
freemangordondo you have anything in /etc/powervr.ini?22:28
uvosno dont have it22:28
freemangordonok22:28
freemangordoncrashed again]22:28
freemangordonnot in the menu, but when I choose to play22:29
freemangordon3d, again22:29
uvosdo you have https://github.com/Neverball/neverball/issues/280?22:29
freemangordonno22:29
uvoshmm22:29
freemangordonfonts are not fuzzy22:29
uvosthe version i installed was at https://github.com/Neverball/neverball/commit/5508e07ac566ae71a5cb54735a8ffa64daff349322:30
uvos(as requested to test by upstream)22:30
uvosthats probubly it22:30
uvosthe version in the extras repo is whatever debian packages22:30
freemangordoncould you provide your binary, to test here?22:30
uvossure22:31
uvoshttp://uvos.xyz/maserati22:32
freemangordonnope, still crashes22:34
uvosblurry fonts?22:34
freemangordonno22:34
uvoswtf22:34
uvosit has blurry fonts also on llvmpipe here and also on desktop+amdgpu22:34
freemangordonno idea22:34
freemangordonoh, could it be -common or -data package?22:35
uvosmaybe22:35
freemangordonwhere do you get those from?22:35
freemangordondebian repo?22:35
uvosnot sure, i might have installed something custom built before i dropped building those in neverball-gles22:36
uvossec22:36
freemangordoncommon/data are 1.6.0+git20180603-222:37
freemangordongles is 1.6.0+git20180603-422:37
freemangordonso that might be the issue22:37
uvosRemoving neverball-data (1.6.0+git20180603-2)22:38
uvosi have the same thing22:38
uvosor had22:38
freemangordonno idea then22:38
uvosturns out neverball is very toublesome :P22:38
freemangordon:D22:38
uvosand in the end its almost unplayable without a mouse22:38
uvos:P22:38
freemangordonwhy mouse, I remeber seeing that on n900 using the accel22:39
uvosyou can make it use accel22:39
freemangordonah22:39
uvosi even built a utility22:39
uvosto make that happen22:39
uvosbut it still sucks if you cant control the mouse22:39
uvos /camera22:39
freemangordonI seee22:40
uvossee https://github.com/IMbackK/iio-uinput22:40
freemangordonok, I really wonder why you see blurred fonts and I don;t22:40
freemangordoncool22:41
uvosi reinstalled data22:41
uvosand it still works fine22:41
freemangordonstill blurry?22:41
uvosyeah22:41
uvosill rebuild a new copy later22:41
uvosmaybe its becuse i bult the current one against _something_ old22:41
uvosbut even then idk why the binary dosent work22:41
freemangordonmay I have your neverballrc?22:42
freemangordonugh aniso                     822:43
uvossure its on my host22:43
uvosmultisample               022:44
freemangordonI would say those are not default setting22:44
uvosi gues, not that i remember changing anything22:44
freemangordonhttps://pastebin.com/wdxy81M022:44
uvosi also deleted the whole directory for  https://github.com/Neverball/neverball/issues/28022:44
uvosdident change anything then22:45
uvosbut i also have .neverball-dev and .neverball22:45
uvoshmm22:45
freemangordonI have only .neverball-dev22:45
freemangordonyeah, it works now22:46
uvosso dose it hang wiht my rc22:46
uvosok22:46
uvosmutisample causes it to hang then i gues?22:46
freemangordonI would guess that 8x anosotropic filtering could be it22:46
freemangordonlemme check which one it is22:47
uvosmaybe i deleted .neverball22:47
uvosisntead of -dev22:47
freemangordonhmm, but now fonts are way smaller22:47
freemangordonlemme revert to my config22:48
uvosha22:49
uvosyeah22:49
uvosif i delete .neverball-dev it hangs22:49
uvosaha22:49
uvosso i deleted .neverball thinking it was using that22:49
freemangordonmultisample is set to 0, so it is not that one22:50
uvosit also has the wrong aspect ratio22:50
uvosafter deletion22:50
uvoshm22:52
uvosi rebooted my device and sgx has not recoverd22:52
uvosit renders hildon just black22:52
freemangordonnever seen such issue here22:53
uvosyeah its not comeing back, ill remove the battery for a while22:53
freemangordonok, got it working with good fonts23:00
freemangordonuvos: the option that crashes it seems to be 'shadow'23:06
freemangordonuvos: yep, that's it, the shadow23:14
uvosand what causes blurry fonts?23:14
uvosthat would be good to know for 28023:14
freemangordonwell, you can do diff betwwen your 'blurry' setting and default and start playing with differences23:15
freemangordonwhat is "textures 4" ?23:16
freemangordonuvos: yep, that's it23:18
freemangordonsetting textures to 4 makes fonts blurry23:18
uvosright23:19
uvoslet me update upstream on this23:20
freemangordonI don;t see this documented23:21
uvossee what documented?23:22
freemangordonwhat "textures" setting means23:22
freemangordonhttps://github.com/Neverball/neverball/blob/master/doc/manual.txt23:22
uvosit reduces the texure size by 423:22
uvosif set to for23:22
uvosyou can see it on the foor etc23:22
uvosbut in gles1 mode it also changes the size of the fonts23:23
uvosin ogl mode it dosent23:23
freemangordonok, but is not documented :)23:23
uvossure23:23
uvosanyhow i informed upstream about that bug23:23
uvosnot sure what to do about the shadow bug, thats in sgx obv.23:23
freemangordonand what about "shadow"?23:23
freemangordonnot sure23:23
uvoswell sgx should never crash no matter what the application dose23:23
freemangordonyeah23:24
uvosso even if the shadow implementaion is terrible it should not crash the device23:24
freemangordonagree23:24
freemangordonbut not much we can do23:24
uvosyeah23:24
uvosdisable it by default it gues23:24
uvosis all we cant do23:24
uvos*can23:24
freemangordonyeah23:24
uvosin unrelated news: the shadow implementation in openlara is also broken on sgx23:27
freemangordonI wonder if we can somehow get some support from TI23:27
uvosdosent crash the device.. but23:27
uvoswell try it23:27
freemangordonuvos: maybe it is SDL that does not plays nice23:27
uvosi doubt23:28
uvosin openlara its a shader23:28
uvosalso its gles223:28
freemangordonyeah23:30
freemangordonI wonder if we can somehow workaround that with powervr.ini settings23:31
freemangordonanyway, its late here, going to zzz23:31
freemangordonnight!23:31
uvosnight23:31

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