| mighty17[m] | freemangordon: https://github.com/xc-racer99/mesa-pvr/blob/mesa-20.3.2-pvr-musl-2/src/egl/generate/eglFunctionList.py#L200-L202 we need 2 functions from what i understand am i right? | 08:06 |
|---|---|---|
| mighty17[m] | also `LibGL: DRI_IMAGE (supported version: 8 - max version: 19)` ig DRI_IMAGE is working fine | 09:27 |
| mighty17[m] | `PVR:(Error): PVRContextCheckGLES2: Unsupported context flags: 0x4. [0, ]` also what is this? | 09:28 |
| mighty17[m] | `libEGL debug: EGL user error 0x3009 (EGL_BAD_MATCH) in eglCreateContext: dri2_create_context` | 10:24 |
| bencoh | hm, I'm trying to run torus-trooper with gl4es, but it dies with "abagames.util.sdl.sdlexception.SDLInitFailedException@src/abagames/util/sdl/sdlexception.d(13): Unable to create SDL screen: X11 driver not configured with OpenGL" | 12:01 |
| bencoh | ldd reports both libSDL-1.2 and libSDL-2 | 12:01 |
| bencoh | how is it linked against both? | 12:01 |
| bencoh | (apparently both libsdl1 and libsdl2 come from the leste repos) | 12:05 |
| Wizzup | uvos suggested we set SDL_RENDER_DRIVER=opengles2 | 12:06 |
| Wizzup | maybe this is related? | 12:06 |
| Wizzup | https://github.com/maemo-leste/leste-config/commit/a878a66b157eff7e8a9d8b7d1415f8559057844b | 12:06 |
| bencoh | I still get the sdl/x11/opengl error with SDL_RENDER_DRIVER=opengles2 | 12:07 |
| Wizzup | can you unset it? | 12:07 |
| bencoh | it's not set by default, only the hint is | 12:07 |
| Wizzup | weird, | 12:08 |
| bencoh | but I think it really tries to use GLX | 12:08 |
| Wizzup | what about export SDL_VIDEO_X11_FORCE_EGL=1 | 12:08 |
| Wizzup | FORCE_EGL doesn't sound like letting it use glx | 12:08 |
| bencoh | same | 12:09 |
| Wizzup | (I mean unset SDL_VIDEO_X11_FORCE_EGL) | 12:09 |
| bencoh | not sure for which sdl version it is | 12:09 |
| Wizzup | eright | 12:09 |
| bencoh | FORCE_EGL isn't set by default here | 12:09 |
| Wizzup | maybe it's a -devel only change | 12:09 |
| bencoh | oh, I don't have that file btw, probably -devel yeah | 12:09 |
| bencoh | I wonder if it actually uses sdl1 or sdl2 (hence my first question) | 12:10 |
| Wizzup | right, that is weird @ linked | 12:11 |
| uvos | forching egl should not stop it from using opengl | 12:17 |
| uvos | glx is broken on video-omap | 12:17 |
| uvos | (so if the application is coded to only use glx for context cration it cant work, even with gl4es) | 12:17 |
| bencoh | damn uvos left | 12:43 |
| bencoh | glxgears works with gl4es btw | 12:44 |
| uvos | bencoh: im here | 12:54 |
| uvos | bencoh: hmm thats pretty wierd it fails in conext creation here | 12:54 |
| mighty17[m] | <bencoh> "glxgears works with gl4es btw" <- wait what really? u didnt face https://github.com/ptitSeb/gl4es/issues/341 ? | 12:57 |
| bencoh | uvos: the fact is I wonder if it fails or if it doesn't try | 12:57 |
| Wizzup | mighty17[m]: did you look at our gl4es package...? | 12:57 |
| mighty17[m] | "our" being?? | 12:58 |
| uvos | bencoh: well sdl should use egl anyhow, thats unaffected | 12:58 |
| uvos | bencoh: but yeah best step with gdb to see what its really trying | 12:58 |
| * bencoh nods | 13:00 | |
| bencoh | mighty17[m]: http://pastebin.notk.org/pastebin.php?show=m124a0862 | 13:00 |
| bencoh | gl4es comes from https://maedevu.maemo.org/leste I didn't bother rebuilding | 13:01 |
| mighty17[m] | :o thats pretty interesting it working for you and not me | 13:02 |
| uvos | it also dosent work for me so idk what bencoh is doing | 13:02 |
| bencoh | uh | 13:05 |
| bencoh | I'm running -stable, maybe that's why | 13:06 |
| bencoh | I haven't fiddled that much with my leste install, so ... :) | 13:06 |
| buZz | i've played with puredata+GEM (its OpenGL library) on a droid4 with gl4es installed | 13:43 |
| buZz | zero issues and pretty decent performance | 13:43 |
| buZz | install gl4es, restart Xorg, get functional OpenGL | 13:45 |
| buZz | honestly, i didnt try glxgears cause i dont really consider that a worthwhile test :P | 13:46 |
| bencoh | alright, so ... that thing uses libsd1.2 apparently, and the error means that sdl was built without GLX support | 15:52 |
| uvos | bencoh: it was disabled for some time | 16:39 |
| uvos | bencoh: but its been reenabled in devel | 16:39 |
| bencoh | oh | 16:43 |
| bencoh | well, I rebuilt libsdl1.2 with it, it looks like it runs, I'll see if it displays something at home | 16:44 |
| bencoh | the issue with gl4es and allowing some kind of broken GLX compat layer to run is that I have a feeling it will break many detection mechanisms | 16:45 |
| bencoh | which is a bit sad | 16:46 |
| uvos | that is correct | 16:46 |
| uvos | most things will prefer gl when offered | 16:46 |
| bencoh | I think mpv 0.29 did that as well, I built the latest version from debian's git repository, and now it doesn't accept x11 anymore (I think it got deprecated / disabled by default) | 16:47 |
| bencoh | (ie it frowns when I pass gpu-context=x11, and works properly with x11egl) | 16:48 |
| freemangordon | ugh, what a nasty bug there is in rtcom_widget_set_account | 17:16 |
| freemangordon | they pass G_STRUCT_OFFSET(RtcomWidgetIface, store_settings) instead of RTCOM_WIDGET_GET_IFACE(obj)->store_settings as a function to g_signal_handlers_disconnect_matched() call | 17:18 |
| freemangordon | or rather it is G_STRUCT_OFFSET(RTCOM_TYPE_WIDGET, store_settings) | 17:38 |
| freemangordon | but still buggy | 17:38 |
| freemangordon | here https://github.com/maemo-leste/rtcom-accounts-ui/blob/master/widgets/rtcom-widget.c#L296 | 17:40 |
| bencoh | the fact that n900/fremantle proves to be usable is some kind of mystery/miracle to me :) | 17:41 |
| freemangordon | :D | 17:41 |
| bencoh | I mean, we've been using it for years, it really works ... but how :] | 17:41 |
| freemangordon | yeah, agree | 17:41 |
| freemangordon | I have a colleague of mine which has a theory about "quantum programming" :D | 17:42 |
| bencoh | ^.^ | 17:42 |
| freemangordon | the real question is whether I introduced more bug with REing that I fixed :p | 17:43 |
| freemangordon | *bugs | 17:44 |
| bencoh | at least debugging it should be more accessible to others | 17:44 |
| freemangordon | yeah | 17:44 |
| freemangordon | deffinitely | 17:44 |
| freemangordon | definitely | 17:44 |
| bencoh | which reminds me, I wonder if we still have that odd memory leak when double-pressing powerbutton to blank screen (I think that was the source of it) | 17:45 |
| freemangordon | first time I hear about that | 17:45 |
| freemangordon | it that in CSSU? | 17:45 |
| freemangordon | *is that | 17:45 |
| bencoh | not -devel, but yeah, years ago we tried to investigate a memory leak around blanking/unlocking screen | 17:46 |
| freemangordon | in tklock? I think I run systemui against valgrind back then and the was no memleak | 17:47 |
| freemangordon | I remember I even added code to run it as a standalone program, (iirc) | 17:47 |
| bencoh | yeah I remember you checked that, I think you even fixed something along the process, and I remember we didn't manage to find the source of it in the end | 17:48 |
| bencoh | but I could still see memory usage rise slightly at each screen blanking (when double-pressing power) | 17:48 |
| freemangordon | hmm | 17:49 |
| bencoh | (I use some gesture to blank, so it hasn't bothered me for years, and I stopped thinking about it) | 17:49 |
| bencoh | well, maybe I'll play a bit with that on droid4 and see if anything pops | 17:49 |
| freemangordon | well, we have all the sources already so you may want to check it in leste | 17:49 |
| freemangordon | mhm | 17:49 |
| freemangordon | actually you can run systemui in valgrind, leaving tklock only | 17:50 |
| bencoh | leaving tklock only? | 17:50 |
| freemangordon | that should give you full details on what's goiing on | 17:50 |
| freemangordon | as a module | 17:50 |
| bencoh | I think the systemui menu is the cause btw | 17:50 |
| freemangordon | what is "systeui menu"? powerkey-menu? | 17:50 |
| bencoh | yeah | 17:51 |
| bencoh | iirc they have some hack, where it catches the first press and waits for a second one | 17:51 |
| freemangordon | then leave powerkey-menu only | 17:51 |
| freemangordon | yeah, I still recall like a little elephant :) https://github.com/maemo-leste/osso-systemui-powerkeymenu/blob/master/osso-systemui-powerkeymenu.c#L48 | 17:52 |
| freemangordon | so, you can uncomment that and run powerkeymenu as a standalone application | 17:52 |
| freemangordon | or, remove all other systemui plugins and leave that one only and run systemui with valgrind | 17:53 |
| bencoh | hmmm, I'll have to try then, thanks :) | 17:54 |
| freemangordon | though I am almost sure I fixed everything in powerkeymenu back then. but... who knows :) | 17:54 |
| bencoh | https://github.com/maemo-leste/osso-systemui-powerkeymenu/commit/98dd7a486ac886c536f812e21cc21b9b20a30706 this somehow rings a bell | 17:56 |
| bencoh | but I think I tried a year or two later, so ... | 17:57 |
| bencoh | anyway, we'll see :) | 17:57 |
| freemangordon | mhm | 17:57 |
| freemangordon | valgrind to the rescue | 17:57 |
| bencoh | yeah (not that I really like valgrind, especially when glib is involved ...) | 17:57 |
| bencoh | do we have/need a valgrind exceptions file? | 17:57 |
| freemangordon | ugh, why? just pass correct exceptions | 17:57 |
| freemangordon | yes, you need | 17:58 |
| freemangordon | I can provide the ones I use(d) for osso-abook | 17:58 |
| freemangordon | they are not 100% complete, but still better than nothing | 17:58 |
| bencoh | yeah, the issue is guessing the correct ones, it's kindof a pain, and I haven't done that in a while | 17:59 |
| freemangordon | --suppressions :) | 17:59 |
| bencoh | huhu :) | 17:59 |
| freemangordon | ok, PM me some email and I'll send them to you | 17:59 |
| freemangordon | sent | 18:01 |
| bencoh | received, thanks! | 18:01 |
| bencoh | for the records, torus-trooper works well | 20:38 |
| bencoh | (with gl4es and sdl1.2 allowing glx) | 20:39 |
| bencoh | and vkcube works (alright, that's just for fun) | 20:49 |
| mighty17[m] | vulkan :O thats awesome | 20:51 |
| mighty17[m] | series6 got it but not series5 :'( | 20:52 |
| uvos | /etc/X11/Xsession.d/03alarmd is useless | 20:54 |
| uvos | there is no way this works | 20:55 |
| bencoh | vulkan actually uses llvm, I've been lied to :( | 22:29 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!