#jogamp @ irc.freenode.net - 20150609 05:05:41 (UTC)
20150609 05:05:41 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20150608050541.html
20150609 05:05:41 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20150609050541.html
20150609 05:52:32 * jk4 (~jk4@anon) Quit (Ping timeout: 258 seconds)
20150609 06:33:46 * bigpet (uid25664@anon) has joined #jogamp
20150609 06:58:17 * monsieur_max (~maxime@anon) has joined #jogamp
20150609 07:05:51 * jk4 (~jk4@anon) has joined #jogamp
20150609 08:39:52 * jvanek (jvanek@anon) has joined #jogamp
20150609 11:06:45 * eclesia (~husky@anon) has joined #jogamp
20150609 11:06:49 <eclesia> hi
20150609 15:53:24 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20150609 16:04:46 * eclesia (~husky@anon) has left #jogamp
20150609 16:28:14 * jvanek (jvanek@anon) Quit (Quit: Leaving)
20150609 16:59:45 * monsieur_max (~maxime@anon) has joined #jogamp
20150609 20:34:16 * noxuma (~markt@anon) has joined #jogamp
20150609 20:35:31 <noxuma> Hey, I was wondering if you guys think this bug could possibyl be jogl related
20150609 20:35:37 <noxuma> or if its linux intel driver
20150609 20:35:39 <noxuma> http://imgbin.org/index.php?page=image&id=24165
20150609 20:37:07 <sgothel> this is a quite vague way of presenting a question :)
20150609 20:38:04 <sgothel> you surely can test rendering results yourself, i.e. whether a driver produces an artifact - simply compare a Mesa3D result (software .. intel .. amd ..) w/ a proprietary driver .. etc
20150609 20:39:00 <sgothel> 'a bug in jogl' .. regarding rendering artifacts? sure its possible, i.e. in case we map the wrong data or even GL function pointer through JNI ..
20150609 20:39:29 <noxuma> Well this bug only happens.. about one frame per 20 seconds or less
20150609 20:39:42 <sgothel> however, this is quite improbable, since our binding is auto-generated and under public review and test
20150609 20:39:59 <sgothel> so my best guess is .. compare different GL implementations
20150609 20:40:01 <bigpet> noxuma: try to record the bug with apitrace or something
20150609 20:40:11 <sgothel> yup
20150609 20:40:12 <noxuma> its definitely in the frame buffer
20150609 20:40:24 <noxuma> what is apitrace?
20150609 20:40:30 <sgothel> i.e. if the recorded apitrace differs from source code -> bug in jogl
20150609 20:40:53 <sgothel> otherwise .. you may send a bugreport to the implementor, but probably you have a bug in your code :)
20150609 20:40:54 <bigpet> a program that captures opengl (or directx) calls and can be used to debug opengl issues
20150609 20:41:19 <sgothel> yeah .. a nice free software GL debugger/trace-app, Mark recommends it
20150609 20:41:39 <noxuma> i doubt its our code here.. it doesn't happen on windows and it didn't happen in our oldder drivers
20150609 20:41:57 <sgothel> you may also try '-Djogl.debug.DebugGL', to enable GL debug profile and GL error checking
20150609 20:42:02 <noxuma> I tried those
20150609 20:42:05 <noxuma> no errors
20150609 20:42:06 <sgothel> or '-Djogl.debug.TraceGL' ..
20150609 20:42:26 <sgothel> don't you have a few machines w/ diff GL implementations ?
20150609 20:42:27 <noxuma> the lack of reproducability of this glitch is the most disturbing
20150609 20:42:37 <noxuma> we have 2 machines
20150609 20:42:41 <noxuma> Intel 4000
20150609 20:42:45 <noxuma> and intel 3000
20150609 20:42:45 <sgothel> so if it only happens w/ one GL impl. .. well ..
20150609 20:42:48 <noxuma> the 3000 seems ok
20150609 20:42:59 <sgothel> free software driver ?
20150609 20:43:00 <noxuma> but the 4000 also runs at 60fps
20150609 20:43:07 <noxuma> linux intel drivers
20150609 20:43:10 <bigpet> well that was among one of the reasons why I suggested apitrace. Capture a trace and see if replaying that trace also causes the issue
20150609 20:43:11 <sgothel> so: yes
20150609 20:43:36 <sgothel> yup .. thats a nice way of reproduction ofc
20150609 20:43:44 <noxuma> interesting i will definitely try this
20150609 20:43:45 <sgothel> i.e. frame by frame
20150609 20:43:51 <bigpet> is there even non-free linux drivers for intel cards anymore?
20150609 20:44:05 <noxuma> i'm not aware of any
20150609 20:44:06 <sgothel> sure there are, on the non-free OS :)
20150609 20:44:27 <bigpet> the non-free version of linux?
20150609 20:44:29 <noxuma> the intel-gfx channel has 200 people in it and they don't even answer a single question on the subject
20150609 20:44:38 <noxuma> haha ya, we're on fedora 20
20150609 20:44:43 <noxuma> or 21 i forget
20150609 20:44:51 <sgothel> I mean Windows, OSX, .. :)
20150609 20:45:06 <noxuma> can't use windows in regulated industry =)
20150609 20:45:10 <noxuma> since it can't be compiled
20150609 20:45:28 <sgothel> just answering questions, not dare to recommend it :)
20150609 20:45:43 <noxuma> =)
20150609 20:46:15 <sgothel> well, but I dare to recommend to setup a few boxes w/ diff. drivers (for god sake) .. to check for diff GL behaviors - just in case the app is important :)
20150609 20:46:48 <noxuma> I run it on windows
20150609 20:46:53 <noxuma> debug versions
20150609 20:46:57 <noxuma> and it runs perfectly there
20150609 20:47:11 <noxuma> i can fix this glitch..
20150609 20:47:18 <noxuma> if i glflush after every gldrawelements call
20150609 20:47:27 <noxuma> but of course that guts my framerate
20150609 20:47:44 <sgothel> sounds quite like a driver issue - OR, you dare to modify the data .. :)
20150609 20:47:56 <noxuma> data is solid. just textures
20150609 20:47:57 <sgothel> (before swap the framebuffers)
20150609 20:48:02 <noxuma> we do read the data
20150609 20:48:11 <noxuma> but we don't modify it ( to record pixels to disk)
20150609 20:48:20 <noxuma> thus how i got that screenshot
20150609 20:49:26 <sgothel> I remember there are read-pixels semantics w/ undef. behavior regarding the read-buffer ..
20150609 20:49:58 <noxuma> the bug does happen when we aren't recording
20150609 20:50:27 <sgothel> and when you do - no issue ? -> memory sync issue, likely, hence your glFinish ..
20150609 20:51:08 <noxuma> when i do glFinish there are 0 glitches, otherwise yes it happens randomly
20150609 20:51:16 <sgothel> mesa3d has a bugzilla and mailinglist .. try to reproduce (smalles native-C test case) .. and file a bug report
20150609 20:51:16 <noxuma> every few seconds or every few minutes
20150609 20:51:32 <noxuma> so mesa, rather than intel?
20150609 20:51:45 <sgothel> the intel devs are there as well -> yes
20150609 20:52:00 <sgothel> some freedesktop site for mesa3d .. hmm
20150609 20:54:31 <sgothel> https://jogamp.org/bugzilla/show_bug.cgi?id=873 <- one example of how we 'tried' to interact w/ Intel/Mesa devs .. ahem :)
20150609 20:56:28 <noxuma> fun
20150609 20:57:20 <noxuma> this bug is driving me insane.. staring at running games.. looking for a super fast glitch..
20150609 20:57:36 <noxuma> and even when u see it.. its random garbage pixels
20150609 20:57:42 <noxuma> for no logical reason
20150609 20:57:55 <sgothel> a typical memory sync issue
20150609 20:58:09 <noxuma> ya, i just dunno where to look
20150609 20:58:20 <noxuma> they dont seem interested in helping
20150609 20:58:50 <sgothel> we have our GLRendererQuirks for that reason .. one is e.g. issue glFlush on OSX/NV .. for some circumstances .. etc
20150609 20:59:04 <noxuma> ya i've been looking there as well
20150609 20:59:08 <sgothel> well, if you can proof the issue - maybe you get some attention
20150609 20:59:24 <noxuma> ya, we got multiple people looking into it here
20150609 20:59:25 <noxuma> =)
20150609 20:59:30 <sgothel> otherwise .. you may want to have that workaround ..
20150609 20:59:45 <sgothel> based on renderer .. etc ..
20150609 20:59:47 <noxuma> ya, higher ups will love that our new os runs slower =)
20150609 21:00:03 <noxuma> we only ever have one hardware renderer. we make our own hardware here
20150609 21:00:08 <noxuma> embedded
20150609 21:00:33 <sgothel> sounds interesting
20150609 21:02:47 <sgothel> http://jogamp.org/wiki/index.php/Maintainer_and_Contacts#Sven_Gothel :)
20150609 21:03:10 <noxuma> haha noted =p
20150609 21:03:53 <sgothel> (for the big-bucks companies only, of course :)
20150609 21:04:32 <sgothel> so what is your security critical device up for (if you can share)?
20150609 21:05:39 <noxuma> its a casino slot machine
20150609 21:05:46 <bigpet> embedded, games and needs source-audit? casino
20150609 21:05:48 <bigpet> was about to gues
20150609 21:05:52 <noxuma> haha correct
20150609 21:05:55 <sgothel> hohoho :)
20150609 21:06:31 <noxuma> whats the current jogl version
20150609 21:06:38 <sgothel> so 'only' financial destruction :)
20150609 21:06:53 <noxuma> well, gamblers prolly dont care if it glitches a bit
20150609 21:06:57 <noxuma> but we do =)
20150609 21:07:09 <sgothel> http://jogamp.org/wiki/index.php/SW_Tracking_Report_Objectives_for_the_release_2.3.1
20150609 21:08:17 <sgothel> jogamp.org <- I know .. some 'folks' like to have a blinking green DOWNLOAD button .. but the 2.3.1 is 'hidden' in the 'Builds/Download' section :)
20150609 21:09:45 <sgothel> off thread: 'Kyocera Ecosys M6526cdn' <- OK? Company needs a printer/scanner/... so complicated to review those specs .. and .. must run w/ GNU/Linux ofc :)
20150609 21:10:39 <noxuma> we'll try latest jogl, and see if that helps this bug at all. thanks for the input guys
20150609 21:21:19 * noxuma (~markt@anon) has left #jogamp
20150609 21:40:07 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20150610 05:05:41 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20150610050541.html