#jogamp @ irc.freenode.net - 20131028 05:05:50 (UTC)
20131028 05:05:50 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20131027050550.html
20131028 05:05:50 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20131028050550.html
20131028 05:15:10 * hharrison (~chatzilla@anon) has joined #jogamp
20131028 05:15:42 <hharrison> sgothel: I wouldn't count additional API as an apit-break, as long as no existing api works differently
20131028 06:02:16 * xranby (~xranby@anon) Quit (Ping timeout: 245 seconds)
20131028 06:06:29 * xranby (~xranby@anon) has joined #jogamp
20131028 06:34:46 * hharrison (~chatzilla@anon) Quit (Remote host closed the connection)
20131028 07:46:37 * monsieur_max (~maxime@anon) has joined #jogamp
20131028 08:00:56 * eclesia (jsorel@anon) has joined #jogamp
20131028 08:24:28 * [Mike] (~Mike]@anon) Quit ()
20131028 10:55:23 * eclesia (jsorel@anon) has left #jogamp
20131028 11:23:02 * xranby (~xranby@anon) has left #jogamp
20131028 11:47:25 <rmk0> sgothel: me neither. would mark the extra functions with @since
20131028 12:04:52 * eclesia (jsorel@anon) has joined #jogamp
20131028 12:46:44 * eclesia (jsorel@anon) Quit (Quit: Leaving.)
20131028 12:48:37 * eclesia (jsorel@anon) has joined #jogamp
20131028 12:50:09 * eclesia (jsorel@anon) Quit (Client Quit)
20131028 13:10:40 * xranby (~xranby@anon) has joined #jogamp
20131028 13:22:39 * xranby (~xranby@anon) Quit (Ping timeout: 272 seconds)
20131028 13:25:28 * xranby (~xranby@anon) has joined #jogamp
20131028 13:47:21 * xranby (~xranby@anon) Quit (Ping timeout: 272 seconds)
20131028 14:00:19 * xranby (~xranby@anon) has joined #jogamp
20131028 14:08:15 * xranby (~xranby@anon) Quit (Ping timeout: 272 seconds)
20131028 14:11:36 * xranby (~xranby@anon) has joined #jogamp
20131028 14:43:30 * [Mike] (~Mike]@anon) has joined #jogamp
20131028 15:56:14 <sgothel> thx - yes, '@since' is a good idea ..
20131028 16:12:01 * hharrison (~chatzilla@anon) has joined #jogamp
20131028 16:12:27 <hharrison> sgothel: starting to digest those patches for bug 776....lots of stuff going on there
20131028 16:16:41 <hharrison> also, I have prelim patches for the interned GL function name strings in the various Impl classes, just not super happy with it yet
20131028 16:17:10 <hharrison> It actually turned out to be a gluegen patch
20131028 16:17:44 * hharrison (~chatzilla@anon) Quit (Quit: ChatZilla 0.9.90.1 [Firefox 25.0/20131022211721])
20131028 16:18:34 * hharrison (~chatzilla@anon) has joined #jogamp
20131028 16:56:20 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20131028 17:13:29 <sgothel> @Harvey: Yes .. Bug 776 .. sort of harmonizes the lazy initialization of things (drawable, ctx) in GLAutoDrawable .. otherwise shared context is almost unusable
20131028 17:13:55 <sgothel> further more .. the 'volatile' of context ref and it's handle .. seem to be required .. especially for this use-case
20131028 17:14:37 <sgothel> w/o concurrent context sharing .. it won't matter that much .. but it was not really accurate then to query e.g. ctx.isCreated()
20131028 17:14:53 <sgothel> (w/o a lock) .. but then .. nobody should care
20131028 17:26:57 <sgothel> adding GLSharedContextSetter to SWT GLCanvas .. hmm
20131028 17:27:03 <sgothel> btw .. better name for GLSharedContextSetter ?
20131028 17:27:15 <sgothel> or .. is it clear ?
20131028 17:37:18 * monsieur_max (~maxime@anon) has joined #jogamp
20131028 17:44:42 <hharrison> Looking a bit closer, we explicitly manage our contexts by-hand, so this doesn't affect us much
20131028 17:45:35 <sgothel> Sure .. this is to allow others to benefit from a generic solution in GLAD
20131028 17:45:41 <rmk0> is there a write-up about shared contexts somewhere?
20131028 17:45:59 <sgothel> GLSharedContextSetter has some remarks .. then go to spec .. i.e ..
20131028 17:46:01 <rmk0> i'd like to look into using them soon... doing texture loading on a separate thread and so on
20131028 17:46:29 * rmk0 eyes it
20131028 17:47:08 <hharrison> I can give you a high-level description of what we did in jaamsim...or you can go read the source if you want
20131028 17:47:24 <rmk0> would appreciate it
20131028 17:47:48 <hharrison> And hey, it will give everyone a chance to point and laugh in case I'm doing something dumb!
20131028 17:48:46 <sgothel> .. http://www.opengl.org/registry/specs/ARB/glx_create_context.txt
20131028 17:49:02 <hharrison> At prgram start, we initialize a resource manager that spawns an initial window just for the pruposes of creating our initial context, and a loader thread
20131028 17:49:16 <sgothel> plus .. the general spec .. just mentioning 'objects' .. while being 'unsure' about FBO
20131028 17:49:28 <hharrison> we then use that context and that thread to load all of our assets from then on, textures, large geomtry...etc
20131028 17:49:47 <sgothel> best stable sharing approach is to use an offscreen/dummy drawable and a context ..
20131028 17:49:56 <sgothel> as mentioned in GLSharedContextSetter
20131028 17:50:05 <masterzen> sgothel: hmm, 2.1.1 bug 862 introduced an issue, where jogl doesn't find any GLES2/3 profiles on some android devices (we see that on a sony xperia). I'll send you the details tomorrow morning.
20131028 17:50:39 <hharrison> Whenever we open a new window, or want to render offscreen, we create a new context sharing the original one created at the start
20131028 17:51:08 <rmk0> right
20131028 17:51:08 <hharrison> Then the 'master' context is only ever 'current' on the resource-loading thread
20131028 17:51:11 <sgothel> @Bride: such thing still exist ? wow :) sorry .. - Also - please add your intel workaround to the matching bug report - thx
20131028 17:52:03 <rmk0> how are drivers with multiple threads these days?
20131028 17:52:11 <sgothel> @Harvey: We also makeCurrent the shared ctx when creating the slave ... to lock resources during this little period of time
20131028 17:52:13 <rmk0> i know some ten years ago they were... not usable
20131028 17:52:26 <sgothel> quite OK .. Mesa best
20131028 17:52:42 <rmk0> i read that john carmack built the doom 3 renderer to be threaded in that respect but had to disable multithreading on release
20131028 17:52:48 <sgothel> AMD-X11 worse (bad X11 Display handling)
20131028 17:52:54 <rmk0> ugh
20131028 17:52:55 <hharrison> We have a bit of message-passing glue where you can ask the resource laoder to load something and you can poll/get a callback when it's done
20131028 17:53:36 <rmk0> hharrison: yeah, i was going to say i'd probably combine it with some actor based concurrency... send a message asynchronously and get one back with a reference to the loaded thing when the request is fulfilled
20131028 17:53:55 <rmk0> most of my resources won't be loaded ahead of time, will be more or less streaming constantly
20131028 17:54:05 <hharrison> sgothel: really, yeah, I think we actually ask the resource thread to hand us a new slave context to avoid problems there
20131028 17:54:37 <hharrison> We essentially implemented a 'Future' object for that use
20131028 17:54:47 <rmk0> right
20131028 17:55:19 <rmk0> well, sounds pleasant enough barring driver problems
20131028 17:55:29 <sgothel> @Harvey: sounds great - yes, this patch is 'only' to accommodate the GLAD lifecycle .. nothing special like offering a whole Thread worker :)
20131028 17:55:54 <hharrison> *phew*, thought I was going to have to understand all those patches ;-)
20131028 17:55:59 <sgothel> Plus fixing the offcreen GLAD ctor in factory .. to _not_ create the GLContext right away
20131028 17:56:17 <hharrison> On the other hand, with my Java3d hat on, that may be useful
20131028 17:57:31 <sgothel> sadly .. (see stable driver constraints in GLSharedContextSetter) .. you cannot dtor the 'master' share
20131028 17:57:51 <sgothel> before the 'slaves' .. the spec actually says nothing about it - i.e. it should work
20131028 17:58:04 <sgothel> so this is a driver issue across at least NV (linux, windows)
20131028 17:58:18 <sgothel> you see it in unit tests ..
20131028 17:58:24 <sgothel> (disabled)
20131028 17:59:53 <sgothel> hmm .. have to simplify SWT test case due to the damn 'main' thread issue there
20131028 18:04:23 <rmk0> i can live with that
20131028 18:05:17 <sgothel> :) .. we should find an SWT maintainer .. since me having no interest here either
20131028 18:05:33 <rmk0> is SWT the eclipse toolkit?
20131028 18:05:34 <sgothel> .. and AWT .. oh well, but what can you do ?
20131028 18:05:36 <sgothel> yup
20131028 18:06:19 <rmk0> didn't know people actually... used it
20131028 18:06:36 <sgothel> Eclipse ?
20131028 18:06:39 <rmk0> not seen a SWT program in the wild, i don't think
20131028 18:06:42 <rmk0> beyond eclipse, i mean
20131028 18:06:46 <rmk0> i sort of assumed it was unique to them
20131028 18:07:10 <sgothel> some dare to add plugins to Eclipse maybe ..
20131028 18:09:11 <rmk0> /o\
20131028 18:25:32 <sgothel> still on java7 OSX w/ remote ssh launch when using AWT: 'No X11 DISPLAY variable was set, but this program performed an operation which requires it.'
20131028 18:25:40 <sgothel> grrr
20131028 18:26:12 <sgothel> this says alot about .. the testing coverage ..
20131028 18:39:38 <sgothel> * Note that on OSX version <= 10.9 (tested w/ 10.7.5 and 10.9) concurrent access of shared GL resources
20131028 18:39:38 <sgothel> * may lead to a crash (SISEGV)!
20131028 18:39:38 <sgothel> :(
20131028 18:40:30 <hharrison> Oh come on....who managed to fuck that up?
20131028 18:40:50 <sgothel> result of last local sharedctx testing ..
20131028 18:41:39 <hharrison> Or maybe I'm reading that pessimistically, what kind of shared GL resources?
20131028 18:41:58 <sgothel> simple VBO
20131028 18:42:11 <sgothel> (guaranteed to be able to be shared - spec)
20131028 18:42:31 <sgothel> estSharedContextVBOES2NEWT* async test case .. will disable it for OSX < 10.9
20131028 18:48:28 <rmk0> sgothel: really appreciate the time you put into this stuff
20131028 18:48:31 <rmk0> that probably doesn't get said enough
20131028 18:49:10 <rmk0> discovering all of these issues whilst trying to write a program (assuming something like JOGL didn't exist) would be utter misery
20131028 18:49:11 <sgothel> thx alot .. but hey, we all do that here
20131028 18:49:37 <sgothel> thats the price to pay to claim 'across platform' :)
20131028 18:49:42 <rmk0> yep
20131028 18:50:17 <hharrison> sgothel: yeah, it's a very finicky problem-space you live in
20131028 18:50:29 <hharrison> But it is very much appreciated
20131028 18:52:29 <masterzen> sgothel: I will update bug with my workaround tomorrow.
20131028 18:53:12 <masterzen> sgothel: do you know if the 2.1.2-rc-20131025 contains this commit: https://github.com/sgothel/jogl/commit/e1ffbf2ae6eb837dc1576eedaacbbb68247139f2
20131028 18:53:43 <masterzen> sgothel: I'm wondering if the issue might not be resolved with this commit. I'll open a bug report tomorrow, but I don't want to add noise unnecessarily
20131028 18:56:54 <sgothel> http://jogamp.org/deployment/archive/master/gluegen_737-joal_492-jogl_1120-jocl_867-signed/log/all.artifact.properties.sorted
20131028 18:57:14 <sgothel> https://jogamp.org/chuck/view/fwd/job/jogl/1115/
20131028 18:57:45 <sgothel> yes
20131028 18:58:45 <sgothel> Fix GLContext: getGLProfile() add missing GLES3; getAvailableGLProfile(device, ..) shall use GLProfile.get(device, ..) (commit: e1ffbf2ae6eb837dc1576eedaacbbb68247139f2) (detail / gitweb)
20131028 18:59:19 <sgothel> ^^^ how you can check yourself .. next time via all.artifact.properties.sorted
20131028 19:08:24 <masterzen> sgothel: thanks, that's a good tip!
20131028 19:54:03 * xranby (~xranby@anon) Quit (Ping timeout: 272 seconds)
20131028 19:55:53 * xranby (~xranby@anon) has joined #jogamp
20131028 20:08:28 <masterzen> I entered bug #875, if any of you can have a look. https://jogamp.org/bugzilla/show_bug.cgi?id=875
20131028 20:08:35 <masterzen> I don't really know what is the issue
20131028 20:18:30 * Schrostfutz (~who@anon) has joined #jogamp
20131028 20:18:51 <Schrostfutz> hi, which jogl versions are stable at the moment?
20131028 21:12:39 <monsieur_max> 2.1.1 ?
20131028 21:24:09 <sgothel> @Frostschutz: you may also try the 2.1.2-rc
20131028 21:31:51 * void256 (~void@anon) has joined #jogamp
20131028 21:39:34 <hharrison> I see fedora and debian have been getting further along with splitting out java-headless packages
20131028 22:04:29 <sgothel> OpenJDK8 / profile .. JiGong ..
20131028 22:05:02 <sgothel> OSX shared context: created resources on a 'slave' are not visible on other slaves ..
20131028 22:05:31 <sgothel> if having a 'master-slave1 and master-slave2' .. works on linux/windows
20131028 22:20:08 <hharrison> Good to know, we always create everythign on the master context
20131028 22:20:45 <hharrison> ....I think ;-)
20131028 22:29:43 <sgothel> yeah .. getting to the guts of it now .. at least resulting in a list of such quirks while making sure it's not our fault :)
20131028 22:31:04 * monsieur_max (~maxime@anon) has left #jogamp
20131028 23:12:37 * void256 (~void@anon) Quit (Quit: ChatZilla 0.9.90.1 [Firefox 24.0/20130910160258])
20131029 01:41:26 <sgothel> Last OSX share ctx .. comment invalid @ [(11:05:02 PM) sgothel:...] - buggy test case .. good
20131029 03:38:06 * [Mike] (~Mike]@anon) Quit ()
20131029 03:54:40 * xranby (~xranby@anon) Quit (Read error: Operation timed out)
20131029 03:56:01 * xranby (~xranby@anon) has joined #jogamp
20131029 04:32:26 * [Mike] (~Mike]@anon) has joined #jogamp
20131029 05:05:51 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20131029050551.html