#jogamp @ irc.freenode.net - 20140930 05:05:52 (UTC)


20140930 05:05:52 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20140929050552.html
20140930 05:05:52 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20140930050552.html
20140930 05:12:34 * [Mike] (~Mike]@anon) has joined #jogamp
20140930 06:10:22 * xranby (~xranby@anon) Quit (Ping timeout: 272 seconds)
20140930 06:25:54 * xranby (~xranby@anon) has joined #jogamp
20140930 06:53:18 * monsieur_max (~maxime@anon) has joined #jogamp
20140930 07:01:55 * eclesia (~husky@anon) has joined #jogamp
20140930 07:04:57 <eclesia> good wet morning
20140930 07:08:39 <monsieur_max> eclesia: hehe, yesterday was pretty wet
20140930 07:19:01 <eclesia> monsieur_max: you are at montpellier too ?
20140930 07:26:30 <monsieur_max> eclesia: yup
20140930 08:05:56 <monsieur_max> interesting java information : http://www.oracle.com/technetwork/oracle-labs/program-languages/overview/index.html
20140930 08:07:23 <monsieur_max> and : https://www.openhub.net/p/graal
20140930 08:09:51 <sgothel> thx Max .. sounds like 'llvm/gcc runtime'
20140930 08:10:09 <sgothel> raining here in Bremerhaven now :)
20140930 08:10:31 * [Mike] (~Mike]@anon) Quit ()
20140930 08:10:40 <sgothel> wetty fall arrived
20140930 08:11:45 <sgothel> GPL + LGPL
20140930 08:20:18 * eclesia javafx invoke dynamic exception hell
20140930 08:51:47 * gouessej (5ee4b442@anon) has joined #jogamp
20140930 08:51:59 <gouessej> Hi
20140930 08:52:20 <gouessej> I can't build JOAL: strip:C:\Users\jgouesse\Documents\joal\build\obj\soft_oal.dll: File format not recognized
20140930 08:52:33 <sgothel> Hi .. I will read your stuff soon (Intel/JOGL)
20140930 08:52:40 <sgothel> oh
20140930 08:53:28 <sgothel> on windows .. we still use your injected blobs .. AFAIK .. hmm
20140930 08:53:43 <gouessej> ok
20140930 08:53:55 <gouessej> I can make a pull request for the bug 1078
20140930 08:53:59 <sgothel> sure .. being able to build it for windows would be nice
20140930 08:54:20 <gouessej> but I'd like to discuss a bit before doing so, to be sure we agree about the fix
20140930 08:54:29 <sgothel> yeah .. great, I was holden back by this other Windows / AMD issues .. horrible, PFD set on HDC .. but should not
20140930 08:54:45 <gouessej> yeah, I saw that yesterday :s
20140930 08:54:52 <sgothel> guess we have to wait for a test case there ..
20140930 08:55:15 <sgothel> however .. unit tests showed .. some [1-2] freezes already .. have to double check
20140930 08:55:55 <gouessej> I plan to add a renderer quirk
20140930 08:56:11 <sgothel> in case you can confirm it's not a bug in JOGL .. sure
20140930 08:57:01 <sgothel> good thing .. the quirks can be easily traced in the source code .. so no risk
20140930 08:57:23 <gouessej> the Intel driver refuses to create a pbuffer when the accumulation buffer might be used
20140930 08:57:30 <sgothel> will merge the fix for Intel back to b22x then
20140930 08:57:34 <sgothel> ah ..
20140930 08:57:45 <gouessej> but ...
20140930 08:57:55 <sgothel> Q: .. could it be that we use chosen-caps and not requested-caps ?
20140930 08:58:00 <gouessej> it fixes only the test case using JCanvas3D (Java3D)
20140930 08:58:07 <sgothel> (I had this the other day ..)
20140930 08:58:24 <gouessej> good question
20140930 08:58:31 <sgothel> i.e. that we try to use already chosen caps via the choosing algo .. which may add accum
20140930 08:58:37 <sgothel> but was never requested
20140930 08:58:48 <sgothel> I fixed that for FBO drawables AFAIK
20140930 08:59:05 <sgothel> maybe you can have a 2nd look ..
20140930 08:59:09 <gouessej> oh... you're right, it's not requested as far as I can tell
20140930 08:59:22 <gouessej> How can I check that?
20140930 08:59:50 <sgothel> at some transition point .. in our code .. using the chosenCaps for pbuffer selection
20140930 09:01:07 <gouessej> chosenCaps is passed to fixOffscreenGLCapabilites
20140930 09:01:40 <sgothel> yeah .. at early init .. we setup chosenCaps .. i.e. it's more a fixed requestedCaps
20140930 09:02:07 <sgothel> 38f6915fedb765313c1f4646acf9e13dfbccef36
20140930 09:02:09 <gouessej> the requested capabilities object already has the accum bits set :s
20140930 09:02:19 <sgothel> http://jogamp.org/git/?p=jogl.git;a=commit;h=38f6915fedb765313c1f4646acf9e13dfbccef36
20140930 09:02:21 <sgothel> ah ..
20140930 09:02:41 <sgothel> user requested ?
20140930 09:02:42 <gouessej> :( \jdk_HEAD_64\jre\bin/jawt.dll: file not recognized: File format not recognized
20140930 09:02:54 <gouessej> I can't build JOGL
20140930 09:03:06 <sgothel> something is really odd ..
20140930 09:03:19 <gouessej> I'm able to build it on another machine
20140930 09:03:25 <xranby> \bin/ correct \ / usage?
20140930 09:03:48 <sgothel> thats ok ..
20140930 09:03:59 <gouessej> yes I just add ";mingw64\bin\" into the PATH
20140930 09:04:00 <sgothel> i.e. our ant stuff works on windows ..
20140930 09:04:19 <sgothel> so do I ofc .. JOGL build HowTo ..
20140930 09:04:20 <gouessej> ";C:\mingw64\bin\"
20140930 09:04:39 <sgothel> jogl/make$ vi scripts/make.jogl.all.win64.bat
20140930 09:04:49 <sgothel> set PATH=%JAVA_HOME%\bin;%ANT_PATH%\bin;c:\mingw64\bin;%PATH%
20140930 09:05:08 <sgothel> set JAVA_HOME=c:\jdk1.8.0_20_x64
20140930 09:05:15 <sgothel> set ANT_PATH=C:\apache-ant-1.9.4
20140930 09:05:25 <sgothel> set SOURCE_LEVEL=1.6
20140930 09:05:25 <sgothel> set TARGET_LEVEL=1.6
20140930 09:05:25 <sgothel> set TARGET_RT_JAR=c:\jre1.6.0_30\lib\rt.jar
20140930 09:05:25 <sgothel> :)
20140930 09:06:00 <sgothel> looks in the gluegen, jogl, .. make/scripts folder
20140930 09:06:19 <gouessej> maybe there is a conflict with another version of mingw
20140930 09:06:25 <sgothel> again: 'user requested capabilities object already has the accum bits set' ?
20140930 09:06:32 <sgothel> true ?
20140930 09:06:41 <gouessej> I have to look at the logs agin
20140930 09:06:44 <gouessej> again
20140930 09:06:58 <sgothel> wonder why we don't reduce the findings then ..
20140930 09:07:14 <sgothel> however .. in FBO case we behave the same: no luck - failure
20140930 09:07:22 <sgothel> i.e. req. stuff n/a
20140930 09:08:01 <sgothel> in such case .. Java3D should change - but who today does request an accum buffer ? :)
20140930 09:08:33 <eclesia> <--
20140930 09:08:35 <sgothel> sure - FBO/Pbuffer should behave like our DefaultGLCapsChooser
20140930 09:08:36 <sgothel> :)
20140930 09:08:42 <eclesia> joking ^^
20140930 09:09:13 <sgothel> http://jogamp.org/git/?p=jogl.git;a=commit;h=38f6915fedb765313c1f4646acf9e13dfbccef36 <- validate w/ this ..
20140930 09:10:42 <sgothel> ^^ sorry .. not relevant I guess
20140930 09:11:15 <sgothel> so we have to verify DefaultGLCapsChooser w/ FBO + Pbuffer for 'not so lucky drivers'
20140930 09:11:39 <sgothel> FBO: we can skip I guess (own restricted thingy)
20140930 09:12:51 <gouessej> I still can't build JOGL
20140930 09:15:58 <sgothel> hmm .. pls verify your installed tools .. by testing to execute them .. mingw64 jdk ..
20140930 09:17:08 <gouessej> gcc --version works
20140930 09:17:19 <sgothel> java -version
20140930 09:17:32 <sgothel> and then .. maybe objdump -T jdk\something.dll
20140930 09:17:52 <sgothel> mingw64/bin/objdump .. ofc
20140930 09:19:31 <gouessej> objdump: \jdk_HEAD_64\jre\bin\jawt.dll: not a dynamic object DYNAMIC SYMBOL TABLE: no symbols
20140930 09:19:39 <sgothel> or -t ..
20140930 09:19:45 <sgothel> something should work :)
20140930 09:20:05 <sgothel> but looks like it could identify the dll properly
20140930 09:20:58 <gouessej> I'll build JOGL on another machine with Eclipse as it worked yesterday
20140930 09:21:17 <sgothel> jdk_HEAD_64\jre\bin/jawt.dll: file not recognized: File format not recognized <- as if using 32bit mingw or tools for 64bit stuff
20140930 09:21:23 <gouessej> now I have to find what requests accum
20140930 09:22:01 <sgothel> yeah, my guess [1] user or [2] we re-use chosen glcaps from the chooser algo or driver-default (maybe even onscreen)
20140930 09:22:13 <sgothel> if [2]: our bug
20140930 09:22:22 <sgothel> if [1] .. well .. lets see :)
20140930 09:22:39 <gouessej> the information should be in the logs then
20140930 09:22:46 <sgothel> yup
20140930 09:24:12 <gouessej> https://jogamp.org/bugzilla/attachment.cgi?id=647
20140930 09:26:10 <sgothel> user chosen caps GLCaps[rgba 5/5/5/0, opaque, accum-rgba 0/0/0/0, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[.]]
20140930 09:26:23 <gouessej> with my hack
20140930 09:26:30 <sgothel> duh! :)
20140930 09:26:31 <gouessej> I force accum-rgba 0/0/0/0
20140930 09:26:53 <gouessej> Without my hack it's rather accum-rgba 16/16/16/16
20140930 09:27:33 <gouessej> https://github.com/sgothel/jogl/blob/master/src/jogl/classes/jogamp/opengl/GLDrawableFactoryImpl.java#L184
20140930 09:29:31 <sgothel> I have an idea .. following my ramblings yesterday ..
20140930 09:29:35 <gouessej> The logs without my hacks are here: https://jogamp.org/bugzilla/attachment.cgi?id=643
20140930 09:29:47 <gouessej> Which idea?
20140930 09:29:56 <sgothel> this is at updateGraphicsConfig .. i.e. post 'preselection'
20140930 09:30:41 <sgothel> preselection of glcaps uses that shared HDC and tries to find a suitable caps
20140930 09:30:52 <sgothel> your logs do not include nativewindow.debug=all ?
20140930 09:31:18 <gouessej> it does but some things doesn't work as expected
20140930 09:31:32 <gouessej> don't
20140930 09:32:18 <sgothel> wglARBPFIDs2GLCapabilities: bucket[1 -> 1]: GLCaps[wgl vid 2 arb: rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 0/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[fbo, pbuffer]]
20140930 09:32:22 <gouessej> Is the culprit in com.jogamp.nativewindow.MutableGraphicsConfiguration?
20140930 09:32:33 <sgothel> claiming to have accum w/ pbuffer
20140930 09:33:04 <gouessej> :s
20140930 09:33:15 <sgothel> MutableGraphicsConfiguration: used for our preselection/final PFD setting
20140930 09:34:19 <sgothel> so I guess the preselection is choosing w/ accum, and then it fails when creating pbuffer
20140930 09:35:01 <sgothel> if this is not a bug in our available glcaps algo .. your quirk may address this path .. hmm, but cont reading
20140930 09:35:30 <gouessej> I just want to be sure that adding a quirk is the right thing to do
20140930 09:36:01 <sgothel> trying to understand .. whether its our fault or not, if not our fault - yes, quirk is great
20140930 09:36:16 <sgothel> but I don't want to cover our bugs w/ it :)
20140930 09:36:27 <gouessej> me neither
20140930 09:37:02 <sgothel> user chosen caps GLCaps[rgba 5/5/5/0, opaque, accum-rgba 0/0/0/0, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[.]]
20140930 09:37:29 <sgothel> updateGraphicsConfigurationARB: hdc 0x70010716, pfdIDCount(hdc) 42, capsChosen GLCaps[rgba 5/5/5/0, opaque, accum-rgba 0/0/0/0, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[.]], WIN DOW
20140930 09:37:34 <sgothel> ^^ preselection thingy
20140930 09:38:03 <sgothel> updateGraphicsConfigurationARB: 2 pfd ids, skipCapsChooser false, WINDOW, 2 glcaps updateGraphicsConfigurationARB: Used wglChoosePixelFormatARB to recommend pixel format 4, idx 0, GLCaps[wgl vid 4 arb: rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 16/0/0, dbl, mono , hw, G LProfile[GL2/GL2.hw], on-scr[.]]
20140930 09:38:04 <sgothel> voila
20140930 09:38:24 <gouessej> updateGraphicsConfigurationARB: Used wglChoosePixelFormatARB to recommend pixel format 4, idx 0, GLCaps[wgl vid 4 arb: rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 16/0/0, dbl, mono , hw, G LProfile[GL2/GL2.hw], on-scr[.]]
20140930 09:38:39 <sgothel> GLCaps[wgl vid 4 arb: rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[fbo, pbuffer]]
20140930 09:38:50 <sgothel> claimed to allow pbuffer ..
20140930 09:39:02 <sgothel> so I look at this claim in our code now .. lets see
20140930 09:39:17 <sgothel> and the preselection must honor pbuffer ofc
20140930 09:40:46 <gouessej> Where should I put the quirk then?
20140930 09:41:14 <sgothel> slow down .. :) I review code regarding the last 2 questions .. i.e. might be our bug
20140930 09:41:28 <gouessej> sorry, there is no hurry
20140930 09:50:30 <sgothel> WGLExt.WGL_DRAW_TO_PBUFFER_ARB -> GLGraphicsConfigurationUtil.PBUFFER_BIT -> glcaps (so we set the glcaps pbuffer bit according to found PFD/ARB ) .. hmm, cont.
20140930 09:58:42 <gouessej> ok
20140930 10:00:05 <sgothel> so that is one bug in the driver .. now checking whether we specify pbuffer in preselection
20140930 10:00:31 <sgothel> WindowsAWTWGLGraphicsConfigurationFactory: Found new AWT PFD ID 4 -> WindowsWGLGraphicsConfiguration[DefaultGraphicsScreen[WindowsGraphicsDevice[type .windows, connection decon, unitID 0, handle 0x0,
20140930 10:00:31 <sgothel> owner false, NullToolkitLock[]], idx 0], pfdID 4, ARB-Choosen true,
20140930 10:00:31 <sgothel> requested GLCaps[rgba 5/5/5/0, opaque, accum-rgba 0/0/0/0, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[.]],
20140930 10:00:31 <sgothel> chosen GLCaps[wgl vid 4 arb: rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[.]]]
20140930 10:00:41 <sgothel> ^^ since it doesn't look like it
20140930 10:03:27 <gouessej> Why doesn't it like it?
20140930 10:04:19 <sgothel> I miss 'pbuffer' being mentioned .. however, a few lines above pbuffer (vid 4 arb) is mentioned
20140930 10:09:56 <sgothel> https://github.com/sgothel/jogl/blob/master/src/jogl/classes/jogamp/opengl/GLDrawableFactoryImpl.java#L184 <- right
20140930 10:10:22 <sgothel> I see you stack trace .. at javax.media.j3d.JoglPipeline.createAwtGraphicsConfiguration(JoglPipeline.java:8635)
20140930 10:10:30 <gouessej> you reassure me, I'm not completely mad :)
20140930 10:11:01 <sgothel> so where you are coming from .. calling into WindowsAWTWGLGraphicsConfigurationFactory.chooseGraphicsConfigurationImpl
20140930 10:11:10 <sgothel> is 'pbuffer' set ?
20140930 10:11:32 <gouessej> yes
20140930 10:11:42 <sgothel> i.e. you do not use our factory model to define the caps for offscreen right ?
20140930 10:11:55 <sgothel> so why don't I see pbuffer in req. bits ?
20140930 10:12:06 <gouessej> chosenCaps already has the pbuffer enable flag set
20140930 10:12:24 <sgothel> user chosen caps GLCaps[rgba 5/5/5/0, opaque, accum-rgba 0/0/0/0, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[.]]
20140930 10:12:49 <sgothel> requested GLCaps[rgba 5/5/5/0, opaque, accum-rgba 0/0/0/0, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[.]], chosen GLCaps[wgl vid 4 arb: rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 16/0/0, dbl, mono , hw, GLProfile[GL2/GL2.hw], on-scr[.]]]
20140930 10:13:18 <gouessej> requested GLCaps[rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 16/0/0, one, mono , hw, GLProfile[GL2/GL2.hw], offscr[fbo]], chosen GLCaps[rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 16/0/0, one, mono , hw, GLProfile[GL2/GL2.hw], offscr[pbuffer]]]
20140930 10:14:19 <sgothel> https://jogamp.org/bugzilla/attachment.cgi?id=643
20140930 10:14:24 <gouessej> yes
20140930 10:14:33 <gouessej> just before creating the pbuffer
20140930 10:14:39 <sgothel> yes .. that late :)
20140930 10:14:44 <sgothel> but earlier ..
20140930 10:14:59 <sgothel> (that late we do our mutations .. but caps already chosen)
20140930 10:15:17 <gouessej> ok
20140930 10:15:40 <sgothel> JoglPipeline.createAwtGraphicsConfiguration( .. <- when you do this, you need to follow that https://github.com/sgothel/jogl/blob/master/src/jogl/classes/jogamp/opengl/GLDrawableFactoryImpl.java#L184
20140930 10:15:52 <sgothel> i.e. patch the requested caps to match on-/offscreen .. etc
20140930 10:16:20 <sgothel> maybe that will do the trick - even if Intel driver claims to support accum w/ pbuffer (see avail caps list)
20140930 10:17:30 <gouessej> https://github.com/hharrison/java3d-core/blob/master/src/classes/share/javax/media/j3d/JoglPipeline.java#L8631 ??
20140930 10:17:35 <sgothel> b/c that earlier we set the requested caps bits .. incl. pbuffer
20140930 10:18:24 <sgothel> the incoming 'capabilities' (requested I assume) shall reflect the offscreen caps if desired
20140930 10:18:45 <sgothel> thats what we do .. e.g. in our factory for creating offscreen stuff ..
20140930 10:18:52 <gouessej> but what should I do when I don't use Java3D? Someone reproduced this bug with a simple offscreen drawable
20140930 10:18:53 <sgothel> the fixOffscreen*(..)
20140930 10:19:28 <sgothel> still it could be that the Intel driver returns accum bits when using the ARB chooser .. a driver bug, sure
20140930 10:19:43 <sgothel> however, when using the chooser .. the bits must be set
20140930 10:19:50 <sgothel> i.e. pbuffer etc
20140930 10:19:53 <sgothel> and they are not
20140930 10:20:10 <sgothel> i.e. give the chooser a chance :)
20140930 10:20:28 <sgothel> a simple jogl test should show .. w/ offscreen pbuffer
20140930 10:20:41 <sgothel> i.e. if that unit tests works .. the Java3D fix should do it
20140930 10:20:53 <sgothel> since we set the pbuffer bit in factory
20140930 10:20:55 <gouessej> The printing test exibits this bug too
20140930 10:21:06 <sgothel> could be another bug .. of ours :)
20140930 10:21:16 <sgothel> i.e. using already choosing caps
20140930 10:21:19 <gouessej> TestOffscreen02BitmapNEWT too
20140930 10:21:30 <sgothel> oh
20140930 10:21:50 <gouessej> I can test again to be 100% sure
20140930 10:21:51 <sgothel> would be nice to see a failing log for a simple test of ours
20140930 10:21:58 <sgothel> pbuffer + bitmap pls
20140930 10:22:37 <sgothel> which test ? lets see
20140930 10:22:49 <gouessej> I'm wrong. Forcing pbuffers causes another bug
20140930 10:22:59 <gouessej> Forcing bitmaps too
20140930 10:23:10 <gouessej> https://jogamp.org/bugzilla/show_bug.cgi?id=1078#c18
20140930 10:23:30 <sgothel> TestGLAutoDrawableFactoryGL2OffscrnCapsNEWT <- that one
20140930 10:23:43 <sgothel> tests everything .. almost
20140930 10:23:54 <gouessej> Ok I'll try this one
20140930 10:23:56 <sgothel> (just not accum :)
20140930 10:24:00 <gouessej> I tested so much things
20140930 10:24:09 <sgothel> thank you!
20140930 10:24:45 <gouessej> You're welcome
20140930 10:24:59 <sgothel> ha - so it works w/ pbuffer being set in req. caps _EARLY_ .. great
20140930 10:25:05 <sgothel> (your comments 18+)
20140930 10:25:30 <gouessej> I remember that this test fails for sure: https://jogamp.org/bugzilla/show_bug.cgi?id=1078#c10
20140930 10:25:33 <sgothel> goody .. so we have one fix in Java3D .. via using fixOffscreenGLCapabilities(..)
20140930 10:25:48 <sgothel> then we need to see about printing .. which is tricky ..
20140930 10:26:30 <sgothel> we probably learn that we cannot use offscreen then, since we cannot find a compatible glcaps for swapping the context to it
20140930 10:26:41 <sgothel> i.e. you cannot swap a context to a less capable glcaps
20140930 10:27:00 <sgothel> Note: we swap context onscreen -> offscreen for hidpi printing, using tile renderer here
20140930 10:27:12 <sgothel> quite fancy if I may say so :)
20140930 10:27:54 <gouessej> I'm going to run the tests very soon...
20140930 10:28:08 <sgothel> Julien, so you already have answered all this now yourself YESTERDAY! :)
20140930 10:28:27 <sgothel> so neither of us is insane .. same conclusion, great!
20140930 10:29:04 <sgothel> TBH: I hate the PFD selection code path .. (Windows)
20140930 10:29:21 <gouessej> TestTiledPrintingGearsAWT
20140930 10:29:40 <sgothel> TestGLAutoDrawableFactoryGL2OffscrnCapsNEWT <- works ?
20140930 10:29:53 <gouessej> JOGL still building without my hacks...
20140930 10:29:56 <sgothel> (baby steps)
20140930 10:30:13 <sgothel> printing .. GLCanvas ? let me see ..
20140930 10:30:19 <gouessej> yes
20140930 10:30:21 <sgothel> (i.e. we need a fallback)
20140930 10:30:27 <gouessej> GLCanvas.setupPrint()...
20140930 10:30:35 <sgothel> yup
20140930 10:31:05 <sgothel> GLDrawableUtil.isSwapGLContextSafe(..) <- the magic one
20140930 10:31:45 <sgothel> which covers accum bits
20140930 10:32:59 <sgothel> *but only on the requested side* :)
20140930 10:34:00 <sgothel> changing things here needs alot of validation .. oh dear
20140930 10:34:18 <sgothel> i.e. can we base newGLADCaps on the requested caps only ?
20140930 10:35:18 <sgothel> - or - you show me the stack trace, and we attempt to recover, maybe more safe
20140930 10:35:27 <gouessej> ok ...
20140930 10:35:32 <gouessej> look at the bug report
20140930 10:35:54 <gouessej> for the printing: https://jogamp.org/bugzilla/attachment.cgi?id=648
20140930 10:38:47 <gouessej> Now I'm running TestGLAutoDrawableFactoryGL2OffscrnCapsNEWT
20140930 10:39:19 <sgothel> good - direct exception from factory.createOffscreenAutoDrawable(..) ..
20140930 10:39:30 <sgothel> before swapping, I guess we can deal w/ that :)
20140930 10:45:04 <gouessej> done
20140930 10:45:21 <gouessej> https://jogamp.org/bugzilla/attachment.cgi?id=649
20140930 10:45:27 <gouessej> TestGLAutoDrawableFactoryGL2OffscrnCapsNEWT
20140930 10:45:33 <gouessej> it seems to work :s
20140930 10:46:40 <sgothel> if you see the pretty pictures .. w/o debug stuff it would be more obvious :)
20140930 10:47:17 <sgothel> great .. working on a fix for printing ..
20140930 10:49:14 <sgothel> chosen GLCaps[wgl vid 4 arb: rgba 8/8/8/8, opaque, accum-rgba 16/16/16/16, dp/st/ms 16/0/0, one, mono , hw, GLProfile[GL2/GL2.hw], offscr[pbuffer]]]
20140930 10:49:18 <sgothel> ^^ so it is possible
20140930 10:49:59 <gouessej> 2 suitable pixel formats found [junit] pixel format 3 (index 0): GLCaps[wgl vid 3 arb: rgba 8/8/8/8, opaque, accum-rgba 0/0/0/0, dp/st/ms 16/0/0, one, mono , hw, GLProfile[GL2/GL2.hw], offscr[pbuffer]] [junit] pixel format 5 (index 1): GLCaps[wgl vid 5 arb: rgba 8/8/8/8, opaque, accum-rgba 0/0/0/0, dp/st/ms 24/8/0, one, mono , hw, GLProfile[GL2/GL2.hw], offscr[pbuffer]]
20140930 10:50:18 <sgothel> ah .. sorry wrong log file :)
20140930 10:50:20 <sgothel> great
20140930 10:50:28 <gouessej> still accum-rgba 0/0/0/0
20140930 10:50:51 <sgothel> that is at least one driver bug then .. setting PBUFFER bit for accum != 0
20140930 10:51:05 <sgothel> but the drivers selection algo .. fixes it :)
20140930 10:51:55 <sgothel> so we still could use a quirk, to disable pbuffer + accum!=0 in PFD -> caps
20140930 10:52:11 <sgothel> i.e. if user like to use her own 'chooser'
20140930 10:52:59 <gouessej> The problem is that I don't think that pbuffers should be skipped. If bitmaps are used, it will be very slow
20140930 10:53:15 <sgothel> pardon me ?
20140930 10:53:31 <gouessej> Maybe I don't fully understand the consequences of your suggestions
20140930 10:53:37 <sgothel> printing ?
20140930 10:53:48 <gouessej> not for printing
20140930 10:54:06 <sgothel> oh .. you mean: so we still could use a quirk, to disable pbuffer + accum!=0 in PFD -> caps
20140930 10:54:07 <sgothel> ?
20140930 10:54:37 <sgothel> the driver claims a PFD/caps w/ accum!=0 is pbuffer capable, which is not
20140930 10:55:01 <gouessej> yes you're right
20140930 10:55:09 <sgothel> so when we query the caps by PFD/ARB .. we can use that quirk to clear the pbuffer flag if accum!=0
20140930 10:55:24 <sgothel> i.e. only valid pbuffer caps will be listed then
20140930 10:55:28 <gouessej> ok
20140930 10:55:35 <gouessej> it seems to be fine
20140930 10:55:42 <gouessej> thanks for the explanation
20140930 10:56:00 <gouessej> then another capabilities (more suitable) will ne picked
20140930 10:56:12 <gouessej> another capabilities object will be picked
20140930 10:56:29 <sgothel> yes - for the 'user chooser algo' case .. or if we use ours
20140930 10:56:54 <gouessej> Where should be put in order to minimize the impact?
20140930 10:57:12 <gouessej> Where should I put the quirk then?
20140930 10:57:22 <sgothel> WindowsWGLGraphicsConfiguration: AttribList2GLCapabilities and PFD2GLCapabilities and PFD2GLCapabilitiesNoCheck
20140930 10:57:35 <sgothel> ^^ producing GLCaps based on GDI PFD or ARB bits
20140930 10:57:58 <gouessej> ok :)
20140930 10:59:10 <gouessej> I have to modify GLRendererQuirks (the quirks count too) and then I have to modify these methods in WindowsWGLGraphicsConfiguration
20140930 10:59:17 <sgothel> it would be a great start to add the quirk in the quirk class w/ proper VERSION/RENDERER infos .. etc
20140930 10:59:43 <sgothel> yup - while the static count is no more used (compatibility)
20140930 11:00:18 <sgothel> (you do it in master branch .. right)
20140930 11:00:26 <sgothel> I will do the backport later
20140930 11:01:35 <gouessej> Where will I have to check whether or not to apply the quirk? directly in WindowsWGLGraphicsConfiguration? I assume that there is something somewhere to decide whether a quirk is suitable for a given configuration
20140930 11:02:21 <sgothel> hmm .. yeah, there is a possibility to return null as well
20140930 11:02:38 <sgothel> look at the WGLGLCaps*.setValuesByARB for started
20140930 11:02:53 <sgothel> AttribList2GLCapabilities
20140930 11:03:07 <sgothel> there you can add the quirk based decision ..
20140930 11:03:37 <sgothel> then we have to see whether the GDI/PFD path needs to be done as well ..
20140930 11:03:38 <gouessej> WGLGLCapabilities?
20140930 11:03:43 <sgothel> yup
20140930 11:04:00 <sgothel> (call tree of setValuesByARB)
20140930 11:04:10 <gouessej> Why is this class public?
20140930 11:04:24 <sgothel> well ..
20140930 11:04:36 <sgothel> it's in our 'private package' though :)
20140930 11:04:53 <sgothel> lazy I guess .. i.e. undecided whether its used outside of package
20140930 11:05:12 <sgothel> we could do a whole review based on that .. go ahead :)
20140930 11:05:21 <sgothel> have to go for around 2-3 hours .. back later, laters .. great hacking
20140930 11:05:33 <gouessej> thank you very much
20140930 11:40:04 * monsieur_max1 (~maxime@anon) has joined #jogamp
20140930 11:42:51 * monsieur_max (~maxime@anon) Quit (Ping timeout: 272 seconds)
20140930 11:47:59 * monsieur_max1 is now known as monsieur_max
20140930 12:03:36 * monsieur_max1 (~maxime@anon) has joined #jogamp
20140930 12:03:42 * monsieur_max1 (~maxime@anon) Quit (Client Quit)
20140930 12:24:06 * eclesia (~husky@anon) Quit (Ping timeout: 244 seconds)
20140930 12:34:03 <gouessej> Addition of the quirk: ok. Modification of WindowsWGLGraphicsConfiguration.AttribList2GLCapabilities: ok.
20140930 12:34:36 <gouessej> I set it in GLContextImpl
20140930 12:38:36 * eclesia (~husky@anon) has joined #jogamp
20140930 13:05:04 <gouessej> Modification of WindowsWGLGraphicsConfiguration.PFD2GLCapabilities: ok.
20140930 13:05:33 <gouessej> Modification of WindowsWGLGraphicsConfiguration.PFD2GLCapabilitiesNoCheck: Actually, it should remain unchanged.
20140930 13:49:44 <gouessej> I can't create a shared resource very early, it hangs
20140930 15:00:37 <sgothel> back
20140930 15:00:55 <sgothel> there should be no change w/ 'shared resource' required IMHO .. hmm
20140930 15:01:32 <sgothel> sounds like you are done .. if you like I can continue - then git pull req.
20140930 15:17:26 <gouessej> ok
20140930 15:17:46 <gouessej> sorry I use 2 machines
20140930 15:18:06 <gouessej> There is asomething annoying...
20140930 15:18:19 <gouessej> There is a broken code path
20140930 15:18:28 <gouessej> NPE when no caps can be returned
20140930 15:19:22 <gouessej> Here: https://github.com/sgothel/jogl/blob/master/src/jogl/classes/jogamp/opengl/windows/wgl/WindowsWGLGraphicsConfigurationFactory.java#L524
20140930 15:20:09 <gouessej> How am I intended to know whether a quirk has to be used without a shared resource?
20140930 15:20:28 <gouessej> I can put my patch into the bug report, you'll see what I mean
20140930 15:24:50 <gouessej> At first, I added a quirk into com/jogamp/opengl/GLRendererQuirks.java
20140930 15:25:34 <gouessej> Then, I detected the case in which this quirk should be added into the list of quirks into jogamp/opengl/GLContextImpl.java
20140930 15:26:19 <gouessej> After that, I modified WindowsWGLGraphicsConfiguration, I needed to check whether the quirk was activated
20140930 15:26:57 <gouessej> I used this code: sharedResource.getRendererQuirks().exist( GLRendererQuirks.NoPBufferWithAccum )
20140930 15:28:14 <gouessej> Actually, fixing bugs isn't very fun but I need to fix this one as soon as possible
20140930 15:32:00 <gouessej> How should I retrieve the GLRendererQuirks without using a shared resource when I have no access to the default device?
20140930 15:35:41 <gouessej> Do you have a machine with a similar driver?
20140930 15:37:45 <gouessej> I should have added several attachments instead of posting my patch as is: https://jogamp.org/bugzilla/show_bug.cgi?id=1078#c25
20140930 15:39:11 * hharrison (~chatzilla@anon) has joined #jogamp
20140930 15:39:35 <hharrison> Morning all, finally back on my feet
20140930 15:39:56 <hharrison> lots of backlog....anything I should be applying to the Java3d side?
20140930 15:40:17 <hharrison> Or will you two finish fighting it out and just spoon feed it to me?
20140930 15:40:40 <gouessej> Hi
20140930 15:40:57 <eclesia> wait until they finish the soup
20140930 15:41:40 <hharrison> :-)
20140930 15:41:44 <gouessej> hharrison: You can set the accum bits to zero in the meantime if you want but it won't be necessary when the real fix is out
20140930 15:42:12 <hharrison> I'll just wait it out for you and sven to tell me what the right thing to do is
20140930 15:42:31 <gouessej> https://jogamp.org/bugzilla/show_bug.cgi?id=1078#c22
20140930 15:43:15 <gouessej> hharrison: Then, you'll have nothing to do except advising developers to use the very latest version of JogAmp as usual
20140930 15:43:40 <gouessej> I tested with a simple test case
20140930 15:43:48 <gouessej> it just creates a JCanvas3D
20140930 15:44:24 <hharrison> Hmm, I wonder if that was what Emmanual was hitting, need to go find his backtrace
20140930 15:44:48 <gouessej> no suitable format?
20140930 15:44:55 <hharrison> I thought it was from the changes made in -pre11 that redid the format selection
20140930 15:45:15 <hharrison> But it was only the JCanavs3D case that wanted pbuffer
20140930 15:45:16 <gouessej> it was reproducible with pre9
20140930 15:45:23 <gouessej> I checked that
20140930 15:45:26 <hharrison> ahh
20140930 15:45:34 <hharrison> nevermind then
20140930 15:46:43 <gouessej> I find some parts of Java3D not very robust. I fear that it will remain inherently weaker than scenegraphs that rely less on offscreen rendering
20140930 15:46:59 <gouessej> GLJPanel works better than JCanvas3D
20140930 15:47:47 <gouessej> I undestand why it was designed that way but it means that we might have some hard time in the future
20140930 15:48:26 <gouessej> There is probably some configurations that don't support Java3D even with OpenGL >= 1.2
20140930 15:48:34 <gouessej> "There are"
20140930 15:48:35 <gouessej> :s
20140930 15:55:54 <gouessej> The most straightforward part of the patch is here: https://jogamp.org/bugzilla/attachment.cgi?id=650&action=diff
20140930 15:57:44 <gouessej> hharrison: Canvas3D has several code paths that use offscreen rendering too, doesn't it?
20140930 15:59:03 <hharrison> Yep, everything goes through Canvas3d....but for the offscreen case it delegates to an offscreen surface, Canvas3D is the 'addressable' object
20140930 15:59:42 <hharrison> From Java3D's perspective, it's all Canvas3D....which is essentially a multiplexer to the correct pipeline method
20140930 16:04:34 * eclesia (~husky@anon) has left #jogamp
20140930 16:06:28 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20140930 16:11:32 * gouessej (5ee4b442@anon) Quit (Quit: Page closed)
20140930 17:21:52 * monsieur_max (~maxime@anon) has joined #jogamp
20140930 19:11:51 <sgothel> good morning - back from loooong nap :)
20140930 19:12:00 <sgothel> Great to see you again, Harvey.
20140930 19:12:16 <sgothel> Reading ^^ ..
20140930 19:12:39 <sgothel> ah .. yeah, maybe you can add an API, like any GLAutoDrawable to be hooked w/ one 'Java3D Context' ?
20140930 19:12:59 <sgothel> then you can use any GLCanvas, GLJPanel or even NEWT GLWindow ..
20140930 19:22:57 * xranby (~xranby@anon) Quit (Ping timeout: 246 seconds)
20140930 19:43:09 * xranby (~xranby@anon) has joined #jogamp
20140930 19:53:32 * xranby (~xranby@anon) Quit (Ping timeout: 245 seconds)
20140930 19:55:19 * xranby (~xranby@anon) has joined #jogamp
20140930 20:46:56 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20140930 21:32:56 * [Mike] (~Mike]@anon) has joined #jogamp
20141001 01:06:52 * xranby (~xranby@anon) Quit (Ping timeout: 245 seconds)
20141001 01:08:55 * xranby (~xranby@anon) has joined #jogamp
20141001 04:33:29 * [Mike] (~Mike]@anon) Quit ()
20141001 05:05:53 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20141001050553.html