Compiling experimental-eglfs-jw
Tyler Mandry
tmandry at gmail.com
Tue Aug 16 18:28:35 UTC 2011
Got it running in raster at least! See below.
On Aug 16, 2011, at 4:13 AM, Thomas Senyk wrote:
> I've added the mailing-list.
>
>
> On Tuesday, August 16, 2011 03:54:38 AM ext Tyler Mandry wrote:
>> On Aug 16, 2011, at 3:22 AM, Thomas Senyk wrote:
>>> Never tried opengl es1
>>> ... I don't see the point, as I've never seen an opengl es1 android
>>> device. Is there one?
>>
>> I'm interested because I have some es1 code that I want to use with Qt. It
>> runs on Android already. Just have to get it to play nice with Qt. Every
>> Android device that support GLES 2 also supports GLES 1.1 and 1.0. It's a
>> documented guarantee.
>
> A, OK! Good information :)
>
>>> Besides that:
>>>> E/libEGL ( 4822): called unimplemented OpenGL ES API
>>>
>>> sounds a lot like "There is no OpenGL ES1!"
>>
>> I'll take a closer look at what's going on.
>
> Would be interesting .... If it's guaranteed then I guess the reference device
> (nexus one) should have it ;)
>
> I will probably have no time this week, but maybe I can give it a try next
> week. Would be interesting to see.
> Although "legacy code" is the only good use-case for that I think.
I would disagree; a lot of people still use GLES v1 because it's "easier" and gives you something to start with. GLES v2 isn't a superset; you have to make custom shaders to do anything, even if you want to just draw lines on the screen. It gives more flexibility but at a cost.
>
>
> How does you setup look like ... You want to use QGLWidget (fullscreen), put
> your GL-code into paintGL() for rendering?
Something like that. The code is flexible and I can easily work it into that kind of setup.
>
>>
>>> About raster:
>>> Good questions ... have you tried to get debugging running?
>>
>> Not until now. I'm pretty certain that it's caused by this:
>> JNI WARNING: inst fieldID 0x4218e084 not valid for class
>> Landroid/view/Surface; The crash signal is SIGSEGV and the fault adds is
>> 0xdeadd00d which indicates that the VM aborting is what caused the app to
>> die.
>>
>> Here's a backtrace from logcat, when using libandroid-9.so. Not very useful.
>> I/DEBUG ( 65): #00 pc 00046240 /system/lib/libdvm.so I/DEBUG
>> ( 65): #01 pc 0003760c /system/lib/libdvm.so I/DEBUG (
>> 65): #02 pc 00038522 /system/lib/libdvm.so I/DEBUG ( 65):
>> #03 pc 00038c12 /system/lib/libdvm.so I/DEBUG ( 65):
>> #04 pc 00033fcc /system/lib/libandroid_runtime.so I/DEBUG ( 65):
>> #05 pc 00054f50 /system/lib/libandroid_runtime.so I/DEBUG ( 65):
>> #06 pc 0000057e /system/lib/libjnigraphics.so
>>
>> Interestingly, this is the same error I got when I tried loading
>> libandroidGL-5.so and libandroid-5.so. Changing to libandroidGL-9.so fixed
>> that problem and the program runs (without displaying anything because of
>> GL errors.) But running with libandroid-9.so for raster pulls up this error
>> again.
>>
>> Here's a backtrace when using libandroid-5.so:
>> I/DEBUG ( 65): #00 pc 00046240 /system/lib/libdvm.so
>> I/DEBUG ( 65): #01 pc 0003760c /system/lib/libdvm.so
>> I/DEBUG ( 65): #02 pc 00038522 /system/lib/libdvm.so
>> I/DEBUG ( 65): #03 pc 00038c12 /system/lib/libdvm.so
>> I/DEBUG ( 65): #04 pc 00033fcc
>> /system/lib/libandroid_runtime.so I/DEBUG ( 65): #05 pc
>> 00054f50 /system/lib/libandroid_runtime.so I/DEBUG ( 65): #06
>> pc 0000ba88 /data/local/qt/plugins/platforms/android/libandroid-5.so
>>
>> Aha.. This looks more useful. And with gdb I got a code listing:
>>
>> (gdb) list *0xba88
>> 0xba88 is in AndroidBitmap_getInfo
>> (../../src/native/graphics/jni/bitmap.cpp:26). 21
>> AndroidBitmapInfo* info) {
>> 22 if (NULL == env || NULL == jbitmap) {
>> 23 return ANDROID_BITMAP_RESULT_BAD_PARAMETER;
>> 24 }
>> 25
>> 26 SkBitmap* bm = GraphicsJNI::getNativeBitmap(env, jbitmap);
>> 27 if (NULL == bm) {
>> 28 return ANDROID_BITMAP_RESULT_JNI_EXCEPTION;
>> 29 }
>> 30
>>
>> Hopefully that will give you more of the information you need. I could keep
>> hacking at it but it's very late here..
>
> So this is backtrace from libandroid-5.so, right?
> If so, this is a known problem when using a <8 plugin on a >=8 device.
>
> If not ... I guess this is a bug in our code :)
>
Yep. So nothing new here then.
>
> A general advice:
> Have you made sure that all linking and including to (Qt)OpenGl was removed?
> (best starting point: projects file remove the QT += opengl)
>
I had done all that on the C++ side. But I looked again and what I missed was that in QtActivity.java, line 135 was "m_surface.applicationStared( true );" where the param is "usesGL". Changing that to false allows it to run in raster mode. This really ought to be commented on or (better yet) not hardcoded.
So it's running! Raster mode is painfully slow :) Now to see if we can get OpenGL...
>
> Greets
> Thomas
>
>
>>
>> Thanks,
>> Tyler
>>
>>> I'm from Germany :)
>>>
>>>
>>> Greets
>>> Thomas
>>>
>>> On Monday, August 15, 2011 04:00:21 PM ext Tyler Mandry wrote:
>>>> Yes, I figured this out after awhile. I ended up compiling on
>>>> experimental, which looks like it's the same as master right now,
>>>> unless my git-fu is lacking.
>>>>
>>>> I've gotten everything compiled, but haven't gotten qmlshowcase to
>>>> run. I'm on OS X (Lion). Here's what I've done so far:
>>>>
>>>> Changed TARGET on line 61 of qmlshowcase.pro from "./bin/qmlshowcase"
>>>> to
>>>> just "qmlshowcase" The app was not finding the app library before I
>>>> did
>>>> this. Also, there seems to be a bug that is causing the build to stick
>>>> that library in "<build-dir>/libs./bin/". It creates a new directory
>>>> called "libs." instead of getting rid of the . or adding a slash
>>>> before it. Removing the path from the target fixed this problem.
>>>>
>>>> Added these lines to the top of startApp() in QtActivity.java:
>>>> getIntent().putExtra("use_local_qt_libs", "true");
>>>> getIntent().putExtra("load_local_libs",
>>>>
>>>> "plugins/platforms/android/libandroid-9.so"); Qt Creator is failing to
>>>> start my application (presumably with the "use_local_qt_libs" intent
>>>> set to true) when I have the option selected to deploy local libs.
>>>> For some reason it says it's starting the app, and nothing happens on
>>>> the Android end. (It does push the libs to the device just fine,
>>>> however.) So I'm manually injecting the option. Next, the app is not
>>>> loading the Android plugin. I don't see any evidence of it trying. As
>>>> a result I force it to load the android-9 lib. (I change it to
>>>> libandroidGL-9.so when trying to use GL.)
>>>>
>>>> So far it fails in both raster and OpenGL modes. Log snippet for
>>>> raster:
>>>>
>>>> D/Qt ( 4768): environmentString
>>>> "HOME=/data/data/eu.licentia.necessitas.industrius.example.qmlshowcase
>>>> /file
>>>> s TMPDIR=/data/data/eu.licentia.necessitas.industrius.example.qmlshow
>>>> case/fi
>>>> les CACHE_PATH=/data/data/eu.licentia.necessitas.industrius.example.q
>>>> mlshowc ase/cache QML_IMPORT_PATH=/data/local/qt/imports
>>>
>>> QT_PLUGIN_PATH=/data/local/
>>>
>>>> qt/plugins"
>>>> ("HOME=/data/data/eu.licentia.necessitas.industrius.example.qmlshowcas
>>>> e/fil es",
>>>> "TMPDIR=/data/data/eu.licentia.necessitas.industrius.example.qmlshowca
>>>> se/fi les",
>>>> "CACHE_PATH=/data/data/eu.licentia.necessitas.industrius.example.qmlsh
>>>> owcas e/cache", "QML_IMPORT_PATH=/data/local/qt/imports",
>>>> "QT_PLUGIN_PATH=/data/local/qt/plugins") D/Qt ( 4768):
>>>> paramsString
>>>> "QtApp -platform android"
>>>> W/Qt ( 4768): WARNING: QApplication was not created in the main()
>>>> thread. D/Qt ( 4768): QAndroidPlatformIntegrationPlugin::create
>>>> "android" D/Qt ( 4768):
>>>> QAndroidPlatformIntegration::QAndroidPlatformIntegration(): creating
>>>> QAndroidPlatformScreen => Using Raster (Software) for painting D/Qt
>>>> ( 4768): QFbScreen::setPhysicalSize QSize(74, 124)
>>>> D/Qt ( 4768): QFbScreen::setGeometry QRect(0,0 480x800)
>>>> D/Qt ( 4768):
>>>> QAndroidPlatformScreen::QAndroidPlatformScreen():QFbScreen() D/Qt
>>>> (
>>>> 4768): setDisplayMetrics
>>>> D/Qt ( 4768): setDisplayMetrics done
>>>> D/Qt ( 4768): setDesktopSize
>>>> D/Qt ( 4768): setDesktopSize done
>>>> I/Qt JAVA ( 4768): surfaceChanged: 480,762
>>>> D/Qt ( 4768): setDisplayMetrics
>>>> D/Qt ( 4768): setDisplayMetrics done
>>>> D/Qt ( 4768): setDesktopSize
>>>> D/Qt ( 4768): setDesktopSize done
>>>> I/Qt JAVA ( 4768): surfaceChanged: 480,762
>>>> D/Qt ( 4768): setDisplayMetrics
>>>> D/Qt ( 4768): setDisplayMetrics done
>>>> D/Qt ( 4768): setDesktopSize
>>>> D/Qt ( 4768): setDesktopSize done
>>>> I/ActivityManager( 96): Displayed
>>>> eu.licentia.necessitas.industrius.example.qmlshowcase/eu.licentia.nece
>>>> ssita s.industrius.QtActivity: +365ms (total +1m41s955ms)
>>>> W/IInputConnectionWrapper( 4148): showStatusIcon on inactive
>>>> InputConnection D/Qt ( 4768): createPlatformWindow false
>>>> D/Qt ( 4768): setFullScreen false
>>>> D/Qt ( 4768): QFbScreen::setPhysicalSize QSize(74, 124)
>>>> D/Qt ( 4768): QFbScreen::setGeometry QRect(0,0 480x762)
>>>> D/Qt ( 4768): QFbScreen::setPhysicalSize QSize(74, 124)
>>>> D/Qt ( 4768): QFbScreen::setGeometry QRect(0,0 480x762)
>>>> D/Qt ( 4768): QFbScreen::setPhysicalSize QSize(74, 124)
>>>> D/Qt ( 4768): QFbScreen::setGeometry QRect(0,0 480x762)
>>>> W/dalvikvm( 4768): JNI WARNING: inst fieldID 0x4218e084 not valid for
>>>> class Landroid/view/Surface; I/dalvikvm( 4768): "Thread-11" prio=5
>>>> tid=9 RUNNABLE I/dalvikvm( 4768): | group="main" sCount=0 dsCount=0
>>>> obj=0x40526710 self=0x300138 I/dalvikvm( 4768): | sysTid=4777
>>>> nice=0 sched=0/0 cgrp=default handle=2776304 I/dalvikvm( 4768): |
>>>> schedstat=( 1585815437 269531244 322 )
>>>> I/dalvikvm( 4768): at dalvik.system.NativeStart.run(Native Method)
>>>> I/dalvikvm( 4768):
>>>> E/dalvikvm( 4768): VM aborting
>>>>
>>>> OpenGL might well be failing because I configured it with "-opengl
>>>> es1" and not es2. From what I gather this should be possible to do,
>>>> as far as mainstream Qt goes, but maybe the eglfs plugin doesn't
>>>> configure it for es1 or something. Here's the log:
>>>>
>>>> D/Qt ( 4822): environmentString
>>>> "HOME=/data/data/eu.licentia.necessitas.industrius.example.qmlshowcase
>>>> /file
>>>> s TMPDIR=/data/data/eu.licentia.necessitas.industrius.example.qmlshow
>>>> case/fi
>>>> les CACHE_PATH=/data/data/eu.licentia.necessitas.industrius.example.q
>>>> mlshowc ase/cache QML_IMPORT_PATH=/data/local/qt/imports
>>>
>>> QT_PLUGIN_PATH=/data/local/
>>>
>>>> qt/plugins"
>>>> ("HOME=/data/data/eu.licentia.necessitas.industrius.example.qmlshowcas
>>>> e/fil es",
>>>> "TMPDIR=/data/data/eu.licentia.necessitas.industrius.example.qmlshowca
>>>> se/fi les",
>>>> "CACHE_PATH=/data/data/eu.licentia.necessitas.industrius.example.qmlsh
>>>> owcas e/cache", "QML_IMPORT_PATH=/data/local/qt/imports",
>>>> "QT_PLUGIN_PATH=/data/local/qt/plugins") D/Qt ( 4822):
>>>> paramsString
>>>> "QtApp -platform android"
>>>> W/Qt ( 4822): WARNING: QApplication was not created in the main()
>>>> thread. D/Qt ( 4822): QAndroidPlatformIntegrationPlugin::create
>>>> "android" D/Qt ( 4822):
>>>> QAndroidPlatformIntegration::QAndroidPlatformIntegration(): creating
>>>> QAndroidEglFSScreen => Using OpenGL painting D/libEGL ( 4822): loaded
>>>> /system/lib/egl/libGLES_android.so
>>>> D/libEGL ( 4822): loaded /system/lib/egl/libEGL_adreno200.so
>>>> I/Qt JAVA ( 4822): surfaceChanged: 480,762
>>>> D/Qt ( 4822): setSurface 0x2a8908 0x2a8900
>>>> I/Qt JAVA ( 4822): surfaceChanged: 480,762
>>>> D/Qt ( 4822): setSurface 0x2a8908 0x2a8900
>>>> D/libEGL ( 4822): loaded /system/lib/egl/libGLESv1_CM_adreno200.so
>>>> D/libEGL ( 4822): loaded /system/lib/egl/libGLESv2_adreno200.so
>>>> I/ActivityManager( 96): Displayed
>>>> eu.licentia.necessitas.industrius.example.qmlshowcase/eu.licentia.nece
>>>> ssita s.industrius.QtActivity: +350ms W/Qt ( 4822): Opened
>>>> display 0x1 W/Qt ( 4822): Initialized display 1 4
>>>> W/Qt ( 4822): An error in Android-EGLFS-QPA-plugin occurred in
>>>> funtion void* q_configFromQPlatformWindowFormat(void*, const
>>>> QPlatformWindowFormat&), when calling eglChooseConfig. eglGetError
>>>> returned: 300c D/Qt ( 4822): ANativeWindow_settings width 480
>>>> height 762 format 1 D/Qt ( 4822):
>>>> QAndroidEglFSScreen::createWindowSurface 0x2a8908
>>>> W/IInputConnectionWrapper( 4148): showStatusIcon on inactive
>>>> InputConnection E/libEGL ( 4822): called unimplemented OpenGL ES API
>>>> E/libEGL ( 4822): called unimplemented OpenGL ES API
>>>> E/libEGL ( 4822): called unimplemented OpenGL ES API
>>>> E/libEGL ( 4822): called unimplemented OpenGL ES API
>>>> E/libEGL ( 4822): called unimplemented OpenGL ES API
>>>> E/libEGL ( 4822): called unimplemented OpenGL ES API
>>>> ...and so on...
>>>>
>>>> I'm running on a Nexus One with 2.3.4.
>>>>
>>>> I can post some or all of this on necessitas-devel if you want. Also,
>>>> what are your working hours GMT? I'm in the US, and I take it you
>>>> aren't, so it would help to know.
>>>>
>>>> Thanks for your help,
>>>> Tyler
>>>>
>>>> On Aug 15, 2011, at 4:39 AM, <thomas.senyk at nokia.com> wrote:
>>>>> No! This branch is deprecated.
>>>>> Use master!
>>>>>
>>>>> There you will see that there are software and opengl plugins.
>>>>>
>>>>> If you got everything build, I can help you get i running tomorrow
>>>>> (Today I'm out-of-office)
>>>>>
>>>>>
>>>>> Greets
>>>>> Thomas
>>>>>
>>>>> ________________________________________
>>>>> From: ext Tyler Mandry [tmandry at gmail.com]
>>>>> Sent: Monday, August 15, 2011 1:12 AM
>>>>> To: Senyk Thomas (Nokia-DXM/Munich)
>>>>> Subject: Compiling experimental-eglfs-jw
>>>>>
>>>>> Hi Thomas,
>>>>>
>>>>> I'm compiling your GL branch of Necessitas. I'm using the
>>>>> "experimental-eglfs-jw" branch, is this correct?
>>>>>
>>>>> Also, is GLES v1.x supported, or only 2.0?
>>>>>
>>>>> When I get this compiled I should be able to get it tested on a
>>>>> handful
>>>>> of different devices and get back to you.
>>>>>
>>>>> Thanks,
>>>>> Tyler Mandry
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/necessitas-devel/attachments/20110816/f1b8cb62/attachment-0001.html>
More information about the Necessitas-devel
mailing list