ERROR: Failed to create KHR image: 12300

[Djoulez]

Hi,

In this simple example, I create/destroy the mediaplayer dynamically. After some times, I always got this error:

09:14:08.076 INFORMATION: Instantiating texture data...
09:14:08.096 INFORMATION: Creating EGLImageKHR...
09:14:08.096 INFORMATION: Instantiating texture data...
09:14:08.532 INFORMATION: Creating EGLImageKHR...
09:14:08.533 INFORMATION: Instantiating texture data...
09:14:08.553 INFORMATION: Creating EGLImageKHR...
09:14:08.553 INFORMATION: Instantiating texture data...
09:14:08.554 INFORMATION: Creating EGLImageKHR...
eglCreateImageKHR: failed to create image for buffer 0x39 target 12465 error 0x300c
09:14:08.556 ERROR: Failed to create KHR image: 12300.

This example is launched with qmlscene.

Do you have any idea about this ?

mediaPlayerTest.zip

[sinadogru]

It is also happening to me, I was just going to declare an issue about it.

[floouuu]

I think this has something to do wih running out of video ram.
To see how much video memory is allocated right now (in bytes) you can run sudo vcdbg reloc |grep "\[" | awk '{split($0,a," "); print a[12]}' | awk '{split($0,a,","); print a[1]}' | awk '{sum+=$1}END{print sum}'
I experienced similar behavior when using the Video component. I wrote a simple video player, that plays urls from a xml file. However, after some time, when gpu memory gets full, I have the same error and the program crashes.
I Have a Loader whichs loads a QML file where a Video component is included. After each video the memory use of the gpu reduces again, but it is 30 MB short. Which means that with each video that I load there are 30 MB more of gpu memory lost. So I think there has to be a memory leak somewhere.

[floouuu]

I took a short look, but I guess too short, but I noticed that in openmaxilplayerservice.cpp contructor many variables are allocated, but I didn't find where they a freed again. It's just a shot into the blue.

[Nightmare204]

@floouuu Does the same happens if you reuse the QML Component without reloading it in the Loader?

I have tested long term loop playback using a playlist and I've found this kind of problem is really related to dynamic creation of new video component with a Loader

[carlonluca]

The fix will be available shortly.

[floouuu]

@Nightmare204 Thank you for the hint. I am abroad now, but I will test this as soon as I am home again - which will be arround tomorrow evening. I will let you know!
@carlonluca Thank you!

[carlonluca]

Someone who can test the new build?

[Nightmare204]

Hi @carlonluca... I'm doing some long test on my code and I'm using POT 5.1.0
Are the fixes for POT 5.2.0 commited already?
As far as I see in github Latest commit bf0e76b on 13 Dec 2015

If you provide the sources I will be able to recompile for qt5.5.1 and give it some long test.

I'm guessing my problem is around this issue because my video loop just freezes and sometimes even gives me kernel panic problems

[carlonluca]

Sources for 5.2.0 will be committed in a few hours.

[carlonluca]

Commit was pushed.

[Nightmare204]

I tested again the 5.6 build and I got some weird behavior:

After several loops (More that 8 hours) between 2 videos of 15 sec the player just stopped with segmentation fault.
If I restart the same player program It fails immediately after starting any video.

stdin: is not a tty
QML debugging is enabled. Only use this in a safe environment.
Program Starts
Init - Creating GUI instance
PlayerStandAlone::main::Creating Instance
Init - Executing App
Setting Source to qrc:/qml/VideoPlayer.qml
16:16:19.916 INFO:   POT build Feb  6 2016 18:15:41.
16:16:19.918 INFO:   Initializing GPU context in media processor...
16:16:19.919 INFO:   Initializing buffer provider...
Requesting control for org.qt-project.qt.metadatareadercontrol/5.0...
Requesting control for org.qt-project.qt.mediaavailabilitycontrol/5.0...
Requesting control for org.qt-project.qt.mediaplayercontrol/5.0...
Requesting control for org.qt-project.qt.medianetworkaccesscontrol/5.0...
Requesting control for org.qt-project.qt.audiorolecontrol/5.6...
Requesting control for org.qt-project.qt.videorenderercontrol/5.0...
Calling viewer->showFullScreen()
viewer->status() == QQuickView::Ready
qml: resType =1, localPath =/home/pi/Video/26_5_1080p_bbb_cut.mp4
qml: check isAlreadyPlaying() = false
qml: play function called
qml: load function called
qml: start function called
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from '/home/pi/Video/26_5_1080p_bbb_cut.mov':
  Metadata:
    major_brand     : qt  
    minor_version   : 512
    compatible_brands: qt  
    encoder         : Lavf56.25.101
  Duration: 00:00:15.04, start: 0.000000, bitrate: 15263 kb/s
    Stream #0:0(eng): Video: h264 (Main) (avc1 / 0x31637661), yuv420p(tv, bt709), 1920x1080, 14818 kb/s, 24 fps, 24 tbr, 19200 tbn, 38400 tbc (default)
    Metadata:
      handler_name    : DataHandler
      encoder         : H.264
      timecode        : 00:00:00:00
    Stream #0:1(eng): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz, 5.1, fltp, 439 kb/s (default)
    Metadata:
      handler_name    : DataHandler
    Stream #0:2(eng): Data: none (tmcd / 0x64636D74), 0 kb/s
    Metadata:
      handler_name    : DataHandler
      timecode        : 00:00:00:00
16:16:21.366 INFO:   Instantiating texture data...
16:16:21.371 INFO:   Instantiating texture data...
16:16:21.377 INFO:   Instantiating texture data...
16:16:21.382 INFO:   Instantiating texture data...
bash: line 1:  8943 Segmentation fault      DISPLAY=':0.0' /home/pi/PlayerStandAlone/PlayerStandAlone

However if I play the video with omxplayer it reproduces normally.
I noticed that one of the videos seem to finish in a bad frame, because it generates garbage...

In Other test this video just never ended but rendered black....

IDK if it could be just some error in the video format... @carlonluca If you can check this video, maybe you find something on it.

I'd reconverted the video and I'm starting tests again... I will tell you

[carlonluca]

Can you reproduce with a minimal QML looping the video? Can you run on gdb?

[carlonluca]

Closing the issue untile someone provides a sample to reproduce. Feel free to reopen.

[floouuu]

Hi,
I'm sorry for the very late response - work, you know :/
@Nightmare204 I figured out that this only happened when the Video was loaded through a Loader component. When using the Video Component directly, I couldn't discover any issues with the 5.6 image from the post of March 12th 2016. @carlonluca Video ram also seems to remain stable with your fixes.