Workspace/Omap4/Pandaboard/History

Revision as of 10:01, 2 April 2013

Solved issues

• OpenGL app crash with SGX: launching an app works first time, but all subsequent attempts fail. Same cause: https://bugs.launchpad.net/linaro-ubuntu/+bug/1034178

* Fixed by updating to pvr-omap4-1.9.0.7.1.5

• Full Nemo installation, after lock screen comes up, fails with:

Xorg invoked oom-killer: gfp_mask=0x44d0, order=2, oom_adj=0, oom_score_adj=0
full log: http://pastie.org/5668605

Less important:

• Moving mouse too early in in Xorg's boot, renders it unmovable - reconnect the cable
• 15/1/2013 FIXED: When mouse cursor is shown, any animation is flickering when mouse is moving
• 11/1/2013 WORKAROUNDED: Impossible to launch qmlcalc twice (or flickr-qml thrice), this what's spat (and qmlcalc dies):

[  121.403869] omapdrm omapdrm.0: already a pending flip

• 11/1/2013 FIXED: Currently, launching any OpenGL app (tested with flickr-qml, qmlcalc and duicontrolpanel) from within lipstick kills Xorg while applauncherd is boosting it for Qt+GLES. Fix: http://www.hastebin.com/notadehohi.coffee

Progress log (newest first)

• 28/3/2013 Pandaboard ES. boot hangs near completion(?); 2 penguins in upper left corner of display, led D1 flashing (2 quick flashes)

U-Boot SPL 2012.04.01 (Mar 22 2013 - 15:24:27) 
 ... 
CPU  : OMAP4460 ES1.1 
 ... 
Image Name:   Linux-3.4.27-1.2-pandaboard 
 ...
[    0.000000] Booting Linux on physical CPU 0
 ...
[    4.632965] generic-usb 0003:045E:007D.0002: input,hidraw1: USB HID v1.00 Mo0

• 15/1/2013 (sledge) OpenGL app crash with SGX: Jake9xx gave a tmp workaround http://pastie.org/5669270 , but it does not work
• 11/1/2013 (sledge) OpenGL app crash with SGX: looking into qmlcalc launch issues. flickr-qml also spits same old

[   54.713897] flickr-qml (214): /proc/214/oom_adj is deprecated, please use /proc/214/oom_score_adj instead.

and now also

[   57.826049] omapdrm omapdrm.0: already a pending flip

but does not crash. Xorg does not crash either (thanks to the previous fix)

qmlcalc also spits oom and "already a pending flip" however crashes (after first successful launch though!), more clues from the app booster stage:

Jan 11 16:56:28 localhost lipstick[201]: Found EGL_KHR_image_pixmap & GL_OES_EGL_image extensions (preferred method)!
Jan 11 16:56:28 localhost lipstick[201]: virtual WindowInfo::~WindowInfo() Destroyed windwo for  6291601
Jan 11 16:56:29 localhost mcompositor[200]: PVR:(Error): ReleaseImageFromTexture: Unsupported pixel format [207, /eglimage.c] (x3 times repeated)
Jan 11 16:56:29 localhost applauncherd[199]: Boosted process (pid=212) was terminated due to signal 11
Jan 11 16:56:29 localhost invoker[264]: error: Can't send signal 11 to application [212]: No such process

• 11/1/2013 (sledge) TODO: what causes out of memory? Reduce full Nemo installation package-by-package, irrelevant (GSM-related etc) ones first
  • Apparently, having on top of minimal-lipstick installed meegotouch-systemui (with its deps), it's triggering OOM (itself or one of its deps: connman-qt contextkit-plugin-cellular contextkit-plugin-connman libiodata libiphb libqmlog qmsystem timed tzdata-timed)

Installation

Full Nemo installation is not working (if you choose @Nemo Complete pattern). Most often while being booted it ends up running out of memory: Xorg rapidly increasing in memory usage, reaching 800MB in ~5seconds, and everything dies.