Hi Denys,

Yes, we also switched to the "thud" branch of u-boot and linux kernel from 
meta-ti. However, we did a check last week whether there are some updates in 
the thud branch we might be missing, and after that we did an update to the 
latest revisions of meta-ti, meta-arago, ... to see whether these new changes 
improve anything.

After the update we encountered a fresh error which we did not see before: when 
running any Qt GUI application, no graphics are shown and the following errors 
are in the console:

root@am335x-evm:~#  export QT_QPA_EGLFS_INTEGRATION=eglfs_kms
root@am335x-evm:~# /run/media/sda/qt_examples/analogclock
Failed to move cursor on screen LVDS1: -14 Could not queue DRM page flip on 
screen LVDS1 (Invalid argument)

or:
root@am335x-evm:~# export QT_QPA_EGLFS_INTEGRATION=none root@am335x-evm:~# 
/run/media/sda/qt_examples/analogclock
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-root'
error: XDG_RUNTIME_DIR not set in the environment.
error: XDG_RUNTIME_DIR not set in the environment.
error: XDG_RUNTIME_DIR not set in the environment.
Could not initialize egl display
Aborted (core dumped)

I also tried setting the /etc/qt5/eglfs_kms_cfg.json to:
{
  "device": "/dev/dri/card0",
  "hwcursor": false,
  "pbuffers": true,
  "outputs": [
    {
      "name": "LVDS1",
      "format": "xrgb8888",
      "primary": true
    }
  ]
}

We are obviously missing some important piece of the puzzle here. I tried 
running the glmark2 and it worked fine. This is how the environment looks like 
now:
QT_QPA_EGLFS_KMS_CONFIG=/etc/qt5/eglfs_kms_cfg.json
QT_QUICK_CONTROLS_1_STYLE=Flat
QT_QPA_ENABLE_TERMINAL_KEYBOARD=1
QT_QPA_EGLFS_ALWAYS_SET_MODE=1
QT_QPA_EGLFS_NO_LIBINPUT=1

I did some digging and I found out that there was a recent change in 
"ti-sgx-ddk-um" that replaces some plugins (libpvrDRMWSEGL.so, 
libpvrGBMWSEGL.so, libpvrws_WAYLAND.so) with their Mesa counterparts 
(libGLESv1_PVR_MESA.so, libGLESv2_PVR_MESA.so). Link to the change: 
http://git.yoctoproject.org/cgit/cgit.cgi/meta-ti/commit/?h=thud&id=05e41c29e2eab2337a0781ac137a1e40ae45b3e8
Could this be the reason why our Qt applications are not working?

Kind regards,
Ondrej

-----Original Message-----
From: [email protected] 
<[email protected]> On Behalf Of Denys Dmytriyenko
Sent: Thursday, 20 February 2020 12:52
To: Ankur Tyagi <[email protected]>
Cc: [email protected]
Subject: Re: [meta-arago] [Thud branch][AM3354]OpenGL/QT font rendering problem 
- corrupt glyphs

That is quite strange. We are wrapping up our Thud support. But it was used in 
our products for about a year with rigorous testing and we haven't seen 
anything like that on any of our platforms.

This does look like some memory corruption - are you sure you have all the 
timings correctly? Have you also switched any of the low-level components, like 
U-boot, Linux kernel, etc.?

Denys


On Wed, Feb 19, 2020 at 07:04:12PM +0000, Ankur Tyagi wrote:
> Hi,
> 
> Is there anything can be done to fix it? Let me know if some more information 
> needs to be shared in order to help with diagnosis.
> 
> Regards
> Ankur
> 
> 
> ________________________________
> From: [email protected] 
> <[email protected]> on behalf of Ankur Tyagi 
> <[email protected]>
> Sent: Tuesday, February 18, 2020, 11:11 AM
> To: [email protected]
> Subject: [meta-arago] [Thud branch][AM3354]OpenGL/QT font rendering 
> problem - corrupt glyphs
> 
> Resending it as previous one didn’t made it to the mailing list
> 
> And we are stuck in production after upgrade to thud branch because of 
> this issue, any help is appreciated
> 
> Regards
> Ankur
> 
> From: Ankur Tyagi
> Sent: Tuesday, 11 February 2020 12:39 PM
> To: [email protected]
> Subject: [Thud branch][AM3354]OpenGL/QT font rendering problem - 
> corrupt glyphs
> 
> Greetings,
> 
> I am using AM3354 processor. I just did a major Yocto upgrade from "daisy" to 
> "thud" branch, and I noticed that some characters/glyphs happen to be 
> slightly corrupt.
> The problem occurs randomly, but when a corrupted character is rendered 
> multiple times on the screen, all of its instances are corrupted the same way.
> 
> It looks like this: https://ibb.co/LYP4JCx
> 
> I tried using various QT environment variables, but the only one that worked 
> was QT_QUICK_BACKEND=software. However, this led to a very significant 
> slowdown, so it's not an option.
> 
> I'm not using X11 or wayland. I'm not modifying the graphics drivers or the 
> QT layer in any way.
> 
> Following are Qt environment variables on target:
> 
> export QT_QPA_EGLFS_INTEGRATION=none
> export QT_QUICK_CONTROLS_1_STYLE=Flat
> export QT_QPA_EGLFS_NO_LIBINPUT=1
> 
> Has anyone experienced similar problem? Is there any configuration detail I 
> might be missing?
> 
> Regards,
> Ankur
> 
> ________________________________
> This email is confidential and may contain information subject to legal 
> privilege. If you are not the intended recipient please advise us of our 
> error by return e-mail then delete this email and any attached files. You may 
> not copy, disclose or use the contents in any way. The views expressed in 
> this email may not be those of Gallagher Group Ltd or subsidiary companies 
> thereof.
> ________________________________
> 

> _______________________________________________
> meta-arago mailing list
> [email protected]
> http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago

_______________________________________________
meta-arago mailing list
[email protected]
http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago
_______________________________________________
meta-arago mailing list
[email protected]
http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago

Reply via email to