Re: Chromium layout rendering issue

[Gary Thomas <gary@mlbassoc.com>]

On 2015-05-08 16:52, Eric Nelson wrote:
> Hi Gary,
>
> On 05/08/2015 03:47 PM, Gary Thomas wrote:
>> On 2015-05-08 06:49, Otavio Salvador wrote:
>>> On Thu, Apr 30, 2015 at 11:41 PM, Nikolay Dimitrov <picmaster@mail.bg>
>>> wrote:
>>> ...
>>>> So I have 2 questions:
>>>> 1. The probability that someone knows what's the actual issue is close
>>>> to minus infinity, but at least did someone observed similar behavior?
>>>> If needed, I can snap photos to visually demonstrate the issue.
>>>
>>> I have customers using it in Dizzy and I didn't see the error. Please
>>> check today's meta-browser as I merged a fix from Khem which address
>>> some installation issues.
>>>
>>>> 2. Chromium on sabresd & fido gave me at least 1 interesting warning
>>>> about GPU being frozen and pending reset in 10s. I haven't seen such
>>>> error, but this makes me highly concerned about porting my stuff to
>>>> fido. Did someone see such error messages?
>>>
>>> Prabhu?
>>>
>>
>> I just rebuilt Chrome with the latest meta-browser (and meta-fsl-arm-*)
>> There are many rendering problems on the Google main page and then I
>> get these errors:
>>
>> [872:872:0508/165904:ERROR:image_metadata_extractor.cc(111)] Couldn't
>> load libexif.
>> [896:896:0508/165907:ERROR:sandbox_linux.cc(301)] InitializeSandbox()
>> called with multiple threads in process gpu-process
>> Fontconfig warning: FcPattern object size does not accept value 9(i)
>> [958:958:0508/165913:ERROR:image_metadata_extractor.cc(111)] Couldn't
>> load libexif.
>> [870:890:0508/165930:ERROR:channel.cc(258)] RawChannel read error
>> (connection broken)
>> [896:906:0508/165943:ERROR:gpu_watchdog_thread.cc(253)] The GPU process
>> hung. Terminating after 10000 ms.
>> [870:870:0508/165943:ERROR:gpu_channel_host.cc(185)] Failed to send
>> GpuChannelMsg_CreateOffscreenCommandBuffer.
>> [870:870:0508/165943:ERROR:webgraphicscontext3d_command_buffer_impl.cc(230)]
>> Failed to initialize command buffer.
>> [1005:1006:0508/165955:ERROR:gpu_watchdog_thread.cc(253)] The GPU
>> process hung. Terminating after 10000 ms.
>> getrlimit(RLIMIT_NOFILE) failed
>> getrlimit(RLIMIT_NOFILE) failed
>>
>> This is on a board using the Boundary Devices 3.10.53 kernel
>> (I don't have a board here that runs a "pure" FSLC kernel)
>> but as far as I can tell the BD kernel is the FSLC kernel
>> with some of their own patches, mostly to do with cameras.
>>
>
> Was this on master or fido?
>
>> Anything I can look for/test?
>>
>> Also note that this kernel has lots of [rendering] problems
>> with other applications.  For example, if I run the new Gstreamer-1.0
>> based "media player" to play a video, it will bring up the window but
>> not render any video frames or sometimes render a few and then freeze.
>>
> Can you try with the pre-built image here?
> 	http://boundarydevices.com/i-mx6-kernel-3-10-53/
>
> We didn't see any issues, but mostly ran things from the command-line
> and the pre-built Vivante demos to test graphics.

Just tried this on my SabreLite.

Chrome - fails just like on my board as reported above
media player (gst-play-1.0) - thinks it's running, but no video

Also, I don't have any audio, neither analog or HDMI

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------