Logs and Video: https://cloud.google.com/console/storage/chromiumos-test-logs/bugfiles/cr/867903/

Bitrate isn't the problem; the encoded bitrate is reasonably close to the target.

The issue here is clearly video corruption, and one caused by the hw codec based on that fact that it is resolved by setting "WebRTC Hardware Video Encoding" to false.

As far as I can see, Blaze is based on NVIDIA Tegra K1, so would this be a v4l2 encoder?

posciak@ can you triage?

This issue is not reproducible in other devices as of now. Video corruption is not observed in other web applications [hangouts, appear.in etc] in Blaze. Thus this issue is very specific to Blaze device and meet. 

* Updated the title as Video corruption. Thank you @sprang. 

I cannot reproduce this issue with nyan_big even in the same Chrome OS version.
Could you test with nyan_big?

@hiroh: Thank you! I couldn't reproduce in any of the skylab devices we have till now [~27]. nyan_big is one among them. Thus I feel it is very likely device specific. As of now reproducible only in  Blaze. 

I can update here soon if chromebox or chromebit have this issue. 

vasanthakumar@, is there an easy way to do a rough bisect to find a range where this issue started happening? Is it still happening in latest canary?

vasanthakumar@, thank you for informing me that!
I think there is almost no difference between nyan_big and nyan_blaze.
By "device specific", do you mean a device you tested, or blaze?
Do you test other blaze?

I have tested in a blaze device we have. We do not have any other blaze with us at the moment. Do you have one to test? if not I can order a new one. 

I meant the issue is limited to blaze device only and not in other devices. Thanks!

Sorry for delay to reply.
I have one blaze here.
It has not been used for a long time and the image is M36.
In addition, it is not in developer mode. So I am trying to burn the latest Chrome OS image into it.
I think I can test on the blaze tomorrow.
But to gather more data, I think it is better you get another blaze.

Thanks

I tested this morning with nyan_blaze here. This issue is not reproduced.
Could you request a new nyan_blaze device?
Additionally, if you have time, could you bisect?

I have tried asking people here about Blaze device. Unfortunately, I couldn't get any device as of yet. Thus I will  order a new one. 

Sure! I will bisect and update here soon. 

@Hiroh: I couldn't reproduce the issue from the builds after 69.0.3494/0 / 10888.0.0. Seems like the issue is fixed in later versions. 

No reproducible:
69.0.3497.14 / 10895.5.0	
69.0.3497.21 / 10895.10.0	

Some of the commits later could have fixed this issue. Thus I feel we can close this issue for now. 

@Hiroh: Share your thoughts on it. Which build did you test?

I remember I tested the same version of your reported, i.e., 10888.0.0.
Could you bisect more? There are too manychanges between 10895.5.0 and 10888.0.0 to find a culprit.

Thanks

@hiroh: I  have bisected the issue. From this build [10894.0.0 / 69.0.3496.0], I couldn't reproduce the issue at all. Thus the fix should have landed from this point. 

Here comes the details. 

10895.0.0 / 69.0.3497.0 Not reproducible
10894.0.0 / 69.0.3496.0 Not reproducible. [First build where we cannot reproduce this issue] reproducible]
10893.0.0 / 69.0.3494.0 reproducible
10891.0.0 / 69.0.3494.0 reproducible.

Thanks for bisecting.
Looks like one of Chrome changes fix this problem, although I couldn't say which one without examining w/wo the CL in actual device.
This issue no longer happens in beta channel. So I think there is no need to investigate more.
Please feel free to close this.

Thanks

Sure! I will close this. Thank you!

I have reproduced this issue somehow today. This issue is not that frequent when compared to the time it was reported earlier. But it is definitely reproducible. We can observe the video corruption every 5 seconds but now it takes upto 20-30 seconds. 

Sometimes even after there is a major changes right before the camera. 

Reproduced when I used mac and linux devices in the call.

https://buganizer.corp.google.com/issues/112410268

Meet feedback is sent. 

I cannot see the feedback at b/112410268.
But you tested at M68?

Now the new steps to reproduce are as follows:

1. Perform a meet call between 3 devices [I used blaze, macbook pro, linux z440 machine]
2. Leave the call for long time more than half hour [max 45 minus]
3. Video from blaze gets corrupted. This was reproduced several times. When this happens it continues to happen until we end the call. 

This issue sometimes reproducible with 2 users also. this is not 100%. 

https://listnr.corp.google.com/product/5007887/report/85594335035?dateRange=All
https://listnr.corp.google.com/product/5007887/report/85594335528?dateRange=All
https://listnr.corp.google.com/product/5007887/report/85592818624?dateRange=All

To get access for Listnr then go to: https://g3doc.corp.google.com/company/teams/feedback/listnr/faq.md?cl=head

I guess this should be assigned to hiroh now?

hiroh, could you please update this bug. It's a P1 and needs attention. If you don't think it's important enough to be P1 we should reduce the priority.

I don't this this issue is P1. I will take a look when I am free. I am focusing on other bugs these days.