1035531 - Failed CHECK in GonkCameraSource::dataCallbackTimestamp() while stability testing

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::Camera, defect)

Description

[Greg Grisco]

Attachment: decoded minidump

STR are unclear at this point, but we found a this crash during stability test which seems to come from a failed CHECK in GonkCameraSource.

Uploading minidump and EXTRA files (with logs).

Comment 1

[Greg Grisco]

Attachment: EXTRA file attachment (with logs)

Comment 2

[Diego Marcos [:dmarcos]]

What device? What build?

Comment 3

[cafbot (PoC: ggrisco)]

Observed on: 

Device: msm8610
Gonk Version: AU_LINUX_GECKO_B2G_KK_3.6.01.04.00.000.023
Moz BuildID: 20140630000201
B2G Version: 2.0
Gecko Version: 32.0a2
Gaia:  http://git.mozilla.org/?p=releases/gaia.git;a=commit;h=c0c8ad187c0466285f2580531e09f8322996f561
Gecko: http://git.mozilla.org/?p=releases/gecko.git;a=commit;h=ef0781044f85ee7f364236c074ab0b7644c38ff8

Comment 4

[Greg Grisco]

(In reply to Diego Marcos [:dmarcos] from comment #2)
> What device? What build?

Answered by cafbot in Comment 3.

Comment 5

[Diego Marcos [:dmarcos]]

Sorry but I don't know what an msm8610 is. What's the code name?

Comment 6

[Mike Habicher [:mikeh] (high bugzilla latency)]

It's a QRD, but this bug appears on other platforms: see bug 1018129.

Andrew and I discussed this in IRC and we think the solution is to remove the CHECK() that's causing the abort and just discard frames with suspicious timestamps.

Comment 7

[Hema Koka [:hema]]

(In reply to Mike Habicher [:mikeh] from comment #6)
> It's a QRD, but this bug appears on other platforms: see bug 1018129.
> 
> Andrew and I discussed this in IRC and we think the solution is to remove
> the CHECK() that's causing the abort and just discard frames with suspicious
> timestamps.

Thanks Mike. 

Assigning this to Andrew who is investigating.

Comment 8

[Andrew Osmond [:aosmond] (he/him)]

Attachment: bug1035531.patch, v1

Resolve the issue by dropping frames with a bad timestamp while logging an error instead of asserting. Greg, can you give it a whirl to see if you can reproduce when it would have asserted and check if the resulting video is acceptable (i.e. frames do not keep dropping)? I have yet to see this issue in my own testing.

Comment 9

[Andrew Osmond [:aosmond] (he/him)]

Attachment: bug1035531.patch, v1.1

Fix nits that are bound to be brought up in review :). Cancelling feedback after re-reading the description; I doubt we will be able to reproduce this in a satisfactory timeframe. It is best to get the crash resolved in the field and keep an eye out for the new error log in the future for situations where the final recording quality suffers.

try: https://tbpl.mozilla.org/?tree=Try&rev=b763366f1805

Comment 10

[Mike Habicher [:mikeh] (high bugzilla latency)]

Comment on attachment 8453053 [details] [diff] [review]
bug1035531.patch, v1.1

Review of attachment 8453053 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks for taking care of this, Andrew.

I'd really like to see some feedback from CAF on it.

Comment 11

[Greg Grisco]

(In reply to Andrew Osmond [:aosmond] from comment #8)
> Created attachment 8452564 [details] [diff] [review]
> bug1035531.patch, v1
> 
> Resolve the issue by dropping frames with a bad timestamp while logging an
> error instead of asserting. Greg, can you give it a whirl to see if you can
> reproduce when it would have asserted and check if the resulting video is
> acceptable (i.e. frames do not keep dropping)? I have yet to see this issue
> in my own testing.

Ok, we'll give it a try, thanks.

Comment 12

[Greg Grisco]

We haven't seen this crash since applying the patch, thanks.

Comment 13

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/integration/b2g-inbound/rev/904c9b948214

Comment 14

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/904c9b948214

Comment 15

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/e52bc38eac66

Comment 16

[[:RJ]]

No STR is present to create test case to address bug.