985681 - Camcorder crash when recording video for more than 5 minutes

Status: RESOLVED DUPLICATE of bug 988713

(Core :: Graphics: Layers, defect)

Description

[Greg Grisco]

Attachment: decoded minidump of crash

1.open Camera app and switch to camcorder
2.Start recording
3.Try to record the video for more than 5 minutes

Crash occurs with crash signature:

[@ jemalloc_crash | arena_dalloc | mozilla::layers::TextureChild::~TextureChild() | mozilla::detail::RefCounted<mozilla::layers::TextureChild, (mozilla::detail::RefCountAtomicity)0>::Release() const ]

Have seen this at least 5 times

Comment 1

[Greg Grisco]

Attachment: EXTRA file attachment

Comment 2

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

Sotaro, any idea what might be going on here?

Comment 3

[Sotaro Ikeda [:sotaro]]

attachment 8393761 [details] seems like double delete, but TextureChild is already thread safe ref couneted. Hmm.

Comment 4

[Sotaro Ikeda [:sotaro]]

(In reply to Greg Grisco from comment #0)
> [@ jemalloc_crash | arena_dalloc |
> mozilla::layers::TextureChild::~TextureChild() |
> mozilla::detail::RefCounted<mozilla::layers::TextureChild,
> (mozilla::detail::RefCountAtomicity)0>::Release() const ]
> 
> Have seen this at least 5 times

Greg, how often does it happen? Which device did you test?

Comment 5

[Greg Grisco]

(In reply to Sotaro Ikeda [:sotaro] from comment #4)
> Greg, how often does it happen? Which device did you test?

We have seen it exactly 5 times all on the same build.  This was on Madai device.

Comment 6

[Greg Grisco]

Sorry, I used the wrong terminology.  This is on QRD8x26.

Comment 7

[Sotaro Ikeda [:sotaro]]

Bug 985878 might causing the problem.

Comment 8

[Sotaro Ikeda [:sotaro]]

Greg, can you test again by applying attachment 8394044 [details] [diff] [review] in Bug 985878?

Comment 9

[Sotaro Ikeda [:sotaro]]

(In reply to Greg Grisco from comment #5)
> (In reply to Sotaro Ikeda [:sotaro] from comment #4)
> > Greg, how often does it happen? Which device did you test?
> 
> We have seen it exactly 5 times all on the same build.  This was on Madai
> device.

Sorry, I just want to know the frequency(how often), not the number of the crash. Can you tell the frequency of the crash (like 5/20)?

Comment 10

[Greg Grisco]

The tester who reported says every time he is able to reproduce.  I was able to record from 5-10 minutes without seeing the crash, so I am asking for re-test right now.  I'll report the findings when I hear back.

Comment 11

[Greg Grisco]

We couldn't reproduce it, so closing for now.  Will reopen if it happens again.

Comment 12

[Greg Grisco]

We are seeing this multiple times now, latest using:

gaia sha1:  7e705dd4718d528974d99ac31866318d7e201152
gecko sha1: 81826931597f4b330cf2d6c40f444bcce000a6b8

Comment 13

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

Greg, is there some way to narrow the STR? Recording settings, etc? I've got a Nexus 4 here happily into its 10th minute of video recording with no signs of slowing down.

- gecko: b2g-inbound:a4ac15d6912d
- gaia: master:c8e974e69c764b953263c7474d10c25aa68bb862

AFAIK, the camera still isn't working on our QRD devices, so the CAF team may have to take the lead on tracking this one down.

Comment 14

[Preeti Raghunath(:Preeti)]

Upon talking to QC the expectation is that we work off of the mini dump. Is that possible for you to check?

Comment 15

[Sotaro Ikeda [:sotaro]]

(In reply to Sotaro Ikeda [:sotaro] from comment #8)
> Greg, can you test again by applying attachment 8394044 [details] [diff] [review]
> [review] in Bug 985878?

Bug 988713 might also related.

Comment 16

[bhargavg1]

(In reply to Mike Habicher [:mikeh] from comment #13)
> Greg, is there some way to narrow the STR? Recording settings, etc? I've got
> a Nexus 4 here happily into its 10th minute of video recording with no signs
> of slowing down.
> 
> - gecko: b2g-inbound:a4ac15d6912d
> - gaia: master:c8e974e69c764b953263c7474d10c25aa68bb862
> 
> AFAIK, the camera still isn't working on our QRD devices, so the CAF team
> may have to take the lead on tracking this one down.

the tests that we are running are,

Run gaia-ui test script with camera, camcorder, Music, Video and FM test cases sequentially

Comment 17

[Inder]

> AFAIK, the camera still isn't working on our QRD devices, so the CAF team
> may have to take the lead on tracking this one down.

@mikeh: Camera issue is fixed on the QRD build, hopefully that will help expedite the investigation here.

Comment 18

[Milan Sreckovic [:milan] (needinfo for best results)]

Sotaro, anything actionable for you on this one, or are you blocked on something or somebody else?

Comment 19

[Sotaro Ikeda [:sotaro]]

I am checking if QRD device works. Right now, camera preview works, but failed to start recording. I am investigating about it.

Comment 20

[Sotaro Ikeda [:sotaro]]

(In reply to Milan Sreckovic [:milan] from comment #18)
> Sotaro, anything actionable for you on this one, or are you blocked on
> something or somebody else?

Camera becomes partially works on QRD device, but the recording still does not work. I am asking about it to codeaurora.

Comment 21

[Sotaro Ikeda [:sotaro]]

Bug 988713 is not uplifted to b2g v1.4 yet.

Comment 22

[Sotaro Ikeda [:sotaro]]

Greg, can you confirm if attachment 8399258 [details] [diff] [review] in Bug 988713 could fix the problem?

Comment 23

[Inder]

Tapas, can you please check this please.

Comment 24

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

(In reply to Inder from comment #23)
> Tapas, can you please check this please.

I didn't see it in my manual testing in tip of v1.4. But this issue happens randomly. So, I asked our internal test team to confirm this . I will update here once they confirm me.

Comment 25

[Sotaro Ikeda [:sotaro]]

On QRD8x26 device, camera recording still does not work. I can not test the problem on my side.

Comment 26

[Sotaro Ikeda [:sotaro]]

From attachment 8393762 [details], TextureChild::~TextureChild() causes the SIGSEGV by accessing 0x0 address, therefore TextureChild's double delete seems happen. But TextureChild is thread safe reference counted.

> Crash reason:  SIGSEGV
> Crash address: 0x0

Comment 27

[Sotaro Ikeda [:sotaro]]

AtomicRefCounted<>'s thread safety problem was fixed by Bug 985878 on 2014-03-26 on mozilla hg.
 Race condition when destroying TextureClient & TextureChild(Bug 988713) was uplifted to b2g v1.4 on 2014-04-09 on mozilla hg.

Comment 28

[Sotaro Ikeda [:sotaro]]

(In reply to Tapas Kumar Kundu from comment #24)
> (In reply to Inder from comment #23)
> > Tapas, can you please check this please.
> 
> I didn't see it in my manual testing in tip of v1.4. But this issue happens
> randomly. So, I asked our internal test team to confirm this . I will update
> here once they confirm me.

If the crash becomes very rare, it might be Bug 994903.

Comment 29

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

It didn't come in latest stability run. But I will wait to see results for another test run as it happens randomly. I will confirm here once I get results from another run.

Comment 30

[Sotaro Ikeda [:sotaro]]

Tapas, if there is a crash, please attach the new crash log. Thanks.

Comment 31

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

(In reply to Sotaro Ikeda [:sotaro] from comment #30)
> Tapas, if there is a crash, please attach the new crash log. Thanks.

We are not seeing it with latest gaia/gecko. It seems like it is fixed.