782223 - crash in NvGl214081946277 on Asus Nexus 7 (NVIDIA Tegra 3) running JB

Status: RESOLVED WONTFIX

(Firefox for Android Graveyard :: Plugins, defect)

Description

[Scoobidiver (away)]

This bug was filed from the Socorro interface and is 
report bp-e87d9372-6aca-4607-b524-f58be2120811 .
============================================================= 

It's #31 top crasher in 15.0b4, #24 in 16.0a2, and #92 in 17.0a1.
Oddly, 14.0.2 that is JB only is not affected.

More reports at:
https://crash-stats.mozilla.com/report/list?signature=libGLESv2_tegra.so%400x4ae10

Comment 1

[Scoobidiver (away)]

With the release of Android 4.2, it's #6 top crasher in 17.0b7.

Comment 2

[Scoobidiver (away)]

(In reply to Scoobidiver from comment #1)
> With the release of Android 4.2, it's #6 top crasher in 17.0b7.
Android 4.2 users were over-represented in 17.0b7. It's no longer a top crasher in 17.0.

Comment 3

[Robert Kaiser]

I still want us to consider tracking those JB and Tegra3 crashes and trying to figure out what we can do there as I think the Tegra3 devices will grow in volume - probably even soon, as many of those could land under the tree at Christmas.

Comment 4

[Alex Keybl [:akeybl]]

(In reply to Robert Kaiser (:[e-mail]) from comment #3)
> I still want us to consider tracking those JB and Tegra3 crashes and trying
> to figure out what we can do there as I think the Tegra3 devices will grow
> in volume - probably even soon, as many of those could land under the tree
> at Christmas.

Agreed, but before we track, where does this fall in the 4.2 top crash list?

Comment 5

[Alex Keybl [:akeybl]]

We don't yet have a 4.2-specific top crash list, but given comment 1, it's obviously up there.

Sending over to Mark to re-assign.

Comment 6

[Alex Keybl [:akeybl]]

(KaiRo mentions that this is related to Flash on JB, given the stacks found on Socorro)

Comment 7

[Alex Keybl [:akeybl]]

QA - please look at instructions like http://liliputing.com/2012/07/how-to-install-adobe-flash-on-the-nexus-7-or-other-jelly-bean-devices.html to test Flash on JB

Comment 8

[Robert Kaiser]

We only have 6 URLs for this signature, two of which are adult videos which I won't cite publicly (if you really need them for testing, please ping us privately), the four others are:

1 	https://apps.facebook.com/d-slots/?ref1=profile&ref=wall_BCB_112012
1 	http://www.nexttv.com.tw/
1 	http://www.itv.com/itvplayer/video/?Filter=327699
1 	http://www.amazon.com/Sony-DSC-RX100-Sensor-Digital-Camera/dp/B00889ST2G/ref=sr_1_3?s=photo&ie=UTF8&qid=1353347431&sr=1-3&keywords=Canon+eos+m

Comment 9

[Alex Keybl [:akeybl]]

Kevin/Aaron - who's planning on filling this qawanted request?

Comment 10

[Aaron Train [:aaronmt]]

Visiting those URL's above on my Nexus 7 (Android 4.2), and Galaxy Nexus (Android 4.2.1) w/Adobe Player 11.1 for Android 2.x and 3.x (11.1.111.24), I was unsuccessful in being able to reproduce any crash.

Comment 11

[Alex Keybl [:akeybl]]

Mark - do you think blind engineering investigation could yield anything here?

Comment 12

[Scoobidiver (away)]

It's #4 top crasher in 18.0b3 but Android 4.2 users are overrepresented.

Comment 13

[Robert Kaiser]

This crash continues to be #2 on 18 beta.

Mark, is there any way we can get someone to investigate what is going on there?
Note that there are a lot of libflashplayer frames in those stacks.

Comment 14

[Mark Finkle (:mfinkle) (use needinfo?)]

There is nothing but guess work to be done. I can't assign that to someone. We'd have more of a chance of seeing this crash if more devs and QA were using the device. I'll ask about that possibility. I notice this only seems to occur when we hit beta. I don't see any fx20 or fx21 crashes. Our nightly and aurora audiences are small, but I would have figured seeing at least one crash.

Comment 15

[Alex Keybl [:akeybl]]

(In reply to Mark Finkle (:mfinkle) from comment #14)
> There is nothing but guess work to be done. I can't assign that to someone.
> We'd have more of a chance of seeing this crash if more devs and QA were
> using the device. I'll ask about that possibility. I notice this only seems
> to occur when we hit beta. I don't see any fx20 or fx21 crashes. Our nightly
> and aurora audiences are small, but I would have figured seeing at least one
> crash.

Given that, there's no need to track for a specific release. This will remain on the topcrash list.

This crash is currently at #5 on beta.

Comment 16

[Benoit Girard (:BenWa)]

Until we get steps let's get the tegra driver so we can pull symbols. This will give us better crash reports for Nexus 7 tegra crashes.

Can someone who has a tegra 3 device pull the library and email them to ted? They are typically, but not always, in '/system/lib/vendor'.

Comment 17

[Kevin Brosnan [Ex-Mozilla]]

Pulled a full copy of /system/lib @ http://people.mozilla.com/~kbrosnan/tmp/782223/nexus-7-Android-4.2.2.zip handing off to Ted.

Comment 18

[(not currently active) Ted Mielczarek]

Attachment: stack from bp-fe455f1a-978d-4204-a250-0ca4d2130417.

Okay, I finally dumped symbols out of these and fiddled with it. Along the way I found a bug in the symbol dumping code I had written, so after I fixed that I found that this stack goes into libflashplayer.so. :-/

This stack is from bp-fe455f1a-978d-4204-a250-0ca4d2130417.

I'll put these symbols on the symbol server and we'll see if reprocessing them clears anything up.

Comment 19

[(not currently active) Ted Mielczarek]

Adding the symbolized signature. It's not actually in this function (it's libGLESv2_tegra.so!NvGl214081946277 + 0x246d4, so it's *way* outside), but that's just an artifact of how Breakpad deals with export symbols.

Comment 20

[Benoit Girard (:BenWa)]

Pulling in snorp since flash is causing the crash.

Ted should we look into getting you that libflashplayer.so?

Comment 21

[(not currently active) Ted Mielczarek]

kbrosnan pointed me at the Adobe download page, I'll track it down.

Comment 22

[(not currently active) Ted Mielczarek]

It doesn't appear to be any of the versions of Flash 11.5 on the Adobe download page. We have a symbol file for it on the symbol server that I uploaded in January, but it doesn't seem to have unwind info covering the function in question, so it doesn't make the stack any better. I have no idea where I got that symbol file from.

Comment 23

[(not currently active) Ted Mielczarek]

These reports are still coming in, they're using the new symbols I uploaded:
https://crash-stats.mozilla.com/report/list?product=FennecAndroid&query_search=signature&query_type=exact&query=NvGl214081946277&reason_type=contains&date=04%2F18%2F2013%2011%3A39%3A43&range_value=1&range_unit=weeks&hang_type=any&process_type=any&do_query=1&admin=1&signature=NvGl214081946277

(adjusting signature because I flubbed it the first time)

The stack ends in libflashplayer.so, unhelpfully:
https://crash-stats.mozilla.com/report/index/f3c1431d-769e-46d3-ac45-8eb302130418

I guess that's better than it used to be, though.

Comment 24

[(not currently active) Ted Mielczarek]

Attachment: everything on the stack

I ran my dump-lookup tool against the crash from comment 18, and once you get past the flashplayer stuff there's a plausible looking stack in Gecko on the bottom. The last bit of Gecko on the stack appears to be PluginTimerCallback:
http://hg.mozilla.org/releases/mozilla-release/annotate/48255272a995/dom/plugins/base/nsNPAPIPluginInstance.cpp#l1430

Comment 25

[James Willcox (:snorp) ([e-mail]) (he/him)]

(In reply to Ted Mielczarek [:ted.mielczarek] from comment #22)
> It doesn't appear to be any of the versions of Flash 11.5 on the Adobe
> download page. We have a symbol file for it on the symbol server that I
> uploaded in January, but it doesn't seem to have unwind info covering the
> function in question, so it doesn't make the stack any better. I have no
> idea where I got that symbol file from.

I generated symbol files for you a while back (apparently January) by running the appropriate things on all the libraries on the Flash archive web page.

Comment 26

[bhavana bajaj [:bajaj]]

Adding needsurl's to see if we have any new/popular urls here for QA to start here as the url's posted in comment# 8 are ~5months old.

Comment 27

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

There's only 2 URLs that I saw in one of the crash signatures

https://apps.facebook.com/houseoffun/?hof_gift=eyJpZCI6IjI2NTE4ODM0MWMzMGE4MSIsI
http://animeon.pl/anime.php?type=player&episode_id=1474

I don't see crashes in others.

Comment 28

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

http://news.walla.co.il/?w=/551/2639126
http://st.keirin.jp/portal/pc/index.html
http://diggio.altervista.org/diggiotvandroidapps/prove/canale5.php
http://www.heroeswm.ru/
http://weeb.tv/channel/t-v-n-h-d
http://pgatour.cbssports.com/live/?media=387929
http://www.globalsp.ho8.com/
http://eu.battle.net/d3/en/console/
http://www.hlntv.com/video/2013/05/06/jodi-arias-tweets-prosecutor-juan-martinez
http://live.nicovideo.jp/watch/lv134638940
http://www.gamethrones.ru/online/3_season/s03e02.php
https://apps.facebook.com/houseoffun/?fb_source=bookmark_favorites&ref=bookmarks
http://www.tv4e.gr/livestreaming.php
http://www.putlocker.com/file/AAC67590485E2FAF#
https://sites.google.com/site/nterfoxlady/los-dibujos-animados-de-tu-vida
about:home

Comment 29

[Scoobidiver (away)]

Adding the JB 4.3 signature.

Comment 30

[Alex Keybl [:akeybl]]

Hey Snorp, we're leaning towards adding [unactionable] to the whiteboard since this has had such little traction. If you agree, please add [unactionable]

Comment 31

[James Willcox (:snorp) ([e-mail]) (he/him)]

(In reply to Alex Keybl [:akeybl] from comment #30)
> Hey Snorp, we're leaning towards adding [unactionable] to the whiteboard
> since this has had such little traction. If you agree, please add
> [unactionable]

I think that's ok unless we get some further information later on.

Comment 32

[Kevin Brosnan [Ex-Mozilla]]

This is not a top crash any more. Its rank in 25 release is in the 40s.

Comment 33

[u597032]

Flash is going away: Bug 1381916.

Comment 34

[Chris Peterson [:cpeterson]]

Removing steps-wanted keyword because this bug has been resolved.

Comment 35

[Chris Peterson [:cpeterson]]

Removing steps-wanted keyword because this bug has been resolved.