1139870 - [B2G][PDF viewer] PDF viewer crashes when opening/closing PDF files

Status: RESOLVED WONTFIX

(Firefox OS Graveyard :: Gaia::PDF Viewer, defect)

Description

[Noemí Freire (:noemi)]

Attachment: PDF_Viewer_Crash.txt

Sometimes PDF viewer crashes when opening/closing it from E-mail app (downloaded attachment) and Download Manager (downloaded file). We've seen it in the following scenarios:
1- Being in E-mail app
1a- click on the attachment viewer icon
1b- close the PDF viewer and go to 1a
2- Being in Download Manager (Settings->Downloads)
2a- click on a pdf entry -> Open
2b- close the PDF viewer and go to 2a

The crash has been seen in 1a, 1b, 2a and 2b steps. There are no specific STRs that makes this issue 100% reproducible. We thought it was related to the size of the pdf file since it appeared to be more reproducible with files bigger than 200KB but it has been also seen with files around 32KB. Please see below a pair of videos gathering PDF viewer crashes:

*https://bugzilla.mozilla.org/attachment.cgi?id=8573173
*https://www.youtube.com/watch?v=Ao7filtnUkY

Please find attached a logcat gathering 1a scenario

Environmental variables:
Flame 3.0
BuildId: 20150305082433
Gecko: fcde1ca
Gaia: a18b457
Platform version: 39.0a1
Firmware Version: v18D

Comment 1

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

I/Gecko   (  215): 
I/GeckoDump(  215): Crash reporter : Not online, postponing.
I/GeckoDump( 5600): [33mWAR: Problem with "open" activity ActivityCanceled[0m
E/HWComposer(  215): Non-uniform vsync interval: 283550311
W/GeckoConsole(  215): [JavaScript Error: "NS_ERROR_NOT_IMPLEMENTED: SetNFCFocus for in-process mode is not yet supported" {file: "jar:file:///system/b2g/omni.ja!/components/BrowserElementParent.js" line: 902}]

Seems like there's a couple of things going on.  Going to try to repro to get the crash id.

Comment 2

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

I can't reproduce this bug.  I think it might be something in the PDF?

Can you connect to wifi and send the report please?  I would like to get the crash id.

Comment 3

[Maria Angeles Oteo (:oteo)]

Attachment: bug1139870_traces.rtf

Hi Naoki,
I am also able to reproduce the bug easily whith a 287K pdf file. Here I send you my traces in case it can help.

Note: after downloading and viewing the pdf file (as the STR explains) kill the e-mail app and launch it again to try to increse the reproduction rate, I don't a better STR :(

Comment 4

[Maria Angeles Oteo (:oteo)]

Environmental variables:
Flame 3.0
BuildId: 20150308082154
Gecko: d5d934b
Gaia: 373c9dd
Platform version: 39.0a1
Firmware Version: v18D

Comment 5

[Noemí Freire (:noemi)]

(In reply to Naoki Hirata :nhirata (please use needinfo instead of cc) from comment #2)
> I can't reproduce this bug.  I think it might be something in the PDF?
> 
> Can you connect to wifi and send the report please?  I would like to get the
> crash id.

Hi Naoki,

Please find below the ids corresponding to 3 reports I sent related to this issue. Thanks

*bp-37b02114-eb1e-45dd-9a58-21b3b2150308
*bp-8e3ced33-16aa-4cac-81bc-008322150308
*bp-cd5fbb16-37ea-4539-a795-aad212150308

Comment 6

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

Thanks Noemí.  unfortunately, I don't have the symbols for the build you used.  All three of those crashes seem to be the same crash; or at least they all crashed with the pdf viewer active is all I can see.

Comment 7

[Antonio Manuel Amaya Calvo (:amac)]

What do you need, Naoki? We can generate a debug build for Noemi, and we can upload the symbols to some place you can reach, or Noemi can download an install a build provided by you if you prefer that.

Comment 8

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

Hi Antonio, 

What I would like to do is try to get a crash stack associated with symbols; this will tell you how to do this manually ( https://developer.mozilla.org/en-US/docs/Debugging_a_minidump
 ).  Otherwise, you could potentially send the symbols to us, have Noemí crash the pdf viewer, w/ wifi access and then place in the crash id.  We can place the symbols on the symbol server and then Socorro ( https://wiki.mozilla.org/Socorro ) will do the matching.

Currently all I see is libxul : https://crash-stats.mozilla.com/report/index/37b02114-eb1e-45dd-9a58-21b3b2150308 ; which isn't helpful.

Comment 9

[Noemí Freire (:noemi)]

(In reply to Naoki Hirata :nhirata (please use needinfo instead of cc) from comment #8)
> Hi Antonio, 
> 
> What I would like to do is try to get a crash stack associated with symbols;
> this will tell you how to do this manually (
> https://developer.mozilla.org/en-US/docs/Debugging_a_minidump
>  ).  Otherwise, you could potentially send the symbols to us, have Noemí
> crash the pdf viewer, w/ wifi access and then place in the crash id.  We can
> place the symbols on the symbol server and then Socorro (
> https://wiki.mozilla.org/Socorro ) will do the matching.
> 
> Currently all I see is libxul :
> https://crash-stats.mozilla.com/report/index/37b02114-eb1e-45dd-9a58-
> 21b3b2150308 ; which isn't helpful.

Hi,

please find the crashreporter-symbols in the following link: https://www.dropbox.com/s/gpog4y4y5k1xdwf/crashreporter-symbols.tgz?dl=0
hope it helps. Thanks!

Comment 10

[Maria Angeles Oteo (:oteo)]

Setting ni to Naoki to check the crashreporter attached in comment 9 by Noemí

Comment 11

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

Attachment: crashstack.txt

Hi.  Thanks for the needinfo.  That's the best way to get my attention in bugs.

The symbols helped some.  Having said that, are these the symbol files from the build you crashed with?  The symbols have to match with the build you used to crash, otherwise I will only get a partial match of the dump to understand the crash stack.

Comment 12

[Noemí Freire (:noemi)]

(In reply to Naoki Hirata :nhirata (please use needinfo instead of cc) from comment #11)
> Created attachment 8580728 [details]
> crashstack.txt
> 
> Hi.  Thanks for the needinfo.  That's the best way to get my attention in
> bugs.
> 
> The symbols helped some.  Having said that, are these the symbol files from
> the build you crashed with?  The symbols have to match with the build you
> used to crash, otherwise I will only get a partial match of the dump to
> understand the crash stack.

Yes, it should...

Comment 13

[BMO Automation]

Firefox OS is not being worked on