851366 - [B2G] OS crashes upon sending SMS

Status: RESOLVED WORKSFORME

(Firefox OS Graveyard :: Gaia, defect)

Description

[Deepa Subramanian]

Description:
Use two devices with build ID 20130314070204. On device(A) have FM radio palying and send a msg from device (b) to device (A). SMS not sent but have Firefox OS crash message.  

Repro Steps:
1) Updated to Unagi Build ID: 20130314070204
2) Device A playing FM radio
3) Compose an SMS on Device B and click "send"
4) SMS not received 
5) Device B crashes  

Expected:
Device A to receive SMS 

Actual:
Device B Crashes

Repro frequency:100%

Environmental  Variables:
Kernel Date: Dec 5
Gecko: http://hg.mozilla.org/releases/mozilla-b2g18/rev/dc3add076016
Gaia: 1e1c8c0ff2bc7f252fbe95016f108e38ece691a9

Notes:
Optional: Log attached

Comment 1

[Deepa Subramanian]

Attachment: Attachment 851366

Comment 2

[Deepa Subramanian]

Attachment: attachment 2 - 851366

Comment 3

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

Barker had checked for us.

Looks like either the device has been flashed or no crash report was made since the time of the report.

Comment 4

[cbarker]

Attachment: Crash Report Pending Files

Comment 5

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

Hrm.  I can't seem to reproduce the bug using : 2013-03-14-11-49-15 and 2013-03-15-07-02-20. I wonder if it's a residue issue from the bad merge...

Comment 6

[Allen Maxwell]

This issue no longer repros on Mozilla RIL OTA build 20130315070220, but does occur on Commercial RIL Flash Build 20130315070220 both same Gaia and Gecko.

Gecko: http://hg.mozilla.org/releases/mozilla-b2g18/rev/78be5f74e410
Gaia: cd028015a499e4fc3a652d7aa6eb9f81a1ea8cd7

Comment 7

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

Use ted's script to submit the crash reports attached (see http://shizen008.wordpress.com/2013/03/16/submitting-pending-crashes/ for instructions and downloading the script)

It spit out output for the submitted crashes:
https://crash-stats.mozilla.com/report/index/bp-9986e266-fa62-4ffc-a97c-3a6bc2130316
https://crash-stats.mozilla.com/report/index/bp-4492657a-e078-4a5d-90e3-24d622130316
https://crash-stats.mozilla.com/report/index/bp-6f2b83dc-a1a5-4dd1-b006-a436f2130316
https://crash-stats.mozilla.com/report/index/bp-64991539-0c46-4976-91d0-d44742130316

Comment 8

[Anshul]

Which apps AU was used for this test?

Comment 9

[Michael Vines [:m1] [:evilmachines]]

Pretty sure this is a gecko/ril mismatch, especially given there was a recent sms-related ril interface change in gecko recently.  Who is responsible for creating these builds from Moz, I think we need to check with them on what ril binary was used?
(clearing Alhad as the assignee for now)

Comment 10

[Vicamo Yang [:vicamo][:vyang]]

About the interface mismatch, we landed changes to enable MMS support in bug 839436. Changes made to nsIRilMobileMessageDatabaseCallback is a must, but we can do it in a more C++ friendly way and is already in progress by Gene in bug 847738 (reviewing). Note that even bug 847738 is landed, the interface is still different from the one commercial RIL is using.

Comment 11

[Anshul]

Vicamo, is this issue reported for v1.1 or v1.0.1?

Comment 12

[Michael Vines [:m1] [:evilmachines]]

v1.1.  Interface change itself is fine really, of course not changing is preferred but changes will be necessary at times. 

However in the AU builds on CAF we advertise the compatible moz build ID precisely to avoid this from happening!  So my question is really about how we can start using the x-mozbuild attribute [1] in the Mozilla builds to prevent a mismatch like this from happening again.


[1] https://www.codeaurora.org/gitweb/quic/b2g/?p=manifest.git;a=blob;f=caf_AU_LINUX_GECKO_ICS_STRAWBERRY.01.01.00.19.031.xml;h=5d18c6d3a72fbe933e867e0a0f8af010a59b237b;hb=7a224b27b6df8c9043aa0e0e4112b241706fd993#l3

Comment 13

[Kai-Chih Hu [:khu]]

Vicamo, thank you for your detailed information. Would you mind if we assign you as the owner? Or, who do you think could be the best owner to handle this case? Thanks.

Comment 14

[Vicamo Yang [:vicamo][:vyang]]

(In reply to khu from comment #13)
> Vicamo, thank you for your detailed information. Would you mind if we assign
> you as the owner? Or, who do you think could be the best owner to handle
> this case? Thanks.

The idea Michael has is nice, but that's something needs help from release team. Gene and I can help landing bug 847738 asap, but we have really little idea about commercial RIL integratin. :(

Comment 15

[Michael Vines [:m1] [:evilmachines]]

Yep, I think this bug could be assigned to the team that is building the releases with the commercial RIL to check how we can avoid the binary incompatibility next time.

Also maybe there's a way that we can express a RIL interface version independent of a Gecko SHA1?  The interface doesn't change that often, maybe we could use the |sha1sum| of nsIRadioInterfaceLayer.idl and perhaps a couple other Gecko headers to allow us to be less restrictive about the gecko/RIL combo.

Comment 16

[Tony Chung [:tchung]]

(In reply to Michael Vines [:m1] [:evilmachines] from comment #9)
> Pretty sure this is a gecko/ril mismatch, especially given there was a
> recent sms-related ril interface change in gecko recently.  Who is
> responsible for creating these builds from Moz, I think we need to check
> with them on what ril binary was used?
> (clearing Alhad as the assignee for now)

cc'ing folks from mozilla release engineering to take a look.

Comment 17

[Tony Chung [:tchung]]

Rail just built me a ril package with .030, and i do not see the crash.  
the crash still happens with .041.

mismatch confirmed.

Comment 20

[Marcia Knous [:marcia]]

I am using .032 and SMS is crashing on both send and receive. amaxell is reporting that SMS send/receive is crashing for him on .030 using today's build.

Comment 21

[Rail Aliiev [:rail]]

If Friday build + .030 doesn't crash and Monday build + 0.030 does crash it sounds like a regression/bug in the Monday build, not in RIL.

If we want reliable results we shouldn't use latest RIL (.032 in today case), but QA signed off RIL.

Another possibility would be creating 2 repacks: latest and signed off.

Comment 22

[Rail Aliiev [:rail]]

FYI, v1-train RIL has been updated to .033 tonight.

Comment 23

[John Hammink]

We're still seeing this one tuesday build, with .033 RIL.

Comment 24

[Anshul]

Alhad, could you please have a look?

Comment 25

[Allen Maxwell]

issue no longer persists on:
Unagi Build ID: 20130320070206
Kernel Date: Dec 5
Gecko: http://hg.mozilla.org/releases/mozilla-b2g18/rev/778da49486f0
Gaia: 6c3767c2dea43b5e9aff7d156d36d69649005621

Works For Me

Comment 26

[Alhad Purnapatre]

Looks like its working now.