1047618 - MOZ_ASSERT(NS_IsMainThread()) at ../../hal/Hal.cpp:77

Status: RESOLVED FIXED

(Firefox OS Graveyard :: GonkIntegration, defect)

Description

[Gregor Wagner [:gwagner]]

on current trunk with debug gecko on nexus 4 when pressing the power button

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 1024.1077]
0xb48a5ae6 in AssertMainThread () at ../../hal/Hal.cpp:77
77	  MOZ_ASSERT(NS_IsMainThread());
(gdb) bt
#0  0xb48a5ae6 in AssertMainThread () at ../../hal/Hal.cpp:77
#1  mozilla::hal::(anonymous namespace)::AssertMainThread () at ../../hal/Hal.cpp:75
#2  0xb48a5d1e in mozilla::hal::GetScreenEnabled () at ../../hal/Hal.cpp:373
#3  0xb50ad8ae in android::EventHub::getEvents (this=0xb2c8e1d0, timeoutMillis=-1, buffer=0xb0c71108, bufferSize=256) at ../../../widget/gonk/libui/EventHub.cpp:865
#4  0xb50b6fbe in android::InputReader::loopOnce (this=0xb0c71000) at ../../../widget/gonk/libui/InputReader.cpp:268
Backtrace stopped: previous frame inner to this frame (corrupt stack?)

Comment 1

[Gregor Wagner [:gwagner]]

Mike, Dave, any idea where this comes from?

Comment 2

[Dave Hylands [:dhylands]]

Probably related to bug 932698

https://hg.mozilla.org/mozilla-central/rev/07c91b928cd7#l4.36

Comment 3

[Michael Wu [:mwu]]

FYI, changes to widget/gonk/libui are not permitted *except* to update to upstream code. Bug 932698 needs to be backed out.

Comment 4

[Ryan VanderMeulen [:RyanVM]]

Bug 932698 backed out.
https://hg.mozilla.org/mozilla-central/rev/6551323d96e6

Comment 5

[viral [:viralwang]]

(In reply to Gregor Wagner [:gwagner] from comment #0)
> on current trunk with debug gecko on nexus 4 when pressing the power button
> 
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 1024.1077]
> 0xb48a5ae6 in AssertMainThread () at ../../hal/Hal.cpp:77
> 77	  MOZ_ASSERT(NS_IsMainThread());
> (gdb) bt
> #0  0xb48a5ae6 in AssertMainThread () at ../../hal/Hal.cpp:77
> #1  mozilla::hal::(anonymous namespace)::AssertMainThread () at
> ../../hal/Hal.cpp:75
> #2  0xb48a5d1e in mozilla::hal::GetScreenEnabled () at ../../hal/Hal.cpp:373
> #3  0xb50ad8ae in android::EventHub::getEvents (this=0xb2c8e1d0,
> timeoutMillis=-1, buffer=0xb0c71108, bufferSize=256) at
> ../../../widget/gonk/libui/EventHub.cpp:865
> #4  0xb50b6fbe in android::InputReader::loopOnce (this=0xb0c71000) at
> ../../../widget/gonk/libui/InputReader.cpp:268
> Backtrace stopped: previous frame inner to this frame (corrupt stack?)

Hi Gregor,

May I have more information about how to reproduce the symptom you met?
Here are my steps:
1) ./config.sh nexus-4
2) move gecko to the commit before bug 932698 back out. (e953ea9c38f362bced31ff142157dd592c23d862)
commit 00471b3504de90d3774387984e8b13a3f62e43d9
Author: Ryan VanderMeulen <[e-mail]>
Date:   Fri Aug 1 16:46:21 2014 -0400

    Backed out changeset 07c91b928cd7 (bug 932698) for causing bug 1047618. a=me

commit e953ea9c38f362bced31ff142157dd592c23d862
Merge: 65b0793 60e99f3
Author: Ryan VanderMeulen <[e-mail]>
Date:   Fri Aug 1 15:59:05 2014 -0400

    Merge inbound to m-c. a=merge

3) ./build.sh
4) ./flash.sh

I expect I can see the Segmentation fault when press power key, but it looks normal after several times suspend/resume. Any suggestion I can have?
Thank you.

Comment 6

[Gregor Wagner [:gwagner]]

You have to build a debug version of gecko. You either set B2G_DEBUG=1 when you configure or you can modify gonk-misc/default-gecko-config to have 'ac_add_options --enable-debug' enabled.
Make sure to clobber before you create a new build.