941985 - [1.3] Browser crashes when running HTML5Test benchmark

Status: RESOLVED FIXED

(Core :: Networking, defect)

Description

[Mandyam Vikram]

Attachment: html5test1.3failedrun.log

This occurs on 1.3 on a 8x10 device (multi-core) on the following build:
Gecko:
56e157765fee56690a296bfdc7dc9206fb7e8ffc
Gaia:
ac42cb33f21b3f13595432c965f44615daae2225


Launch the browser pointing to the following URL:
http://html5test.com/

(The HTML5Test benchmark has been updated recently). When running this benchmark on the browser, the Browser crashes "Well this is embarassing..."
I've attached a logcat of the run. The content process is getting killed. Here is a snippet from the failed run:
------
I/Gecko   (  327): 
I/Gecko   (  327): ###!!! [Parent][DispatchAsyncMessage] Error: Value error: message was deserialized, but contained an illegal value
I/Gecko   (  327): 
I/Gecko   (  327): [Parent 327] WARNING: pipe error (127): Connection reset by peer: file /local/mnt/workspace/lnxbuild/project/release_dev_msm8610_1973053/checkout/gecko/ipc/chromium/src/chrome/common/ipc_channel_posix.cc, line 446
I/Gecko   (  327): 
I/Gecko   (  327): ###!!! [Parent][MessageChannel] Error: Channel error: cannot send/recv
I/Gecko   (  327): 
I/Gecko:ProcessPriorityManager(  327): [Browser, child-id=6, pid=1861] Scheduling reset timer to fire in 1000ms.
I/Gecko   (  327): -*- QCMessageManager_QC_B2G: receiveMessage: child-process-shutdown arrived from content process
I/Gecko:ProcessPriorityManager(  327): [Browser, child-id=6, pid=1861] ScheduleResetPriority bailing; the timer is already running.
I/Gecko:ProcessPriorityManager(  327): [child-id=6, pid=-1] Destroying ParticularProcessPriorityManager.
I/Gonk    (  327): NFC: OnConnectError
I/Gonk    (  327): NFC: OnDisconnect
I/Gecko   (  327): [Parent 327] WARNING: waitpid failed pid:1861 errno:10: file /local/mnt/workspace/lnxbuild/project/release_dev_msm8610_1973053/checkout/gecko/ipc/chromium/src/base/process_util_posix.cc, line 254
I/Gonk    (  327): NFC: OnConnectError
I/Gonk    (  327): NFC: OnDisconnect
D/wpa_supplicant( 1072): RX ctrl_iface - hexdump(len=11): 53 49 47 4e 41 4c 5f 50 4f 4c 4c
D/wpa_supplicant( 1072): wlan0: Control interface command 'SIGNAL_POLL'
D/wpa_supplicant( 1072): nl80211: survey data missing!
I/Gecko   (  327): [Parent 327] WARNING: waitpid failed pid:1861 errno:10: file /local/mnt/workspace/lnxbuild/project/release_dev_msm8610_1973053/checkout/gecko/ipc/chromium/src/base/process_util_posix.cc, line 254
I/Gecko   (  327): [Parent 327] WARNING: Failed to deliver SIGKILL to 1861!(3).: file /local/mnt/workspace/lnxbuild/project/release_dev_msm8610_1973053/checkout/gecko/ipc/chromium/src/chrome/common/process_watcher_posix_sigchld.cc, line 118
-------

As a note of comparison, this benchmark runs successfully on 1.2 on a 7x27 device on the following build:
Gecko:
14432474328e7558902329c3fc2a8238c3222a7f
Gaia:
7a23f8c53ba97da9c63a7275b36d155b4526a639

Comment 1

[Mandyam Vikram]

Attachment: html5test1.2goodrun.log

Good run on 1.2

Comment 2

[[:fabrice] Fabrice Desré]

What I get on a debug build is:
JavaScript warning: http://html5test.com/scripts/engine.js, line 1: WebGL: Can't get a usable WebGL context
JavaScript warning: http://html5test.com/scripts/engine.js, line 1: WebGL: Can't get a usable WebGL context
[Child 13207] WARNING: NS_ENSURE_TRUE(aSelection->GetRangeCount()) failed: file /home/fabrice/dev/birch/editor/libeditor/base/nsEditor.cpp, line 3789
[Child 13207] WARNING: NS_ENSURE_SUCCESS(rv, rv) failed with result 0x80004005: file /home/fabrice/dev/birch/editor/libeditor/base/nsEditor.cpp, line 3768
[Child 13207] WARNING: NS_ENSURE_SUCCESS(res, res) failed with result 0x80004005: file /home/fabrice/dev/birch/editor/libeditor/html/nsHTMLEditRules.cpp, line 316
[Child 13207] WARNING: NS_ENSURE_TRUE(aSelection->GetRangeCount()) failed: file /home/fabrice/dev/birch/editor/libeditor/base/nsEditor.cpp, line 3789
[Child 13207] WARNING: NS_ENSURE_SUCCESS(rv, rv) failed with result 0x80004005: file /home/fabrice/dev/birch/editor/libeditor/base/nsEditor.cpp, line 3768
[Child 13207] WARNING: NS_ENSURE_SUCCESS(res, res) failed with result 0x80004005: file /home/fabrice/dev/birch/editor/libeditor/html/nsHTMLEditRules.cpp, line 8482
[Child 13207] WARNING: NS_ENSURE_TRUE(mDisabledJSAndPlugins) failed: file /home/fabrice/dev/birch/editor/composer/src/nsEditingSession.cpp, line 208
[Parent 13075] WARNING: Not support UDP socket in content process, aborting subprocess: file /home/fabrice/dev/birch/netwerk/ipc/NeckoParent.cpp, line 386

###!!! [Parent][DispatchAsyncMessage] Error: Value error: message was deserialized, but contained an illegal value


###!!! [Parent][MessageChannel] Error: Channel error: cannot send/recv

[Parent 13075] WARNING: NS_ENSURE_TRUE(mOwnerContent) failed: file /home/fabrice/dev/birch/content/base/src/nsFrameLoader.cpp, line 2383

It looks like it's because we don't have webrtc ipc enabled: https://mxr.mozilla.org/mozilla-central/source/netwerk/ipc/NeckoParent.cpp#384 (I checked that the pref is not set except in tests).

Eric, does this sound reasonable to turn this on in b2g?

Comment 3

[Jason Smith [:jsmith]]

Adding some webrtc folks to the bug to get input on the above comment

Comment 4

[Jason Smith [:jsmith]]

This looks like this was broken by the changes in bug 869869. According to jesup, the real fix we need is bug 870660.

If QC's blocked here, then we might need to consider backing out the regressing patch.

Comment 5

[Eric Rescorla (:ekr)]

Based on talking to Fabrice it seems like the problem is that they are testing PeerConnection.

But PeerConnection isn't ready on either 1.2 or 1.3 (though it should be ready on 1.3 as soon as
this patch lands).

We should make sure the PeerConnection pref is turned off on B2G.

Comment 6

[Jason Smith [:jsmith]]

Can someone double check this isn't reproducing on 1.2? I don't think it should be, but we should double check.

Comment 7

[Eric Rescorla (:ekr)]

It probably isn't reproducing on 1.2 because this crash is new code,
but it might be that PeerConnection is on and busted.

Comment 8

[bhavana bajaj [:bajaj]]

(In reply to Eric Rescorla (:ekr) from comment #5)
> Based on talking to Fabrice it seems like the problem is that they are
> testing PeerConnection.
> 
> But PeerConnection isn't ready on either 1.2 or 1.3 (though it should be
> ready on 1.3 as soon as
> this patch lands).
> 
> We should make sure the PeerConnection pref is turned off on B2G.
Is there a known bug on file to do that, else happy to file one. I think that's the way to go to resolve this bug for now.

Comment 9

[Jason Smith [:jsmith]]

Spoke with mreavy & jesup in IRC - there's agreement that we should pref off Peer Connection on B2G for now until we know the implementation is ready for general consumption.

Comment 10

[Jason Smith [:jsmith]]

I filed bug 942343 to track the pref off. Removing the blocking flag as the pref off will resolve the bug.

Comment 11

[Mandyam Vikram]

This issue has been verified to be resolved on the following build:
Gecko:
e5638ae70aad34961a65e6a51b15ffe2f3553b0d
Gaia:
1abda08e450cb66a61a31bdcfd3352e2df9d9ace

This can be closed.