White/blank screen on Motorola devices

RESOLVED FIXED in Firefox 14

Status

()

Core
Graphics: Layers
--
major
RESOLVED FIXED
5 years ago
5 years ago

People

(Reporter: yury, Unassigned)

Tracking

Trunk
mozilla14
ARM
Android
Points:
---

Firefox Tracking Flags

(firefox14 fixed, firefox15 fixed, blocking-fennec1.0 beta+)

Details

(Whiteboard: [MTD])

Attachments

(5 attachments)

(Reporter)

Description

5 years ago
Web page or screen you were on when you saw the issue: 

Steps to reproduce (Droid Bionic):
1. Type any URL (even "about:crashes")
2. Nothing is displayed
3. Hit "list of tabs" icon of the tabs
4. Preview images display the page(s) content

What you expected:
Content of the page is visible on step 2

Crash report ID (if applicable):
Confirmed, using the Droid Bionic from DeviceAnywhere.
This has Android2.3.4 on it.

From gsmarena.com:
http://www.gsmarena.com/motorola_droid_bionic_xt875-3710.php
	OS 	Android OS, v2.3.4 (Gingerbread)
Chipset 	TI OMAP 4430
CPU 	Dual-core 1 GHz Cortex-A9
GPU 	PowerVR SGX540
blocking-fennec1.0: --- → ?
Created attachment 621027 [details]
Logcat when loading about:crashes

This is a catlog while loading about:crashes, which turns up white.
Note that loading any website turns up white on that phone, so Fennec is unusable on that phone.

I also crashed twice while starting up (and tapping into the url bar, I think):
https://crash-stats.mozilla.com/report/index/1b48de7e-c4f2-4706-ad8f-e42bf2120504
https://crash-stats.mozilla.com/report/index/7900cbf1-685f-428e-8e11-57ad22120504
bug 737928 NEW crash in js::detail::RegExpCode::compile
Other phones similar specs; same issues?

* Motorola Atrix 2
* Motorola RAZR XT910
This might be a regression (from the Maple landing?), we should look and see if we can find a regression range here.
Hopefully the catlog attach has some useful info here.
Keywords: regressionwindow-wanted
Version: Firefox 14 → Trunk

Comment 5

5 years ago
Some highlights from the log:

05-04 06:09:28.755 D/dalvikvm( 3298): No JNI_OnLoad found in /data/data/org.mozilla.fennec/lib/libmozglue.so 0x4051bc50, skipping init

05-04 06:09:28.857 I/dalvikvm( 3298): Failed resolving Lorg/mozilla/gecko/gfx/SurfaceTextureLayer; interface 105 'Landroid/graphics/SurfaceTexture$OnFrameAvailableListener;'
05-04 06:09:28.857 W/dalvikvm( 3298): Link of class 'Lorg/mozilla/gecko/gfx/SurfaceTextureLayer;' failed
05-04 06:09:28.857 I/dalvikvm( 3298): Could not find method org.mozilla.gecko.gfx.SurfaceTextureLayer.create, referenced from method org.mozilla.gecko.GeckoApp.createSurface

Later, we get:

05-04 06:09:30.380 I/Gecko   ( 3273): === SHADER COMPILATION FAILED ===

05-04 06:09:30.404 I/Gecko   ( 3273):  -- creating basic, not accelerated

At this point, we've failed to create the compositor, and this will indeed cause a blank screen to appear.

Kats, any idea what could be causing the dalvikvm errors above?

In any case, a regression range will be helpful.
I can confirm that this same issue is present on the Motorola Atrix 2:

Android OS, v2.3 (Gingerbread)
Chipset	TI OMAP 4430
CPU	Dual-core 1 GHz Cortex-A9
GPU	PowerVR SGX540

Updated

5 years ago
Summary: White/blank screen → White/blank screen on Motorola devices
(In reply to Ali Juma [:ajuma] from comment #5)
> 05-04 06:09:28.755 D/dalvikvm( 3298): No JNI_OnLoad found in
> /data/data/org.mozilla.fennec/lib/libmozglue.so 0x4051bc50, skipping init
> 

This one is normal I think, I see it often.

> 05-04 06:09:28.857 I/dalvikvm( 3298): Failed resolving
> Lorg/mozilla/gecko/gfx/SurfaceTextureLayer; interface 105
> 'Landroid/graphics/SurfaceTexture$OnFrameAvailableListener;'
> 05-04 06:09:28.857 W/dalvikvm( 3298): Link of class
> 'Lorg/mozilla/gecko/gfx/SurfaceTextureLayer;' failed
> 05-04 06:09:28.857 I/dalvikvm( 3298): Could not find method
> org.mozilla.gecko.gfx.SurfaceTextureLayer.create, referenced from method
> org.mozilla.gecko.GeckoApp.createSurface

It looks like android.graphics.SurfaceTexture is only available on API level >= 11. Adding snorp - I'm not sure if this needs to be in an API level guard or what.
Attachment #621027 - Attachment mime type: audio/mpeg → application/zip
(In reply to Kartikaya Gupta (:kats) from comment #7)
> (In reply to Ali Juma [:ajuma] from comment #5)
> > 05-04 06:09:28.755 D/dalvikvm( 3298): No JNI_OnLoad found in
> > /data/data/org.mozilla.fennec/lib/libmozglue.so 0x4051bc50, skipping init
> > 
> 
> This one is normal I think, I see it often.
> 
> > 05-04 06:09:28.857 I/dalvikvm( 3298): Failed resolving
> > Lorg/mozilla/gecko/gfx/SurfaceTextureLayer; interface 105
> > 'Landroid/graphics/SurfaceTexture$OnFrameAvailableListener;'
> > 05-04 06:09:28.857 W/dalvikvm( 3298): Link of class
> > 'Lorg/mozilla/gecko/gfx/SurfaceTextureLayer;' failed
> > 05-04 06:09:28.857 I/dalvikvm( 3298): Could not find method
> > org.mozilla.gecko.gfx.SurfaceTextureLayer.create, referenced from method
> > org.mozilla.gecko.GeckoApp.createSurface
> 
> It looks like android.graphics.SurfaceTexture is only available on API level
> >= 11. Adding snorp - I'm not sure if this needs to be in an API level guard
> or what.

That's normal to see when you're on a Gingerbread device. It should be harmless. We don't ever actually call the code in question on that device.
Looking at the log, I agree with Ali that the shader compilation failure is the likely culprit here.
Can you give me 'built from' changeset from about:buildconfig? I'm curious to see if on demand shader loading (from bug 716439) is in your build.
Nevermind, you linked a crash report that had your buildid. That change is not in yet.

Updated

5 years ago
Severity: normal → major

Updated

5 years ago
Duplicate of this bug: 751589
Reproduced over DeviceAnywhere with a

* Motorola Droid 3 (2.3, TI OMAP 4430, PowerVR SGX540)

Updated

5 years ago
Assignee: nobody → joe
blocking-fennec1.0: ? → beta+
It looks like we're failing to create a context - people see problems very much like this when eglCreateWindowSurface fails.

Unfortunately, the Rogers Motorola RAZR doesn't reproduce this problem. Can anyone verify whether the RAZR reproduces it in general?

The failure to compile a) already has all the error messages output and b) is likely only a symptom of the problem.
What would help us right now is if someone could take a debug Android build and run it with MOZ_GL_DEBUG=1 and MOZ_GL_DEBUG_ABORT_ON_ERROR=1, e.g.

  adb shell am start -a android.activity.MAIN -n org.mozilla.fennec_<user>/.App --es env0 MOZ_GL_DEBUG=1 --es env1 MOZ_GL_DEBUG_ABORT_ON_ERROR=1

Comment 16

5 years ago
Chris, a Droid 3 as per #13 seems to have the problem - we think its available in Europe, can you expense one and see if you can reproduce ASAP?
Milestone 3 is available in Europe, I think. It seems that no carriers carry it here. :(
bug 752153 has reported a similar issue, probably the same on the HTC One X (Tegra 3)

Comment 19

5 years ago
Bug 730890 which is implicated in bug 752153 has not been uplifted to aurora.  Has anyone seen this on Aurora?

Aaron, can you bring your Droid 3 to the office tomorrow?
I tested over DeviceAnywhere services, I don't own one.

Updated

5 years ago
Status: NEW → RESOLVED
Last Resolved: 5 years ago
Resolution: --- → DUPLICATE
Duplicate of bug: 752153
Keywords: regressionwindow-wanted
Not a duplicate. Using the builds mentioned in comment #11 of bug 752153, I could not reproduce.

I'm a bit confused because the current Nightly (05/07) works on the Atrix II where I initially saw the problem with 05/04.
Status: RESOLVED → REOPENED
Resolution: DUPLICATE → ---
Hey Yury,

Would you be able to test current Nightly and see if the problem is resolved? If not, can you try the instructions in comment #11 here? https://bugzilla.mozilla.org/show_bug.cgi?id=752153#c11

Comment 24

5 years ago
(In reply to Aaron Train [:aaronmt] from comment #23)
> Hey Yury,
> 
> Would you be able to test current Nightly and see if the problem is
> resolved? If not, can you try the instructions in comment #11 here?
> https://bugzilla.mozilla.org/show_bug.cgi?id=752153#c11

Based on Yury's response (Bug 752153, Comment 28), it's possible that whatever was causing this bug on May 3rd had been fixed by May 4th.
(Reporter)

Comment 25

5 years ago
(In reply to Ali Juma [:ajuma] from comment #24)
> (In reply to Aaron Train [:aaronmt] from comment #23)
> > Hey Yury,
> > 
> > Would you be able to test current Nightly and see if the problem is
> > resolved? If not, can you try the instructions in comment #11 here?
> > https://bugzilla.mozilla.org/show_bug.cgi?id=752153#c11
> 
> Based on Yury's response (Bug 752153, Comment 28), it's possible that
> whatever was causing this bug on May 3rd had been fixed by May 4th.

Nightly works on the Bionic.
I have the Motorola droid 3 running android 2.3.4. I've had this problem on nightly for the past ~6 weeks.  No webpage will render, it's just a white screen. Also, nightly pretty regularly crashes before loading the white page.

I've been submitting crash reports. Let me know if I can provide any sort of identifier to help track those down to correlate.
Also reproduced issue on Aurora - droid 3 running android 2.3.4.

Only difference is that is doesn't crash as often as nightly, but same white/blank screen behavior.
Created attachment 621765 [details]
blank-page-logcat.txt

I tested mcoates' Droid 3. Here are my log files.

I could load the Fennec Home and About screens, but I was NOT able to load any web pages. I was able to reproduce the blank page problem once. Every other time I tried to load a page, Fennec crashed.

The logs report some suspicious error messages about dalvikvm failing to load org.mozilla.gecko.gfx.SurfaceTextureLayer:

 D ActivityRenderTarget: onPause
 W dalvikvm: Link of class 'Lorg/mozilla/gecko/gfx/SurfaceTextureLayer;' failed
 W dalvikvm: VFY: unable to resolve static method 7598: Lorg/mozilla/gecko/gfx/SurfaceTextureLayer;.create ()Lorg/mozilla/gecko/gfx/SurfaceTextureLayer;
 W dalvikvm: VFY: unable to resolve virtual method 71: Landroid/app/DownloadManager;.addCompletedDownload (Ljava/lang/String;Ljava/lang/String;ZLjava/lang/String;Ljava/lang/String;JZ)J
 W dalvikvm: Link of class 'Lorg/mozilla/gecko/gfx/SurfaceTextureLayer;' failed
 E dalvikvm: Could not find class 'org.mozilla.gecko.gfx.SurfaceTextureLayer', referenced from method org.mozilla.gecko.GeckoApp.showSurface
05-07 15:23:36.052 24821 24821 W dalvikvm: VFY: unable to resolve check-cast 1435 (Lorg/mozilla/gecko/gfx/SurfaceTextureLayer;) in Lorg/mozilla/gecko/GeckoApp;
Created attachment 621766 [details]
aurora-crashes-logcat.txt
joedrew, I tested your try build e5f988326565 on mcoates' Droid 3. I was successfully able to view web pages! Do you want my logcat logs from that test run?

Comment 31

5 years ago
https://tbpl.mozilla.org/?tree=Try&rev=e5f988326565 for a fast link

Updated

5 years ago
Duplicate of this bug: 752708
Mary Colvig has a Droid 3 too (duplicate bug 752708).
for me, this build is broken as the reporter describes:
https://ftp.mozilla.org/pub/mozilla.org/mobile/nightly/2012-05-04-03-05-09-mozilla-central-android/fennec-15.0a1.en-US.android-arm.apk

and this build works fine
https://ftp.mozilla.org/pub/mozilla.org/mobile/nightly/2012-05-05-03-05-10-mozilla-central-android/fennec-15.0a1.en-US.android-arm.apk
those two builds have this range https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=2db9df42823d&tochange=0a48e6561534
Given Brad and Yury's comments, this was fixed 'by accident' somehow. I'm going to bisect tomorrow to see how it got fixed.
Status: REOPENED → RESOLVED
Last Resolved: 5 years ago5 years ago
Resolution: --- → WORKSFORME
bisect results:
Due to skipped revisions, the first good revision could be any of:
changeset:   93046:10cd5a5210f2
parent:      93045:d32d6823f0af
parent:      92905:cc91f155491b
user:        Ehsan Akhgari <ehsan@mozilla.com>
date:        Thu May 03 17:33:52 2012 -0400
summary:     Merge from mozilla-central

changeset:   93047:43b2b050af51
user:        Bas Schouten <bschouten@mozilla.com>
date:        Thu May 03 23:41:37 2012 +0200
summary:     Bug 738413 - Followup: Move ToIntRect to a Rect class method. r=roc

changeset:   93048:26738df8a4e0
user:        Matt Brubeck <mbrubeck@mozilla.com>
date:        Thu May 03 14:55:52 2012 -0700
summary:     Back out 682bf201edde, ef35ba222ac8, and 6a20cf61289d (bug 750907, bug 751003, bug 751377) because of build failure on a CLOSED TREE

that merge has: 
eakhgari@mozilla.com
Thu May 03 14:33:59 2012 -0700	10cd5a5210f2	Ehsan Akhgari — Merge from mozilla-central
← 4 hidden changesets [Collapse]
	cc91f155491b	Ehsan Akhgari — Bug 751611 - Add mozconfig files for building Win32 binaries on our Win64 bit platforms; r=khuey DONTBUILD
	2db9df42823d	Gregory Szorc — Bug 749957; r=rnewman
	c045085c0436	Brian R. Bondy — Bug 670514 - LNK file test for Windows share security. r=bz
	60613f18435b	Brian R. Bondy — Bug 670514 - Arbitrary File + Directory read via .lnk files on Windows Share. r=bz

no clear winner there
I'm assuming this is still a problem on Aurora.  Is this correct?  If so, we still need to figure out what fixed it so we can make sure it gets backported.
status-firefox14: --- → affected
status-firefox15: --- → fixed
Target Milestone: --- → Firefox 15
Duplicate of this bug: 753152
(Reporter)

Comment 40

5 years ago
(In reply to Matt Brubeck (:mbrubeck) from comment #38)
> I'm assuming this is still a problem on Aurora.  Is this correct? 

That's correct. (Double verified with http://mzl.la/mtd-aurora)
Reopening until this is settled.
Status: RESOLVED → REOPENED
Resolution: WORKSFORME → ---
I'm in the process of an hg bisect --extend (which includes the contents of merges). Once I figure out which patch we need to uplift, I'll mark it as such.
changeset:   92942:77bf50b33a05
user:        Nicholas Cameron <ncameron@mozilla.com>
date:        Wed May 02 14:54:40 2012 +1200
summary:     Bug 716439; compile OGL shaders lazily. r=BenWa
Depends on: 716439
Created attachment 622487 [details] [diff] [review]
attempt to port/rewrite nrc's patch

Nick's patch needed to be more-or-less rewritten to apply to Aurora, and even still it doesn't fix this bug. I'm now going to try uplifting all of bug 716439.
Uplifting all of bug 716439 (and its dependent bug 733894) fixes this problem, but actually doing that uplift and shipping it to our users is crazypants.

Still, I pushed this patch queue on top of aurora to try: https://tbpl.mozilla.org/?tree=Try&rev=354e84cfc357
(In reply to Joe Drew (:JOEDREW!) from comment #45)
> Uplifting all of bug 716439 (and its dependent bug 733894) fixes this
> problem, but actually doing that uplift and shipping it to our users is
> crazypants.

You could try working your way through the patch queue to see which one of Nick's patches actually fixed this; that might give us useful information.
Created attachment 622530 [details] [diff] [review]
disable component alpha shaders

We don't know why, but these shaders don't compile on Motorola devices. Since we don't use component alpha on GLES, just disable their content there.
Attachment #622530 - Flags: review?(jmuizelaar)
Attachment #622530 - Flags: approval-mozilla-aurora?
Comment on attachment 622530 [details] [diff] [review]
disable component alpha shaders

Splendid.
Attachment #622530 - Flags: review?(jmuizelaar) → review+

Comment 49

5 years ago
Comment on attachment 622530 [details] [diff] [review]
disable component alpha shaders

Approved, should be mobile only.
Attachment #622530 - Flags: approval-mozilla-aurora? → approval-mozilla-aurora+
Seems like we should get some kind of usable and easily diagnosable feedback if our shaders fail to compile!
http://hg.mozilla.org/releases/mozilla-aurora/rev/1faea99701db
Assignee: joe → nobody
Status: REOPENED → RESOLVED
Last Resolved: 5 years ago5 years ago
status-firefox14: affected → fixed
Component: General → Graphics: Layers
Product: Fennec Native → Core
QA Contact: general → graphics-layers
Resolution: --- → FIXED
Target Milestone: Firefox 15 → mozilla14
I have a similar issue to this one on my Motorola Photon that is still happening with http://ftp.mozilla.org/pub/mozilla.org/mobile/candidates/14.0b1-candidates/build3/unsigned/android/en-US/ which is the latest beta. In my case the pages load black instead of white. I can reproduce it just by loading about:crashes

Android 2.3.4

I can file another bug if it doesn't belong here but when I mentioned it the other day I was directed to this bug and I believe they said it would be fixed in the latest beta.
(In reply to Marcia Knous [:marcia] from comment #52)
> I have a similar issue to this one on my Motorola Photon that is still
> happening with
> http://ftp.mozilla.org/pub/mozilla.org/mobile/candidates/14.0b1-candidates/
> build3/unsigned/android/en-US/ which is the latest beta. In my case the
> pages load black instead of white. I can reproduce it just by loading
> about:crashes
> 
> Android 2.3.4
> 
> I can file another bug if it doesn't belong here but when I mentioned it the
> other day I was directed to this bug and I believe they said it would be
> fixed in the latest beta.

marcia, Joe confirmed this is bug 751732.   I'll move your comments over there and we can track discussion in that bug.   thanks for the report!
(In reply to Tony Chung [:tchung] from comment #53)

> marcia, Joe confirmed this is bug 751732.  

Correction, reference bug is bug 754253.
FYI - The Firefox beta worked perfectly on my droid 3 (which previously has had the white-blank screen issue in this bug).  

However, after we released the first beta update (the next day I think) the white-blank screen issue returned.
Also, is this bug supposed to be marked resolved? I'm happy to verify in nightly if that's where the fix is. Please let me know.

Comment 57

5 years ago
(In reply to Michael Coates [:mcoates] from comment #56)
> Also, is this bug supposed to be marked resolved? I'm happy to verify in
> nightly if that's where the fix is. Please let me know.

Is this still a problem Michael?  If so, please re-open the bug.
You need to log in before you can comment on or make changes to this bug.