719910 - White graphical anomalies in web content

Status: VERIFIED WORKSFORME

(Firefox for Android Graveyard :: General, defect, P2)

Description

[JR Conlin [:jrconlin,:jconlin]]

Attachment: corrupted screen photo

Screen noise seems sporadic. I cannot always reproduce it because I am guessing that the page image is being cached somehow.

Comment 1

[Aaron Train [:aaronmt]]

CC'ing :snorp, :cwiis. 

JR Conlin, you can take screenshots using a utility named 'ddms' found in the Android SDK [1].

[1] http://developer.android.com/sdk/index.html

Which device is that?

Comment 2

[Tony Chung [:tchung]]

I can reproduce this all the time on a recent nightly build runing on Kindle Fire.  Granted the Fire isnt a shipped device, but if you need another device to debug, let me know.

Comment 3

[JR Conlin [:jrconlin,:jconlin]]

(In reply to Aaron Train [:aaronmt] from comment #1)
> CC'ing :snorp, :cwiis. 
> 
> JR Conlin, you can take screenshots using a utility named 'ddms' found in
> the Android SDK [1].
Thanks, didn't know about that one.

> 
> Which device is that?

Droid Razr

Attaching screen cap and log dump

Comment 4

[JR Conlin [:jrconlin,:jconlin]]

Attachment: device screen capture

Comment 5

[JR Conlin [:jrconlin,:jconlin]]

Attachment: Log snippet around about:firefox url

No obvious error present.

Let me know if you want more of this log.

Comment 6

[Chris Lord [:cwiiis]]

I've seen this sort of artifacing on android devices with dodgy drivers before (Tegra 1, if memory serves) - I'm not sure if there's much we'll be able to do about it, and I don't have the hardware to reproduce it.

I used to see this on the Android Market on said device, so if it's happening during checking for updates, perhaps there's something specific to that (disk I/O during texture uploads?) that causes it?

Resetting assignee, I think somebody with the hardware should take a look at it and I have enough on my hands at the moment. If this only happens while checking for updates from the about:firefox screen on certain devices, I'm not sure I'd class it as a P2.

Comment 7

[Chris Lord [:cwiiis]]

That said, this is a device with gralloc enabled, judging by the logs - I'd be interested to know if it's reproducable with gralloc turned off...

Comment 8

[Brad Lassey [:blassey] (use needinfo?)]

(In reply to Chris Lord [:cwiiis] from comment #6)
> I've seen this sort of artifacing on android devices with dodgy drivers
> before (Tegra 1, if memory serves) - I'm not sure if there's much we'll be
> able to do about it, and I don't have the hardware to reproduce it.
> 
> I used to see this on the Android Market on said device, so if it's
> happening during checking for updates, perhaps there's something specific to
> that (disk I/O during texture uploads?) that causes it?
> 
> Resetting assignee, I think somebody with the hardware should take a look at
> it and I have enough on my hands at the moment. If this only happens while
> checking for updates from the about:firefox screen on certain devices, I'm
> not sure I'd class it as a P2.

What tegra 1 device were you using to see this?

Comment 10

[Aaron Train [:aaronmt]]

Reported Samsung Galaxy SII in bug above, bug 720946.

Comment 11

[Aaron Train [:aaronmt]]

Reported in web content too

https://bug720946.bugzilla.mozilla.org/attachment.cgi?id=591379.

Comment 12

[Chris Lord [:cwiiis]]

(In reply to Brad Lassey [:blassey] from comment #8)
> What tegra 1 device were you using to see this?

That was an Advent Vega.

(In reply to Aaron Train [:aaronmt] from comment #11)
> Reported in web content too
> 
> https://bug720946.bugzilla.mozilla.org/attachment.cgi?id=591379.

Ok, that makes it more worrying/high priority... Has this happened on any devices not using gralloc?

Comment 13

[Channy Yun [:channy]]

(In reply to Chris Lord [:cwiiis] from comment #12)
> (In reply to Aaron Train [:aaronmt] from comment #11)
> > Reported in web content too
> > 
> > https://bug720946.bugzilla.mozilla.org/attachment.cgi?id=591379.
> 
> Ok, that makes it more worrying/high priority... Has this happened on any
> devices not using gralloc?

After update of nightly today right now, this problem is disappeared from my Samsung Galuxy S2.

Comment 14

[Aaron Train [:aaronmt]]

Can we identify if a check-in fixed this issue as per Channy's word?

Comment 15

[Mark Finkle (:mfinkle) (use needinfo?)]

(In reply to Aaron Train [:aaronmt] from comment #14)
> Can we identify if a check-in fixed this issue as per Channy's word?

Not easily. We have no build id range. However, Tony and JR both can reproduce the issue too.

Tony & JR - Has the issue been fixed for you as well?

Comment 16

[Tony Chung [:tchung]]

(In reply to Aaron Train [:aaronmt] from comment #14)
> Can we identify if a check-in fixed this issue as per Channy's word?

I agree with Channy, the 1-25 build no longer reproduces the proble on the Fire.   

Going to dig through back builds now to see if i can locate a reverse regression range.

Comment 17

[Tony Chung [:tchung]]

(In reply to Tony Chung [:tchung] from comment #16)
> Going to dig through back builds now to see if i can locate a reverse
> regression range.

So strangely, the regression i found broke up to 1/19 build and starting working after 1/20 build.  Not sure if this lines up with anyone else reporting the issue.  But this is my reverse regression range to look at.

http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=58e933465c36&tochange=5c2bc94d359c

Channy, which nightly build dated were you running when you reported bug 720946?

Comment 18

[Channy Yun [:channy]]

(In reply to Tony Chung [:tchung] from comment #17)
> (In reply to Tony Chung [:tchung] from comment #16)
> > Going to dig through back builds now to see if i can locate a reverse
> > regression range.
> 
> So strangely, the regression i found broke up to 1/19 build and starting
> working after 1/20 build.  Not sure if this lines up with anyone else
> reporting the issue.  But this is my reverse regression range to look at.
> 
> http://hg.mozilla.org/mozilla-central/
> pushloghtml?fromchange=58e933465c36&tochange=5c2bc94d359c
> 
> Channy, which nightly build dated were you running when you reported bug
> 720946?

Actually I found this bug in earlier than my reporting date. (Sorry for my ignorance of thig bug.) Orignally I knew this bug in Jan, 9th. 2012 according to my first screen capture. 

https://bug720946.bugzilla.mozilla.org/attachment.cgi?id=591380
(You can see that this screen was captured in Jan 9th. in EXIT data)

Comment 19

[Channy Yun [:channy]]

Attachment: Orinial Screen Capture of this bug.

I attached my original screen capture in the first time. We have to examine push logs around Jan. 9th.

Comment 20

[Chris Lord [:cwiiis]]

This is almost certainly because of this change: http://hg.mozilla.org/mozilla-central/rev/aba1658ce66c - save some effort and try a build without and with this first.

Like I said before, has this happened on any devices not using gralloc?

Comment 21

[JR Conlin [:jrconlin,:jconlin]]

Still seeing the problem with 11.0a2 (01-27 build) on Razr

Comment 22

[JR Conlin [:jrconlin,:jconlin]]

Attachment: latest screencap with noise

Comment 23

[JR Conlin [:jrconlin,:jconlin]]

Attachment: logcat dump of latest

Comment 24

[Aaron Train [:aaronmt]]

(In reply to JR Conlin [:jrconlin,:jconlin] from comment #22)
> Created attachment 592164 [details]
> latest screencap with noise

$62 on pho .... seriously?

Comment 25

[JR Conlin [:jrconlin,:jconlin]]

Attachment: Real latest screenshot (doh)

Comment 26

[Channy Yun [:channy]]

I confirmed this bug still in 11.0a2 (2012-01-29) native version.(In reply to JR Conlin [:jrconlin,:jconlin] from comment #21)
> Still seeing the problem with 11.0a2 (01-27 build) on Razr

I confirmed this bug still in 11.0a2 (2012-01-29) native version on Samsung Galuxy SII. But, It shows when I move a screen by finger gesture in updown.

Comment 27

[JR Conlin [:jrconlin,:jconlin]]

(In reply to Aaron Train [:aaronmt] from comment #24)
> (In reply to JR Conlin [:jrconlin,:jconlin] from comment #22)
> > Created attachment 592164 [details]
> > latest screencap with noise
> 
> $62 on pho .... seriously?

I was feeling peckish.

Comment 28

[Channy Yun [:channy]]

Attachment: Reproduce in nightly 2012-02-01 build on galuxy s2

Comment 29

[Tony Chung [:tchung]]

Attachment: screenshot on aurora 2/15/2012 build

Also reproduced today. 

see screenshot.  Logcat is useless.

Repro
1) Install Aurora build (02-15-2012), Samsung Galaxy s2
2) go to mobile.twitter.com/signup
3) on the username/password sign in page, tap any of the textfields (username, password)
4) See the fuzz.

Comment 30

[Channy Yun [:channy]]

Also you can see that:
1) Go to setting and About Aurora
2) Click Check Updates
3) See the fuzz

Comment 32

[Brad Lassey [:blassey] (use needinfo?)]

STR from bug 707968 comment 0:
1. go to http://people.mozilla.com/~nhirata/html_tp/formsninput.html
2. click on the first text box
3. dismiss the vkb
4. click the second text box

and bug 707968 comment 5:
If you tap on the web page to dismiss the VKB, the rendering artifacts go away.

The Kindle Fire's VKB has a "dismiss keyboard" button (in the lower right corner). If you dismiss the VKB with the button, the rendering artifacts DO NOT go away.

I can reproduce the rendering artifacts on the Kindle Fire (Gingerbread), but NOT on the Galaxy S2 (Gingerbread), Droid Pro (Gingerbread with HW KB), or Galaxy Nexus (ICS).

Comment 33

[Mark Finkle (:mfinkle) (use needinfo?)]

We saw this on Samsung Galaxy IIs we are using at MWC. The phone updated its Android version and suddenly the white artifacts appeared on the screen in the content area. I disabled gralloc using the about:config pref and the artifacts went away.

Comment 34

[Joe Drew (not getting mail)]

Has anyone tested whether this happens on Maple builds?

Comment 35

[Cristian Nicolae (:xti)]

I tried to reproduce this issue on the latest regular Nightly Maple build, but everything works as expected. No white artifacts were noticed on any page.

--
Maple (2012-02-03)
Device: HTC Desire
OS: Android 2.2

Comment 36

[Martijn Wargers (dead)]

This might be a testcase to reproduce this issue more easily: http://people.mozilla.org/~mwargers/tests/contenteditable/contenteditable_focus_blur.htm

Comment 37

[Brad Lassey [:blassey] (use needinfo?)]

Talked to snorp about this, the suspicion is that we're touching a buffer at the same time the hardware is because don't have appropriate locking in place to prevent that. This is very much gralloc-releated and thus would not be on maple. That is further supported by the fact that turning off gralloc fixes the issue (see comment 33)

Comment 39

[Brad Lassey [:blassey] (use needinfo?)]

We expect this to be "fixed" by maple since it doesn't use the code path that looks to be the issue. Please retest once that lands.

Comment 40

[Brad Lassey [:blassey] (use needinfo?)]

Tony, can you retest now that maple has landed?

Comment 41

[Tony Chung [:tchung]]

(In reply to Brad Lassey [:blassey] from comment #40)
> Tony, can you retest now that maple has landed?

i updated to the 3-18 nightly build on the kindle fire, and tried my steps from comment 29. (showing and hiding VKB) There's no fuzz appearing anymore on this device with this build.

That said, when trying to switch between two open tabs, the screen will flicker quickly before loading the site of that tab.  But that's a different issue.

Can someone else on this thread also confirm the issue is also fixed for them?  (my samsung galaxy S2 also seems fine)

Comment 42

[Brad Lassey [:blassey] (use needinfo?)]

resolving wfm, if anyone else sees it, reopen

Comment 43

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

I don't seem to see it currently either on the Samsung G SII.  I see a little bit of flickering, but no white fuzz.