770088 - Rendering issues after updating to Aurora 15

Status: RESOLVED WORKSFORME

(Core :: Networking: HTTP, defect)

Description

[George Billios]

Attachment: wrong.png

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/20120701 Firefox/15.0a2
Build ID: 20120701042008

Steps to reproduce:

Browse forums, wiki etc eg
-http://forum.xda-developers.com
-http://www.adslgr.com/forum/forum.php
-www.wikipedia.org
-internal mediawiki

aurora version: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/20120701 Firefox/15.0a2
beta version: 14b10



Actual results:

Page is rendered distorted, simple refresh doesn't solve the issue, need to Reload (override cache) to render the page correctly.


Expected results:

Render the page correctly.

Have tried with fresh profile and only way to resolve it is to go back to v14.

Comment 1

[George Billios]

Attachment: correct rendering

Comment 2

[George Billios]

The screenshots are from an internal mediawiki. I am able to capture everything from this server (logs, packet traces etc) so please ask if you think is relevant.

Comment 3

[Loic]

Are you sure the browser cache is not corrupt?

Comment 4

[George Billios]

(In reply to Loic from comment #3)
> Are you sure the browser cache is not corrupt?

Can't be, first of all I use only memory cache and secondly I've tried more than one profiles and also cleared the cache many times during my tests.

Comment 5

[Matthias Versen [:Matti]]

Please try it on the Firefox safemode
http://support.mozilla.org/en-US/kb/Safe+Mode

Do you use a proxy ?

Comment 6

[George Billios]

(In reply to Matthias Versen (Matti) from comment #5)
> Please try it on the Firefox safemode
> http://support.mozilla.org/en-US/kb/Safe+Mode
> 
> Do you use a proxy ?

I have used safe mode with the same results. Please note that I can reproduce this with aurora(15) and nightly(16) but firefox beta (14) renders everything just fine. Everything tested with new clean profile.

I do use a proxy to access the internet sites but even with my internal mediawiki - were I don't use a proxy -  I get the results you see on the screenshots.

Comment 7

[Matthias Versen [:Matti]]

I don't have any issues with Seamonkey trunk and there must be something special on your system/network. A Proxy could be the difference...
Is your profile stored on a unusual location like a network share ?
How frequently do you get this issue ?

You screenshot looks like as if the css failed to load.
Are https and http pages affected or only https ?

Comment 8

[George Billios]

(In reply to Matthias Versen (Matti) from comment #7)
> I don't have any issues with Seamonkey trunk and there must be something
> special on your system/network. A Proxy could be the difference...
> Is your profile stored on a unusual location like a network share ?
> How frequently do you get this issue ?
> 
> You screenshot looks like as if the css failed to load.
> Are https and http pages affected or only https ?

It's a local profile and since this happens also without the proxy (on the internal server). I can replicate this issue within seconds after starting aurora and have seen it in both http/https pages.

The strange this is that I create a new profile and can replicate the issue with Aurora/Nightly but can't replicate it when I use the same profile with the beta version!

Comment 9

[Matthias Versen [:Matti]]

Can you switch network.http.spdy.enabled to false via about:config and test again ?
There is not much what I can suggest besides this, a developer have to take a look at this.
It's possible a http log is necessary and creating it should be easy if you can reproduce this within seconds.

Another option would be to find a regression range with https://github.com/mozilla/mozregression/

Comment 10

[George Billios]

(In reply to Matthias Versen (Matti) from comment #9)
> Can you switch network.http.spdy.enabled to false via about:config and test
> again ?
> There is not much what I can suggest besides this, a developer have to take
> a look at this.
> It's possible a http log is necessary and creating it should be easy if you
> can reproduce this within seconds.
> 
> Another option would be to find a regression range with
> https://github.com/mozilla/mozregression/


Nope spdy doesn't apply in this, I have followed that bug hoping that that is the issue but with no luck. I have also replicated this on another computer.

Regarding the log, you need the packet trace between the client and the server or the server log ? I can capture both. 

The regression range tool is very interesting and will probably try it.

Comment 11

[George Billios]

Seems I've found the issue after testing nightlies from different days.

The first version that I can reproduce the issue is the 2012-06-01 one built from 
http://hg.mozilla.org/mozilla-central/rev/73783bf75c4c

The 2012-05-31 version and earlier versions that I tested (2012-05-30,28,20) do not have the rendering issue, the 2012-06-01 and later versions I've tested (2012-06-06,12,16 etc.) have this issue. 

I'm not an expert but seems that bug 722034  caused this, or to be more precise the patch for that bug.

Comment 12

[Loic]

(In reply to George Billios from comment #11)
> I'm not an expert but seems that bug 722034  caused this, or to be more
> precise the patch for that bug.

It makes sens because the bad rendering you described in comment #0 is typical of broken cache system.

Did you test the nightlies with a fresh profile or your old/current one?

Comment 13

[Matthias Versen [:Matti]]

It would be nice if you could post the build revision from the last working build, too

Thanks for doing the regression search !

Comment 14

[George Billios]

(In reply to Loic from comment #12)
> (In reply to George Billios from comment #11)
> > I'm not an expert but seems that bug 722034  caused this, or to be more
> > precise the patch for that bug.
> 
> It makes sens because the bad rendering you described in comment #0 is
> typical of broken cache system.
> 
> Did you test the nightlies with a fresh profile or your old/current one?

I did with no luck, same results/

(In reply to Matthias Versen (Matti) from comment #13)
> It would be nice if you could post the build revision from the last working
> build, too
> 
> Thanks for doing the regression search !

Sure, last working nightly is from 2012-05-31 build revision:
http://hg.mozilla.org/mozilla-central/rev/3aa566994890

Comment 15

[u279076]

Seems to be a regression in Firefox 15. Do you have any third party software installed and running on your system? What hardware do you have in your computer? Could you attach a copy of your about:support page in this bug? How often does this reproduce?

Comment 16

[George Billios]

(In reply to Anthony Hughes, Mozilla QA (irc: ashughes) from comment #15)
> Seems to be a regression in Firefox 15. Do you have any third party software
> installed and running on your system? What hardware do you have in your
> computer? Could you attach a copy of your about:support page in this bug?
> How often does this reproduce?

Well your questions are very generic, of course I have 3rd party softwares installed and certainly not a hardware issue, I can reproduce it constantly, every  minutes on 2 systems. 

I will attach a copy of about:support but may ask what is the purpose of this since it is pretty clear that a patch between 2012/05/31 an 2012/06/01 caused this issue? You believe it's not the patch the problem but a combination of software/hardware etc?

Comment 17

[George Billios]

Attachment: about:support copy from affected firefox version

Comment 18

[u279076]

I'm merely trying to most accurately reproduce the conditions of your environment. In terms of hardware, I'm looking to see if you have a particular GPU/Driver which may expose this bug. In terms of software, I'm looking to see if you've installed something which affects rendering (window theming, UI additions, and the like).

I don't have much doubt that the regression range is correct. I'm merely trying to reproduce this bug as I can't right now under a fairly clean and vanilla environment. It's standard practice for us to find a locally reproducible testcase so that we can verify this when it is fixed and test for it in the future.

For example, your about:support page shows that you have an integrated Intel card so it might be GPU related since my system uses an NVidia card and my VM uses a VMWare emulated graphics driver.

Another example, your driver says it was last updated in January 2012 but checking Intel's site shows that their latest version is from May 2012. Perhaps you can try updating your driver to the latest version to eliminate that as a variable.

I apologize if my intentions were not clear enough earlier.

Comment 19

[George Billios]

Don't get me wrong, I'm pretty sure you know better tracking these bugs :) Sorry if I sounded a bit strange.

Regarding the driver, I did the update to the latest on but I still have the same issue.

Comment 20

[u279076]

If you go into about:config and type "accel" you should see three layers.acceleration prefs. What are their values?

Comment 21

[George Billios]

layers.acceleration.disabled;false
layers.acceleration.draw-fps;false
layers.acceleration.force-enabled;false

All false and this seems to be the default

Comment 22

[u279076]

Okay, so this can't be layers related then. I'll defer further comment until an Engineer has a chance to comment.

Comment 23

[Lukas Blakk [:lsblakk] use ?needinfo]

tracking, looks like it would be nice to see this regression addressed in 15.

Comment 24

[Brian Smith (:briansmith, :bsmith, use NEEDINFO?)]

I would not be surprised if the patches for bug 722034 caused this problem as we've had similar reports. The good news is that those patches were backed out of Aurora already, and we are working on fixing the related issues (e.g. bug 770243).

I am going to mark this as depending on bug 770243, as it may be a dupe.

Comment 25

[George Billios]

(In reply to Brian Smith (:bsmith) from comment #24)
> I would not be surprised if the patches for bug 722034 caused this problem
> as we've had similar reports. The good news is that those patches were
> backed out of Aurora already, and we are working on fixing the related
> issues (e.g. bug 770243).
> 
> I am going to mark this as depending on bug 770243, as it may be a dupe.

I updates today to aurora built from changeset - 98450:50963e16d1dc  and can't replicated the issue, I assume because of the backed out patches. 

Thanks for your input on this issue.

Comment 26

[Alex Keybl [:akeybl]]

Untracking for FF15 given the backout of bug 722034. Tracking bug 770243 for the longterm fix in FF16 instead.

Comment 27

[Matthias Versen [:Matti]]

marking wfm based on comment#25