20185 - absolute position elements repaint wrong w/GFX scrollbars on

Status: VERIFIED FIXED

(Core :: Web Painting, defect, P2)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

DESCRIPTION:  Pages with absolutely positioned elements (these have their own
views, I believe) are not repainting correctly.  This shows up on my homepage,
http://www.fas.harvard.edu/~dbaron/ , where:
 * scrolling down and back up causes the header at the top to disappear (and be
replaced by light gray)
 * switching to another virtual screen and back causes the entire viewport to be
light gray

A simplified testcase will also be attached, where the page in question displays
nothing.  (The correct display flashes quickly by before the screen goes blank.)

STEPS TO REPRODUCE:
 * load testcase

ACTUAL RESULTS:
 * blank

EXPECTED RESULTS:
 * the letters A, B, and C, not necessarily in that order

DOES NOT WORK CORRECTLY ON:
 * Linux, mozilla, 1999-11-28-08-M12

WORKS CORRECTLY ON:
 * builds more than a week before (I haven't tested between)

ADDITIONAL INFORMATION:

I'm marking this DOGFOOD because it prevents me from using my homepage.  It
likely affects other pages with absolute positioning (e.g., hotbot), although I
haven't checked.  (It hit 2 of 2 pages with absolute positioning that I tried.)
It's also a major regression.

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: testcase described in bug

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

ALTERNATE STEPS TO REPRODUCE:
 * load http://www.hotbot.com/
 * resize the window

ACTUAL RESULTS:
 * all of the page except for the animated GIF advertisement disappears


This might be Linux-only.  I can't tell...

Comment 3

[leger]

Putting on PDT+ radar.

Comment 4

[Patrick C. Beard]

Definitely linux-only regression. I don't see this on the Mac.

Comment 5

[Patrick C. Beard]

No ETA on this yet. I need a linux helper to diagnose this.

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Marking [PP] since this is Linux-only.

This gets fixed when I turn GFX scrollbars off.  cc:ing evaughan.  However, it
could be another bug that only showed up once GFX scrollbars were turned on...

Could it be related to bug 19142?

Comment 7

[Patrick C. Beard]

The page and test case work on Mac. Handing to pavlov to look into problems on
linux.

Comment 8

[Stuart Parmenter]

i see it with gfx scrollbars on but it is fine without them.  reassigning to
eric

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Hmmm.  I wonder if Pavlov's fix to bug 20496 could've helped this?  I guess I
can check for myself tomorrow evening...

Comment 10

[Christopher Blizzard (:blizzard)]

I think this might be the same as the bug that I've been chasing for the last
couple of days:

http://bugzilla.mozilla.org/show_bug.cgi?id=18802

Comment 11

[Christopher Blizzard (:blizzard)]

*** Bug 18802 has been marked as a duplicate of this bug. ***

Comment 12

[Christopher Blizzard (:blizzard)]

If I turn the #ifdef NEW_COMPOSITOR to an #if 0 in nsViewManager::RenderViews
this problem goes away.  Looks like a problem in that compositor code.  Also,
this test case works _so_ much better than the test cases that I had.

Comment 13

[Christopher Blizzard (:blizzard)]

Woah.  Why is this assigned to me?  I'm clueless about the compositor but I'll
try to track it down.  Someone in layout should look at it, really.

Comment 14

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Readding evaughan to cc: list.  Note Chris Blizzard's last comment.

If this isn't fixed for M12, then, IMO, you should consider turning GFX
scrollbars off by default for the Linux M12 release.

Comment 15

[Christopher Blizzard (:blizzard)]

...or use the workaround that I've mentioned above unless that causes more
problems than it solves.

Comment 16

[rickg]

This should go to eric for examination. Reassigning, and clearing fix estimate
date.

Comment 17

[Peter Trudelle]

Rick, could you comment on Blizzard's assertion that NEW_COMPOSITOR is also a
necessary precondition, and his proposed workaround?
In the meantime, assigning to pavlov, who will work with evaughan on this.
p2 for m12
If this only affects a small number of pages, why is it PDT+ dogfood?

Comment 18

[jst]

I see this on http://www.nokia.com/phones/8210/main.html, if I comment out all
<INPUT type="image" ...> on a local copy of that page the problem disappears
(no absolute position involved AFAIK)...

I see this problem only on linux with GFX scrollbars turned on.

With blizzards workarroud the page looks OK, with or without GFX scrollbars.

Comment 19

[chris hofmann]

pav sez, down deep in ugly code. hoping for early m13.

Comment 20

[Christopher Blizzard (:blizzard)]

I've seen this with my workaround so I think it's a fluke that it works at all.
Also, I've verified that it's an img submit button problem.  It makes many pages
completely unusable.  Here are some:

http://cnn.com/
http://www.abcnews.com/
http://www.etrade.com/
http://www.redhat.com/
http://www.linux.com/

This was just in 10 minutes of browsing.  I beg.  I plead.  Please fix this for
M12.

Comment 21

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Blizzard - what do you mean by "I've verified that it's an img submit button
problem."  Do you mean that image submit buttons are a sufficient condition to
cause the problem?  They certainly aren't a necessary condition, because
absolutely positioned elements are also a sufficient condition (on their own).

Comment 22

[Christopher Blizzard (:blizzard)]

I meant that I've been able to verify that it happens on pages that have img
buttons on them and the problem happens when those images come into view.  It
may or may not be related to the problem with elements.  It might be a seperate
bug.

Comment 23

[Stuart Parmenter]

*** Bug 20377 has been marked as a duplicate of this bug. ***

Comment 24

[Stuart Parmenter]

*sigh*  this bug sucks.  i'm working on it.

Comment 25

[Christopher Blizzard (:blizzard)]

*** Bug 22520 has been marked as a duplicate of this bug. ***

Comment 26

[Stuart Parmenter]

blizzard?  got any new ideas on this bug?

Comment 27

[Christopher Blizzard (:blizzard)]

Nope.  I tracked it down deep into the view manager.

Comment 28

[Hervé Renault]

i simply took a screenshot of http://www.nokia.com/phones/8210/main.html

is it really what you are discussing here ?

the same kind of grey areas appear on David Baron's homepage for me too.

Comment 29

[Hervé Renault]

Attachment: screenshot for bug 20185 as seen on http://www.nokia.com/phones/8210/main.html

Comment 30

[Braden]

Well, you don't need absolute positioning *or* an image submit button to trigger

this bug. Observe <URL:http://www.openvrml.org>. The left frame typically has

this drawing problem on the first load. The top frame grays out when I increase

the browser window width beyond a certain threshold. I'll attach a screen shot.

Comment 31

[Braden]

Attachment: Screen shot of <URL:http://www.openvrml.org>

Comment 32

[Christopher Blizzard (:blizzard)]

That has absolutely position elements.  The left frame has html in it like this:

<DIV STYLE="position:absolute; width:130px; height:53; z-index:1; left: 0;
top:   0; visibility: visible; overflow: visible"><A HREF="news.html"
TARGET="main"><IMG SRC="pics/menunews.gif" WIDTH="130" HEIGHT="53" BORDER="0"
ALT="NEWS"></A></DIV>

Comment 33

[Christopher Blizzard (:blizzard)]

*** Bug 18750 has been marked as a duplicate of this bug. ***

Comment 34

[waqar]

*** Bug 22105 has been marked as a duplicate of this bug. ***

Comment 35

[Stuart Parmenter]

beard just checked in a fix.  he rocks.

Comment 36

[Tom Mraz]

Great!!! It's fixed now!

Comment 37

[jst]

This one's back, reopening. I see the exact same problems in todays build on
linux :(

Comment 38

[Tom Mraz]

*** Bug 23115 has been marked as a duplicate of this bug. ***

Comment 39

[leger]

Adding dogfood and pp to keywords.

Comment 40

[Paul Gartside]

Why does this bug have severity=`normal'?

This is a BAD bug(s). It makes many html pages unviewable. Styling using

stylesheets instead of tables etc is ruled out (contrary to W3c). And for xml

only stylesheets are possible...



I too have this bug. (M13, 27th Jan build, i686 linux on a pc) Hope you can fix

it soon. --And THANKS!! for all your work on mozilla!

Comment 41

[Stuart Parmenter]

beard, you fixed this before, so i'm assigning it to you.

Comment 42

[sarnold]

*** Bug 21664 has been marked as a duplicate of this bug. ***

Comment 43

[Patrick C. Beard]

This requires a slight rewrite of the classic view manager to get this working 
correctly.

Comment 44

[Tom Mraz]

*** Bug 25782 has been marked as a duplicate of this bug. ***

Comment 45

[Damian Christey]

Can someone who's been working on this check bugs 23626 and 25227 (list boxes
cause gray stripe) to see if they are duplicates of this one?

Comment 46

[Tom Mraz]

*** Bug 23626 has been marked as a duplicate of this bug. ***

Comment 47

[Tom Mraz]

When this bug will be fixed? I can't use mozilla for common browsing due to this

bug - it shows up on every third page which I visit.

Comment 48

[Stuart Parmenter]

*** Bug 22936 has been marked as a duplicate of this bug. ***

Comment 49

[Stuart Parmenter]

*** Bug 22936 has been marked as a duplicate of this bug. ***

Comment 50

[Stuart Parmenter]

*** Bug 22936 has been marked as a duplicate of this bug. ***

Comment 51

[leger]

Putting dogfood in the keyword field.

Comment 52

[waqar]

*** Bug 26285 has been marked as a duplicate of this bug. ***

Comment 53

[prass]

Adding syd and myself to cc list.

Comment 54

[Christopher Blizzard (:blizzard)]

What horrible, degrading act do I have to perform and who do I have to do it to
in order to get this bug fixed?

Comment 55

[Stuart Parmenter]

heh, we're having a meeting in a minute.

Comment 56

[Sean Richardson]

*** Bug 25492 has been marked as a duplicate of this bug. ***

Comment 57

[Christopher Blizzard (:blizzard)]

*** Bug 18224 has been marked as a duplicate of this bug. ***

Comment 58

[alla]

Ok.I verified that it works for me too using viewmanager2. 
Any ideas when it is gonna be turned on by default? This bug has really annoyed
me for a long time.

Comment 59

[waqar]

*** Bug 24553 has been marked as a duplicate of this bug. ***

Comment 60

[Patrick C. Beard]

There's now a pref checked-in to turn this on by default. Warning, it doesn't 
handle opacity yet (coming soon).

Comment 61

[Sitsofe Wheeler]

*** Bug 27215 has been marked as a duplicate of this bug. ***

Comment 62

[Dan Erikson]

*** Bug 27515 has been marked as a duplicate of this bug. ***

Comment 63

[Richard Zach]

*** Bug 27657 has been marked as a duplicate of this bug. ***

Comment 64

[Stuart Parmenter]

*** Bug 27562 has been marked as a duplicate of this bug. ***

Comment 65

[Damian Christey]

*** Bug 27341 has been marked as a duplicate of this bug. ***

Comment 66

[Damian Christey]

*** Bug 24141 has been marked as a duplicate of this bug. ***

Comment 67

[Damian Christey]

*** Bug 27637 has been marked as a duplicate of this bug. ***

Comment 68

[Damian Christey]

*** Bug 22573 has been marked as a duplicate of this bug. ***

Comment 69

[Damian Christey]

*** Bug 27228 has been marked as a duplicate of this bug. ***

Comment 70

[Damian Christey]

*** Bug 27443 has been marked as a duplicate of this bug. ***

Comment 71

[Damian Christey]

*** Bug 25409 has been marked as a duplicate of this bug. ***

Comment 72

[Damian Christey]

*** Bug 27516 has been marked as a duplicate of this bug. ***

Comment 73

[waqar]

*** Bug 23120 has been marked as a duplicate of this bug. ***

Comment 74

[Damian Christey]

*** Bug 25672 has been marked as a duplicate of this bug. ***

Comment 75

[Sam Black]

Seeing this on cnn.com and pathfinder.com (in linux, build 2000021510), whcih,
ironically enough, are two of the "verification" sites listed.  This also
appears on virtually any large/long page with a lot of tables if you scroll up
and down fast enough.

Comment 76

[Patrick C. Beard]

This is fixed by nsViewManager2, which will be turned on by default for beta1.

Comment 77

[leger]

*** Bug 22907 has been marked as a duplicate of this bug. ***

Comment 78

[Sitsofe Wheeler]

*** Bug 28061 has been marked as a duplicate of this bug. ***

Comment 79

[Richard Zach]

*** Bug 28613 has been marked as a duplicate of this bug. ***

Comment 80

[jst]

Verified with todays build on linux, thinks look great now!

Comment 81

[Richard Zach]

*** Bug 28620 has been marked as a duplicate of this bug. ***

Comment 82

[Richard Zach]

*** Bug 28620 has been marked as a duplicate of this bug. ***

Comment 83

[Asa Dotzler [:asa]]

*** Bug 29587 has been marked as a duplicate of this bug. ***

Comment 84

[Sean Connor]

*** Bug 32405 has been marked as a duplicate of this bug. ***

Comment 85

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

For those who don't know already, this bug has come back, and has been
reincarnated as bug 37360 (which is nsbeta2+).