477236 - single pixel horizontal white lines in iframes when scrolling

Status: RESOLVED FIXED

(Core :: Graphics, defect, P1)

Description

[Kevin Brosnan]

Attachment: Flash ad displaying the bug

Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.2a1pre) Gecko/20090205 Minefield/3.2a1pre

1) load the url above
2) scroll down on the page
3) as the flash ad appears white lines run through the ad 250x300 at the end of the story

The site does not always show flash ads. Flash ads with motion don't exhibit the issue. Does not appear in the 1.9.1 branch. Does not appear if you load a problem page in the dom inspector.

Working on a regression date.

Comment 1

[Ria Klaassen (not reading all bugmail)]

I also still see the lines in the locationbar autocomplete.

Comment 2

[Ria Klaassen (not reading all bugmail)]

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2a1pre) Gecko/20090206 Minefield/3.2a1pre

Sorry, FWIW, with the latest build I see no lines anymore.

Comment 3

[Kevin Brosnan]

Looks to work in Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.2a1pre) Gecko/20090114 Minefield/3.2a1pre

Fails in Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.2a1pre) Gecko/20090115 Minefield/3.2a1pre

Comment 4

[Kevin Brosnan]

Still see the issue with Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.2a1pre) Gecko/20090206 Minefield/3.2a1pre

It helps to hold down the middle mouse button to scroll to produce the number of white lines shown in the top of the attached image. The last white line in the image is from rolling the scroll wheel one click.

Comment 5

[Kevin Brosnan]

Change sets involved http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=89811ac1b35a&tochange=73fbb12f48e1

Comment 6

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Just to confirm before I panic -- this is not happening in 1.9.1, correct? Just 1.9.2?

Comment 7

[Kevin Brosnan]

That is correct. Only 1.9.2 builds show the issue.

Comment 8

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Thanks; from looking at the changeset range, the one that stands out is bug 448830.

Comment 9

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Yeah, it could be 448830.

Comment 10

[Jeremy Lea]

I think I managed to generate the bug once without flash.  On the square ads below the article.

Looking at one example with a flash ad, it had a 300x250 iframe with the flash object inside that.  White lines always appear at the edge of the screen after scrolling, so that means the rectangle for either the object or the iframe is being clipped to the client area and then rounded in such a way that it is coming up one pixel short.

I've looked over the patch from bug 448830 and I can't find anything.  All the major cases of changes in rounding are either not on Windows or only change the cursor, I think...  The minor cases are all float rounding, where the math should be more accurate now.

I did notice the page generating a lot of assertions for "computed overflow area must contain frame bounds".

Comment 11

[Robert O'Callahan (:roc) (email my personal email if necessary)]

That assertion is quite bad. We should probably track it down. A reduced testcase would be useful.

Comment 12

[Martijn Wargers (dead)]

Attachment: testcase

When scrolling down, I get white lines.
I tried to make a testcase with drawWindow canvas to capture the painting bug, but I didn't succeed.

Comment 13

[Robert O'Callahan (:roc) (email my personal email if necessary)]

OK, I see a white line at the bottom of the window on Mac.

Comment 14

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Thanks!

Comment 16

[Martijn Wargers (dead)]

I'm also seeing this bug at http://alleskanhier.blogspot.com/

Comment 18

[Justin Dolske [:Dolske]]

FWIW, I don't think iframes or scrolling are required here. I encountered this problem in bug 477519 with just a scaled image. No iframe, and scrolling actually made the glitches disappear.

Comment 19

[Jeremy Lea]

I suspect that bug 477519 is not the same bug, although it was also exposed by bug 448830.  When I can find some time to look at this, this bug will be a rounding problem with views which are not pixel aligned, but which bug 448830 didn't fix.  bug 477519 will be a similar problem but with the computation of the damage rectangle for a scaled image.  Both will be computations relying on an inner and outer rectangle being done in exactly the same way, and one is using height = round(height/scale) and the other is now using height = round(height+y)/scale-round(y/scale) (for some value of round).

Comment 20

[Mike Beltzner [:beltzner, not reading bugmail]]

Why do we think this is a 3.6a1 blocker?

Comment 21

[Timothy Nikkel (:tnikkel)]

I think this should be fixed by bug 503814 (at least, if not roc's compositor landings), but I couldn't reproduce using the testcases in this bug so I didn't mark the dependencies. Anyone able to verify this is fixed?

Comment 22

[Rob Arnold [:robarnold]]

I could not reproduce the issue with the testcase using my build from this morning on Windows 7. I also checked the url and none of the many flash ads had lines when I scrolled.

Comment 23

[Timothy Nikkel (:tnikkel)]

I was hoping someone who could reproduce in this bug before could check if they can reproduce now, since I was never able reproduce on the testcases in this bug, but could reproduce what sounded like the same problem on my own testcase.

Comment 25

[Mike Beltzner [:beltzner, not reading bugmail]]

Mass change: adding fixed1.9.2 keyword

(This bug was identified as a mozilla1.9.2 blocker which was fixed before the mozilla-1.9.2 repository was branched (August 13th, 2009) as per this query: http://is.gd/2ydcb - if this bug is not actually fixed on mozilla1.9.2, please remove the keyword. Apologies for the bugspam)