112410 - background-attachment: fixed does not work on tables

Status: RESOLVED FIXED

(Core :: Layout: Tables, defect, P3)

Description

[Burton Radons]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.6) Gecko/20011120
BuildID:    2001112012

The CSS style background-attachment in a style sheet is not updated while
scrolling and exhibits the smearing when scrolling onto regions that should be
updated, as described in bug 68499 (for the record, all of his examples work
properly), when the object target is a table.  The first block with a background
is a <p> block; the second is a <table> block where the style is set in the
table element; the third is a <table> block where the style is set in the <td>
element.

The URL exhibits this problem.  Selecting and deselecting some text in the
tables shows the correct position beneath the update.  Also:

http://amateur-scrolls.sourceforge.net/fixed-background-error2.html

Sorry about this example complexity, but I can't reduce it significantly without
the bug disappearing.  The second example shows the same problem (The top bar is
wrapped with <p> to do the background, the second two use tables).  However, the
new problem occurs during reload, which leaves the table background in places it
shouldn't be, with a bar of the background going from the far top-left of the
"Main Menu" to the end of it, another bar from above the top table to the end of
that (but not touching it), and a third that causes the top table's background
to extend to the left to the end of the "Main Menu" panel.  These artifacts
disappear on refresh (Switching to this window and back), and when a window is
moved over it and off again.  It sometimes doesn't appear or appears only
partially.  If the bug is not reproducible and this description is inadequate, I
can provide a screenshot.

The second bug is reproduced on all pages at
(http://amateur-scrolls.sourceforge.net).  I'll be modifying things, however, so
this is not guaranteed, but both examples are static and shouldn't change.

Reproducible: Always
Steps to Reproduce:
1. For first example, load page, then scroll around.
2. For second example, load page, then reload page.

Comment 1

[Moied]

*** Bug 131870 has been marked as a duplicate of this bug. ***

Comment 2

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

Lots of things like this don't work on tables -- I think it's a symptom of a
more general architectural problem that we have to implement these things in
multiple places.

Accepting the bug, although it probably belongs in the tables component.

Comment 3

[John Levon]

Basic found this case http://www.futurelooks.com/display.asp?i=80&p=1
which uses background-attachment: fixed on a <td> ... is this the
same bug as this, or different ?

Comment 4

[Sampo Syreeni]

Actually I'm not seeing an instance of background-attachment on this last site.
As for (my) bug 131870, I'm not at all sure that it is actually a duplicate of
this one -- there is not one table, HTML or CSS, on the example page, there. I'm
wondering whether body and/or div (perhaps when selected by body and div>body
inside a @media screen {}, as they are) also constitute special cases, or if
that one simply isn't a redraw events problem instead of a background one, per se.

Comment 5

[John Levon]

Sampo, it's difficult to reproduce, but I've seen it several times
whilst scrolling in a variety of ways.

Comment 6

[Sampo Syreeni]

What I'm saying is, http://www.futurelooks.com/display.asp?i=80&p=1 does not
link to any external stylesheet, either via <link/> or @import, and there is not
a single instance of background-attachment either in style elements or
attributes. The CSS default value for background-attachment is scroll, and I
believe Mozilla's default stylesheet does not override this. In other words, all
backgrounds on the page have background-attachment:scroll. Provided I didn't
somehow manage to miss the one crucial line...

Comment 7

[John Levon]

That page mutated since comment 3, so no wonder I had trouble
reproducing it again :)

Comment 8

[Dimitrios]

*** Bug 157324 has been marked as a duplicate of this bug. ***

Comment 9

[Eli Friedman]

Attachment: What background-attachment:fixed should do for a paragraph

I made this so everyone who sees this will know the expected behaviour of
position: fixed on a paragraph (or anything else) (the way I did it is very
wacky because transparency and position: fixed don't work together well and the
feature isn't supported.)  The source of this is irrelevent, except if somebody
cares to know what I did.

Comment 10

[Eli Friedman]

Attachment: The image I used as a background (call it "coolback.jpg")

Comment 11

[Eli Friedman]

Comment on attachment 92054 [details]
What background-attachment:fixed should do for a paragraph

I meant background-attachment:fixed

Comment 12

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

Comment on attachment 92054 [details]
What background-attachment:fixed should do for a paragraph

This doesn't seem relevant to this bug, since 'position: fixed' and
'background-attachment: fixed' are quite different.

Comment 13

[Rupert Eibauer]

I have a very reproducible way to show this bug.
It can be downloaded from

http://www.at-haunis.de/attachment-bug.tgz

I see the problem on every cell which has background-attachment set _and_ the
content does not reach the bottom of the cell! On a half-full cell, only the top
half has correct background, the bottom half is transparent. Experimenting
showed that the bottom half even does not show any border if I set it in the CSS.

The form also shows some interesting effect: the background stops at the middle
of the submit button.

I hope my example shows enough effects of the bug to make it fixable.
I was able to reproduce it on Mozilla1.4/Windows and Mozilla1.3/Linux

Please fix it soon.

Comment 14

[Boris Zbarsky [:bzbarsky]]

Attachment: Testcase

This worksforme and the original testcase seems to be gone... is there a
testcase showing this problem?

Comment 15

[Bernd]

fantasai fixed this in bug 4510.

Comment 16

[Jure Repinc (JLP)]

I'm not user if I have the same bug or not with todays Linux trunk build (Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a1) Gecko/20060415 Firefox/3.0a1) on this page:
http://www.slo-lanparty.com/

Is this the same bug and should be reopened or should I open a new bug?