53972 - canvas background fixup causes weird incremental repaint

Status: RESOLVED WORKSFORME

(Core :: CSS Parsing and Computation, defect, P2)

Description

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

Something related to the canvas background fixup or body fixup rule is working
strangely in some cases.

STEPS TO REPRODUCE
 * load http://www.people.fas.harvard.edu/~dbaron/css/test/implied

ACTUAL RESULTS:
 * **some of the time**, the correct display (dark green around edges) flashes,
but is then replaced by light green where the dark green should be.  Sometimes
it works correctly.  It becomes correct again after an expose event.

EXPECTED RESULTS:
 * Light green BODY, dark green HTML (in margins of BODY and below bottom of BODY)

BUGGY IN:
 * MacOS 9.0.4 Mozilla 2000-09-22-20

Comment 1

[Marc Attinasi]

This sounds like it could be a mac-specific rendering issue (I cannot see it on 
NT or Linux, and David did report it as a mac-only bug).

Reassigning to rendeing team to investigate.

Comment 2

[Marc Attinasi]

NOTE: I just saw this on WinNT - it is really strange because I exposed the 
window which was partially obscured by my mail window, and the outter are was 
mostly light green (the newly exposed part anyway) and then I hid it again and 
re-exposed it an it repainted correctly. Also, if you resize it after the 
initial exposure it updates correctly.

I believe the problem is in the html.css file, the rule:

:viewport, :viewport-scroll, :canvas {
  display: block;
  background-color: inherit;
}

is causing the problem. If yuo remove 'background-color: inherit' it works fine. 
I believe that this is a problem with the style resolution battling with the 
body fixup rule.

Taking back now...

Comment 3

[Marc Attinasi]

Accepting and updating the platform fields, adding regression kwd.

Comment 4

[Marc Attinasi]

OK, after another 15 minutes of playing with this it is clear that those rules 
are NOT the problem. This bug is just tempermental and I was not being thorough 
enough. Please disregard my prior comment (2000-09-25 10:56).

Comment 5

[Hixie (not reading bugmail)]

FWIW, without those rules it stops using the default background colours.
Hey, if these non-standard extensions were documented somewhere, I could have
been more sure of what I was doing when I edited the rules with those strange
and interesting pseudo-classes, ok... ;-)

Comment 6

[Marc Attinasi]

David, do you have any idea when this started happening? I cannot seem to repro 
it in a build from 9/19, but easily can from a build of 9/22 (as you reported 
it). Have you seen this before 9/22?

Comment 7

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

I reported it the first time I saw it, but I didn't test that page for months
before I saw it.

Comment 8

[Marc Attinasi]

Nominating for RTM since this is really ugly and will probably show up on the
BoxAcidTest as well.

Comment 9

[Marc Attinasi]

Popping up a content-menu over the HTML are and dismissing by clicking in the
BODY will reporduce this problem quite reliably. Covering and exposing the
window will also reporduce it, but much less reliably.

Comment 10

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

*** Bug 54660 has been marked as a duplicate of this bug. ***

Comment 11

[Marc Attinasi]

CC'ing Kevin to check out. Thanks, Kevin.

Comment 12

[Pierre Saslawsky]

CCd myself because the rule above and the origin of the canvas background are of 
interest to me (bug 19329).

Comment 13

[ekrock's old account (dead)]

It's been said that this *may* show up on the Box Acid Test. Can anyone confirm
this?

Comment 14

[Marc Attinasi]

I cannot check the Box Avid Test because verso.style.com is down right now. I'll
keep trying.

Comment 15

[Marc Attinasi]

Actually, I cannot reproduce this at all anymore. I tested David and Ian's pages:

http://www.people.fas.harvard.edu/~dbaron/css/test/implied
http://www.bath.ac.uk/%7Epy8ieh/internet/eviltests/htmlbodyrendering6.html

and a modified Box Acid Test page (since the original is down) at:

http://pessoal.mandic.com.br/~aristeu/boxacidtestIE5corrected/IE5-acidtest.htm

using the context-menu technique and the random hide/show window technique and I
cannot get the bug to show.

Marking WORKSFORME - please reopen if you can still reproduce it. (I'll also
check the Box Acid Test when it is up again.)

Comment 16

[Hixie (not reading bugmail)]

I couldn't reproduce this yesterday either, but did not check this thoroughly.
I wouldn't say it was stop ship, even if it did affect the box acid test. This
is (was) a very minor bug IMHO.

BTW, Marc: the box acid test is at:  http://style.metrius.com/boxacidtest/

Comment 17

[Hixie (not reading bugmail)]

Netscape's standard compliance QA team reorganised itself once again, so taking 
remaining non-tables style bugs. Sorry about the spam. I tried to get this done 
directly at the database level, but apparently that is "not easy because of the 
shadow db", "plus it screws up the audit trail", so no can do...

Comment 18

[Hixie (not reading bugmail)]

Reassigning QA to reporter. I could never reproduce this much on the given 
pages anyway, so I couldn't really verify it.