69075 - elements are forgotton to be painted

Status: RESOLVED WORKSFORME

(Core :: Layout, defect)

Description

[Wolfgang Schwarz]

This page contains two elements:

<div id="L0" style="position:absolute;top:100px;background-color:red">
 bla
 <div id="L1" style="position:relative;top:50px;background-color:yellow">
  bla
 </div>
</div>

But only one the first one is initially visible (in M0.7 and M0.8/win98). Only 
if you hide/show the browser window or otherwise force a redraw the second 
element will appear.

Comment 1

[Wolfgang Schwarz]

Attachment: testcase

Comment 2

[Marc Attinasi]

Hmm. Works for me on NT and Linux...

Comment 3

[Andreas M. "Clarence" Schneider]

I see this sometimes with Mozilla 0.8 on Win NT, but never when I copy the code
and load it locally from a file. Maybe it depends on network speed (tested with
a 33,6k modem).
When the second element is missing, a reload does never change that (tried many
times). When loading into a new window, the behavior sometimes changes. Weird.
Confirming.

Comment 4

[Andreas M. "Clarence" Schneider]

Hm, I see it now *all the time*. IIRC, the only time this worked for me was the
first time I viewed it. Related to caching?

Comment 5

[Andreas M. "Clarence" Schneider]

Copied the code to http://clarence.de/tests/69075.html (for me faster than
bugzilla, from within Netscape probably much slower).
I can now reproduce with it:
- simple reload -> works for me
- shift-reload -> no second element

Comment 6

[Wolfgang Schwarz]

I have exactly the same behaviour on clarence.de. 
But on bugzilla I *never* see the second element. 
On a local server I only get to see the element when I hit enter in the location 
field - but not when I reload or shift-reload (this shows btw that the problem 
does not depend on connection speed). When I set Cache preferences to "Never 
compare..." this is also what I see on bugzilla, but not  on clarence.de.
Viewed from file: protocol the element is always visible.
That is very strange indeed.

Comment 7

[HJ]

If you resize your browser window then that second element is displayed. Cache
has nothing to do with this, set pref to whatever you like but that won't help
you. Now the question is: is this a regression and if so what build does work
correct?

Comment 8

[Wolfgang Schwarz]

The bug shows up also in 0.7, but not in 0.6. I don't have any nightlies from in 
between but that's when this apparently came in.

As for the caching issue, I can reliably reproduce the effect I reported: If I 
load the attachment from bugzilla and hit enter in the location bar I can see 
the second element if and only if the Cache preference "Compare..." is set to 
"Never". However, on clarence.de or localhost I can't see any such effect.
Of course I don't know if this is relevant.

Comment 9

[Fini A. Alring]

I have experienced the same kind of redrawing bug that you describe,
I was trying to get my most complex dhtml app to also cope with mozilla,
and I discovered after many hours, that my layers (div/span) was forgotten by
mozilla (and Netscape 6.0), I recently fixed it by instead of hidding i do a
display:none/display:block, another temporary fix for you folks is too have a
front layer with a transparent gif, which auto resizes to fit the
innnerBrowserDisplay, and then do a hide/show on that.. this causes mozilla to
suddenly remember to draw the whole thing ;) that simple! not!.. Well don't rely
to much on the fix, rather the workaround if possible.. until this bugga is
fixed.. If needed I can create a test suite for this bug.. ;)

Comment 10

[Andreas M. "Clarence" Schneider]

Bug 75537 is similar.

Comment 11

[Marc Attinasi]

This is an incremental loading problem, for sure. I took the testase and moved
it local, and as suggested it works fine. Then I added a boat-load of <p> in
between the divs, making them display:none so the rendering would not be
changed. And, sure enough, if I add enough content between the divs then it
fails even locally.

Attaching the new testcase, which shows the bug when local.

Kevin said he'll check if this is a rendering problem or a layout problem -
thanks Kevin! If it is layout, send it to me please.

Comment 12

[Kevin McCluskey (gone)]

Looking at this bug I discovered a related problem where views are not repainted
if they do not overlap their parent views rect. see bug 76698

Comment 13

[Andreas M. "Clarence" Schneider]

Works for me with 2001-05-12-08, Win NT.
My internet connection speed has increased since my last try :) so I don't
close this bug until others can confirm it.

Comment 14

[Wolfgang Schwarz]

also works for me with 2001-05-13-08, Win 98 - and my internet connection speed
has decreased... :)

Comment 15

[Andreas M. "Clarence" Schneider]

Marking WORKSFORME.