211931 - Inconsistant and usually wrong table layout/interpretation, especially within the composer

Status: RESOLVED WORKSFORME

(Core :: Layout: Tables, defect)

Description

[Ian Batterbee]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4) Gecko/20030624
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4) Gecko/20030624

The main page at www.viewsonic.com.au and www.viewsonic.com.sg is rendered very
badly. One quick look at the page is worth 1000 words. I have a local cache of
the page if necessary, but as there are a lot of images, I won't attach a tgz of
it here unless requested.

The real problem, however, is not simply that Mozilla (and phoenix BTW) render
the page badly and that IE gets it "right", but that Moz renders inconsistantly,
and sometimes almost right, in the Composer.

In the navigator, it always seems to get it wrong as far as I can tell from many
many reloads.

In the composer, switching between the tabs at the bottom of the page (normal /
html tags / source / preview), without ANY change to the document source, keeps
changing the way mozilla renders the page.

For example, if I go to www.viewsonic.com.au, and observe the badly laid out
graphics on the right that are obviously supposed to build up a single image,
then I press ctrl-E to edit it, mozilla may or may not render it correctly in
the composer at that stage. It may or may not look like what you just saw in the
navigator window. If I then change to the tags view, then preview, it may or may
not get it right, but it will almost certainly be differently laid out from what
'normal' showed. Clicking back on source or tags view, then normal gives yet
another layout. There seems to be no obvious pattern as to what it will display
at any one time, nor any obvious pattern as to which images will be displayed in
the wrong place.

This has got to be duplicate, but I can't come up with the right search terms to
locate the master bug for this issue.

I guess the real problem is with the rendering engine, as phoenix suffers from
the same problem. It is odd however that the same rendering engine displays the
same source different ways if called from the composer.

Also, don't get thrown by the fact that the buttons on the left hand side of the
 page don't work in mozilla, or that there are javascript syntax errors when the
page loads. These are caused by the bare word "file:" appearing twice inside
javascript source blocks. I have reported this to their webmaster.



Reproducible: Always

Steps to Reproduce:
View how the page is rendered incorrectly. Press ctrl-E, switch repeatedly
between views. Watch how the rendering changes without reason.
Actual Results:  
It places things in the wrong places

Expected Results:  
If it's going to get it wrong, it should at least be consistant about how wrong
it is going to be.

Comment 1

[Bernd]

wfm, winxp 2003080104 only the .sg link works and one must have flash to see the
page.

Comment 2

[Ian Batterbee]

hmm... yeah.. they've changed the site completely.. it looks nothing like what
it did before.

now where did I put the local copy I made of it......

ah.. here it is.. 

Comment 3

[Ian Batterbee]

Attachment: zip of viewsonic's website when moz screws up the rendering

The page to open to see the problem is ViewSonic2.html

Comment 4

[Bernd]

Ian could you please check wether this is still a problem and if it is create a
minimized testcase (http://www.mozilla.org/newlayout/bugathon.html).

Comment 5

[Ian Batterbee]

I've reinstalled my machine a couple of times since this bug was created, but I
just downloaded the ZIP I attached to this bug a year ago, unzipped it, opened
the html, and yes, it still does it in 1.7.2 (20040803).

I also noticed that I can get it to look more or less right in the browser by
changing the font size up and down.. this makes me wonder if it is in fact a
result of bug 217527, which claims to be fixed in the main (CSV only at the mo)
trunk.

I'll see what I can come up with for a minimal test case if I have time.

Comment 6

[Bernd]

Ian, please please download a nightly before you start to work or even a 1.8.4a,
you might waiste your time and that would be bad.

Comment 7

[Ian Batterbee]

Tested with 1.8.4a - does exactly the same thing.  I have a spare linux box
which I want to trash for various reasons.. I'll install 2K on there and try it
with a completely clean install so that my prefs etc don't come into it. Gimme a
few days..  haven't started on the minimal test case yet either sorry

Comment 8

[Ian Batterbee]

Okay, here's a much more minimal example. I took the original document and
removed all the images, all the CSS, javascript, iframes, etc etc, so I think
all I'm left with is a lot of table code.

I will continue to work on this and try and find the HTML that is causing the
problem, but I think it's at the stage now where someone may click to what's
going on.

Comment 9

[Ian Batterbee]

Attachment: Fairly minimal test case showing problem occuring, with explanation at the top

Comment 10

[Ian Batterbee]

Attachment: testcase 3

Better testcase. If you shorten the line reading 'the quick brown fox...', the
problem goes away.

Comment 11

[Bernd]

the testcase is wfm winxp 20050711, Ian do you still see the problem?

Comment 12

[Johannes Buchner]

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9a1) Gecko/20060531 Minefield/3.0a1

I can confirm this big/feature for the nightly trunk.
The website http://viewsonic.com.au/ renders correctly good, but they might changed smth. in the meantime.

Comment 13

[Johannes Buchner]

Attachment: screenshot

Comment 14

[Ian Batterbee]

Thanks. Viewsonic have changed their site several times since this bug was first logged, so there's little point in checking that any more. 

Comment 15

[Daniel.S]

The zipped testcase and testcase #3 are buggy in Fx 2, but work fine since Fx 3.0

Comment 16

[Daniel.S]

Attachment: reftest

In this case, the width of the cell in the third row get doubled due to the 50% width of the cell in the first row (in Fx2).

I think this is a dupe, but here's a reftest anyway.

Comment 17

[Boris Zbarsky [:bzbarsky]]

Pushed http://hg.mozilla.org/mozilla-central/rev/60a67e590399