On build ID 2001070521, 2001071021, and 2001060703, the See also http://web.thock.com/mozilla/table_reflowa.html, which passed XHTML 1 strict and CSS validation. This was a rather unfun shock since I designed the layout using Moz (which, in May/June, rendered these fine). Konq does a decent rendering of it (it does not honour table attributes, the em padding, and has redraw issues, though).
Seeing on Win2K/2001070904
Could you attach the testcase to the bug, the host is unreachable for me -> no triage.
So first we have the table, which has "padding-right: 10em;" applied. No width is specified. But somehow it ends up being really wide (3-4 screens), and not showing any text. The floating banner css also seems to have been broken, in that it's no longer centred (could be an affect of the table problems). The floaty is copied from http://www.w3.org/Style/CSS/, where it works fine in 2001070521. This makes me think that the one table problem is causing others which aren't directly related.
Ohh, a new facit to this bug. Head to http://thock.com/test/?op=story;sid=140920008782 You'll see it doesn't misrender. Why? I suspect it has to do with the <pre> used in the other article. Why is the pre rendering broken?
I added a simpler testcase (#49208) of the first page. It appears to be some combination of <p> rules, table rendering, "text-indent" and "white-space: pre". (The problem with the table being too wide is not a bug at all, I'd guess -- it's only how things should be when you haven't specified a table width and are using "white-space: pre". The testcase does NOT expose this "bug"/problem.)
problem still there.
Set milestone to Moz1.1
The problem occurs with - *percentage* text-indent - nowrap or pre -moz-pre-wrap doesn't trigger this bug. It has nothing to do with paragraph rules, because the same thing happens with <div>. If you put white-space: nowrap instead of white-space: pre, putting the text directly in the td does cause the first line to render. But if you use white-space: pre, it does. The regression seems to have taken place before 2001-05-31.
As visible from the debug log for the previous testcase the line layout is completely wrong. Once the line-layout bug shown by the debug reflow is fixed, it would be good to verify that no table bug was hidden here. This is yours Marc.
Based on the source in 90297, I think these are the same bug. Marking as such.
*** This bug has been marked as a duplicate of 90297 ***