1) Load URL 2) Watch carefully as page is loaded and drawn once... 3) ...then goes blank and is loaded and drawn again. May require a semi-slow Internet connection. Extremely annoying when browsing Bonsai, for instance, although I think I can reproduce this without involving large tables. Trunk-only regression that started sometime in the last month.

I have seen something like that (a sort of flashing midway through page loading) on Bonsai pages or long Bugzilla search-result pages (even with a fast pipe), but not frequently. But I can't reproduce anything like this now (with a build that is few hours old). Lots of layout related checkins in the past 24hours. (shouldn't that be a core bug, btw ?).

I can still reproduce this with 2007110901, although to see it you need to be watching very carefully on a fast Internet connection. It's much more obvious if you're on a connection that gives a max speed of, say, 20K/sec or less.

Chris, do you have View Menu > Text Encoding > Automatically detect page encoding turned ON ? While loading this short query: http://bonsai.mozilla.org/cvsquery.cgi?module=Camino&branch=HEAD&cvsroot=%2Fcvsroot&date=day&who=karlt%2B%25karlt.net I noticed a reflow of the table. On the first row there is a '-' hyphen, next to the bug number. When loading first, the character was displaying incorrectly. then a brief sort of 'reload'(*)/reflow, and the character displayed correctly. (*) not really a reload, rather reparsing. I could see the table briefly expand/contract. With 'Automatically detect page encoding' disabled, this does not happen. Sure enough, with the URL in the url field, the same thing happens.

This is probably a better, more permanent URL: http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=Camino&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=karlt%2B%25karlt.net&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2007-11-11+20%3A30%3A00&maxdate=2007-11-11+20%3A50%3A00&cvsroot=%2Fcvsroot

Turning off autodetect does seem to fix it on the URL I originally gave as an example, but not on the URL provided in comment 3 (which seems to be too short to trigger the bug for me, at least on the connection I'm currently using). That'd explain why clearing the cache seems to be required between pageloads to re-trigger the bug, too. cl

Heh. This is a dupe of a *really* old Core bug, I think. Why we're not seeing it on branch is beyond me.