STEPS TO REPRODUCE 1. load http://www.cyclingnews.com/ (wait until it's fully loaded, it takes a while) 2. click Reload and while the page is loading type CTRL++ (increase text size). Hitting CTRL++ when a light blue box (containing "World Times") is rendered increases the chances of reproducing the bug. ACTUAL RESULT Most of the page is missing, only the top and left parts are rendered. Reproducible about 50% of the time. Select all (CTRL+A) and pasting to an editor results only in the text that is rendered. EXPECTED RESULT All content visible, with increased text size. PLATFORMS AND BUILDS TESTED Bug occurs in SeaMonkey 2007022001 on Linux. Seems to be a recent regression (ie. not reflow branch landing) but I could be wrong.

It's much harder to reproduce in a debug build. There are a few warnings and assertions on loading this page, also when the bug doesn't occur. When the bug finally occurred, there was this additional warning: WARNING: constructing frames in the middle of reflow: 'mPresContext->mLayoutPhaseCount[eLayoutPhase_Reflow] == 0', file ../../dist/include/layout/nsPresContext.h, line 919

Bug occurred again at http://www.hexus.net/ CTRL+R followed almost directly by CTRL++ rendered a blank page except for the background image. SeaMonkey 2007022001 on Linux. Also reproduced the bug at both URLs in Minefield 20070206 Linux.

Maybe you can reproduce it locally when strategically inserting this script somewhere in the page? netscape.security.PrivilegeManager.enablePrivilege("UniversalXPConnect"); var navigator1 = parent.QueryInterface(Components.interfaces.nsIInterfaceRequestor).getInterface(Components.interfaces.nsIWebNavigation); var docShell = navigator1.QueryInterface(Components.interfaces.nsIDocShell); docviewer = docShell.contentViewer.QueryInterface(Components.interfaces.nsIMarkupDocumentViewer); docviewer.textZoom += 0.2;

Because of the use of enhanced privileges, you need to download the testcase locally. This testcase gives a regression range between 2006-12-07 and 2006-12-08, which makes it a regression of the reflow branch landing. I guess this might be the same issue as Mats is seeing, not sure though.

Should retest once bug 363253 is fixed, given the stack I see to the frame construction reentry stuff.

(In reply to comment #5) > Should retest once bug 363253 is fixed, given the stack I see to the frame > construction reentry stuff. Could you retest?

Testcase is now worksforme, using: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9a4pre) Gecko/20070417 Minefield/3.0a4pre I guess this was fixed by bug 363253 then.