Sometime in early/mid-November, the layout engine was adversely affected. Unfortunately, I do not know the exact date; I do know that it works in 0.9.5 and does not work in 0.9.6 Go to the mentioned site. Notice the headings that extend from the left bar to the right. In newer builds, there is a blank area at the right end of each heading that splits the image in the table apart. This was not present in earlier builds and I cannot seem to get rid of it. It is also not present in other browsers.

Does removing the doctype declaration from the document help? This may well be a standards mode only issue.

Yes, you are correct. It only has the problem in standards mode.

Bernd, how should we be rendering this per spec?

it looks to me like a dupe of bug 22274 or at least I would wonder if this image stuff would not suffer from it.

I'm not exactly sure how it would be bug 22274 since that dealt with spaces underneath the image instead of beside the image. Also, considering how long Mozilla did render this case correctly, I don't see how it could be the old bug 22274.

Seems like the table cell counts the <BR> as 1px wide?

the one px wide <br> tag is probably a uprounding of the one twips it gets at http://lxr.mozilla.org/seamonkey/source/layout/html/base/src/nsBRFrame.cpp#189

Temporarily moving to future until a milestone can be assigned.

Not platform specific (seeing on Linux too).

The source Bernd mentioned references bug 10036, but that was fixed long before this problem ever showed up. Is there something I'm missing?

the fix for bug 10036 was a hack, and now we pay the price for this.

But what was it that actually made the hack apparent? Mozilla worked properly in this respect for over a year without the problem after that hack was entered.

nsbeta1-. Engineers are overloaded with higher priority bugs.

br{display:block;} is a work-around.

Too bad display:block severely breaks NS4.x

This patch simply removes the hack for bug 10036, which is apparently not needed anymore.

Did you run the layout regression tests, this is a must. Please report back the failures of these tests. No 'r' without 'r'test. ;-). How does the patch influence the empty table cells in quirks and standard mode?

Um, no. The fix of bug 10036 is a happy coincidence; the real reason that code is there is so that the <br> is included in the line-height calculation in nsLineLayout::VerticalAlignFrames, apparently... David, do you know what the deal is here? It looks like the nsLineLayout code would not ever trigger for <br> now (for one thing, I don't think the emptyContinuation bool will be true for a <br>). So we should be able to remove that hack safely.... Also, setting the width non-zero in both modes is a little odd if we're only going to ignore the line-height on the <br> in quirks mode...

emptyContinuation would never be true if the BR frame were ever a span (the linelayout term for a container), but it should never be that either.

Yes, I did run layout regression tests and I couldn't find anything that removing the hack would break.

Comment on attachment 87822 [details] [diff] [review] Proposed fix Well I did now. It seems that <br />'s end up with a height of zero.

nsbeta1-. Not a "stop ship" bug

mass reassign to default owner

Seems to be fixed now...