It seems that there are some points where you are considering the size of text in line-height calculations. You seem OK on Ian's test: http://www.bath.ac.uk/~py8ieh/internet/eviltests/lineheight.html However, there are problems in http://www.fas.harvard.edu/~dbaron/csstest/linebox3 in the first and fourth paragraphs. There is too much spacing at the bottom edge of the line containing the big text. You seem to be expanding the line box to hold the text. Similarly, in the *body text* of http://www.fas.harvard.edu/~dbaron/csstest/inlinetest the appearance of an orange X increases the line height at the top, but it should not (or at least not by nearly as much as it does). (Could this, instead, be caused by your interpretation of the block-level element setting the minimum height of an inline box thing, which I think is a mistake in the spec. - I think you should ask H&kon about this). Somehow you need to go down to all *elements* within a line, but not those boxes that aren't elements. (But you still need the minimum thing for lines that don't have any elements within them. I think this should be done using an anonymous inline box that wraps the whole paragraph, but...)

This problem may also occur (but I'm not sure) in the :first-letter in http://www.fas.harvard.edu/~dbaron/sat/na/

David posted about this error in the spec back in January: http://lists.w3.org/Archives/Public/www-style/1999Jan/0027.html

I think another way of describing the problem is that you are apparently defining line-height as max(font-size, line-height) instead of just basing it on the line-height property's value (which might be relative to the font-size, but that is irrelevant). According to the specs, for a piece of text with: TEST { font-size: 6px; line-height: 1px; } The boxes are set out like this: ___________________________________ top of font-size ###### ###### #### ###### ## ## ## ## ## --##---#####---###-----##---------- line- --##---##--------###---##---------- height ## ## ## ## ## ..##...######..####....##.......... baseline ___________________________________ bottom of font-size ...and thus if the above element appeared in a line with XYZ { font-size: 1px; line-height: 1px; } ...on either side, the line should be aligned like this (note that the font-size is irrelevant): ___________________________________ top of font-size ###### ###### #### ###### ## ## ## ## ## --##---#####---###-----##---------- line- --##---##--------###---##---------- height ## ## ## ## ## .XYZ..##...######..####....##..XYZXYZ.. baseline ___________________________________ bottom of font-size ...which would make the inline boxes look like this: ###### ###### #### ###### ## ## ## ## ## --------------------------- TEST inline --------------------------- box ____ ## ## ## ## ## _______ XYZ inline ____ ## ###### #### ## _______ boxes And thus, if the surrounding lines all had: OTHERLINES { font-size: 1px; line-height: 1px; } ...then the surrounding line boxes would be placed above the TEST element's *inline box*, not above the font. This is because the line box of the TEST element's line would be like this: ------------------------------------- top of previous line box above ----------------------------------- top of preceeding line box above ------------------------------------------------- top of line box ## ## ### ## ## ## ## ## ## _XYZ__##___######__####____##__XYZXYZ____________ bottom of line box _________________________________ bottom of following line box below ______________________________________ bottom of next line box below ...thus the final rendered output would be: OTHERLINESOTHERLINESOTHERLINESOTHERLINESOTHERLINES OTHE######O######NE####ER######THERLINESOTHERLINES OTHERL##ESO##ERLIN##OT##RLI##SOTHERLINESOTHERLINES ## ##### ### ## ## ## ### ## ## ## ## ## ## ZXYZ ## ###### #### ## XYZXYZXYZXYZXYZXYZX OTHERLINESOTHERLINESOTHERLINESOTHERLINESOTHERLINES OTHERLINESOTHERLINESOTHERLINESOTHERLINESOTHERLINES I hope that was clear enough. (I also hope it was right. But if it is wrong, I am sure David will correct me now...) Note that the font-size *should* be used to decided where the padding, borders, and background are drawn. This is currently working fine, and is not related to the issue described above.

Exactly. This bug is quickly looking like bug 1990.

Incidentally, this is mainly covered by CSS2 section 10.8.1. http://www.w3.org/TR/REC-CSS2/visudet.html#q22

THe code has been revised to not consider text. Life is good...

Using 9/16 Apprunner and the http://www.fas.harvard.edu/~dbaron/css/test/inlinetest textcase, verified fixed.

Migrating from {css1} to css1 keyword. The {css1}, {css2}, {css3} and {css-moz} radars should now be considered deprecated in favour of keywords. Sorry for the spam...