The URL given is a valid, simple HTML document. I am using a "P + P" selector to indent paragraphs that are preceded by other paragraphs. All other paragraphs should not be indented. But they are. Converting the document to XML makes everything work fine: http://style.verso.com/styleserverb/cgi-bin/styleserver.cgi?doc=localdoc&localdo c_name=XML_and_Java&proprocess=tidy&composites=midnight In my opinion, if the property won't be supported properly in some cases, then it should be entirely ignored in sa

I don't see any adjacent sibling selectors used in this document (let alone any style sheet at all).

Sorry, Peter - that resource isn't very stable. I have attempted to reproduce with Mac M3, happily without success, here: http://style.verso.com/junk/adjsiblingtest.html . I am changing to fixed. Looks great - typographers and markup freaks everywhere bless you for implementing an elegant means toward traditional indented composition!Oddly, though, the final paragraph renders initially with too little top-margin until the window is resized - then it's fine. I'll report as a separate bug if it persists.

I've re-opened, as further investigation shows that the bug persists. Valid butnot-well-formed content screws up the adjacent sibling selector mechanism.The relevant CSS is p+p { text-indent: 5% } .In the simplest case, the bug doesn't appear:http://style.verso.com/junk/adjsiblingtest.htmlhttp://style.verso.com/junk/adj_a.gifHowever, once you interpose headings between runs of paragraphs, then it breaks:* http://style.verso.com/junk/adjsiblingtest_b.html* http://style.verso.com/junk/adj_b.gifThe paragraphs following the heading elements should not be indented, nor shouldany headings themselves be indented.The problem vanishes by making the content well-formed (closing paragraphs):* http://style.verso.com/junk/adjsiblingtest_c.html* http://style.verso.com/junk/adj_c.gif

This is a parser/DTD issue. The H2s in question are winding up inside the preceeding (non-closed) paragraph, so the CSS code sees adjacent paragraphs (which is what's in the content model)

Fixed by changes to CNavDTD.cpp

Fixed in April 6th Build.