Created attachment 570959 [details] Testcase minimized from Facebook ticker Discovered dogfooding on Facebook with Hebrew UI.
This is the: 64 * "When a block element that does not have a dir attribute is transformed to 65 * the style of an inline element by a style sheet, the resulting presentation 66 * should be equivalent, in terms of bidirectional formatting, to the 67 * formatting obtained by explicitly adding a dir attribute (assigned the 68 * inherited value) to the transformed element." */ spec quote in html.css, right? What does HTML5 say about this?
For some values of "this is", yes. By which I mean that in principle we do the right thing -- http://dev.w3.org/html5/spec/Overview.html#flow-content-1 says that block elements should have unicode-bidi: isolate, and bug 613149 changed html.css to give it to them -- but the implementation of bidi isolation is currently buggy. I think I have a fix, currently waiting for results from tryserver.
Created attachment 572295 [details] [diff] [review] Patch This is really the same issue that I was trying to fix in the last round of changes to bug 613149, as discussed in bug 613149 comment 18-20. Reusing isFirstFrame and isLastFrame to determine whether the sub-paragraph is (or has) continuations was wrong: the correct test is on the continuations of the containing element, i.e. the element with "unicode-bidi: -moz-isolate",
Created attachment 572296 [details] [diff] [review] Reftest