On the url, when hovering over the text, it gets underlined. This doesn't happen on branch builds, because the parser terminates the <a>-tag directly. This doesn't happen anymore with the new html5 parser. IE and Google Chrome also don't seem to do this, but they don't underline the text either, so perhaps Mozilla should have some default css rule to prevent text inside tables to get underlined when the table is also inside an anchor?

There are a bunch of competing issues in this testcase: parsing, block-within-inline handling, and the rules for propagation of text-decoration. We have some known bugs with the last, but I don't remember exactly what the spec says for text-decoration rules.

The same parses the same way in Minefield and Chrome beta channel.

This bug is quirksmode only. Request blocking because it affects Ebay forums (bug 601425). I saw it on other sites as well.

> This bug is quirksmode only. Yes, see the "text-decoration" part of comment 1.

According to the (still member confidential, unfortunately) editor's draft of CSS 2.1, our text-decoration behavior here is correct.

The text in question in the editor's draft is the first bit of: http://lists.w3.org/Archives/Public/www-style/2010Aug/0334.html

Then again, Chromium has too much underlining on these tests (whereas we have too little): http://test.csswg.org/suites/css2.1/20101001/xhtml1/text-decoration-086.xht http://test.csswg.org/suites/css2.1/20101001/xhtml1/text-decoration-088.xht http://test.csswg.org/suites/css2.1/20101001/html4/text-decoration-086.htm http://test.csswg.org/suites/css2.1/20101001/html4/text-decoration-088.htm but still doesn't underline the text in attachment 451971 [details]. However, the actual relevant difference in our text-decoration behavior isn't the handling of inlines, it's the handling of tables. Chromium doesn't propagate text-decorations into a table from anything (block or inline) outside of it; we do (as I *think* is required by the spec).

(And I'd note the CSS 2.1 test suite has no tests testing propagation into tables.)

based on attachment 491738 [details] and attachment 491739 [details]: Differences between mozilla-central and Firefox 3.6: * propagation of standards-mode text-decoration from blocks to tables (we now underline) * parsing of <a> without an </a> (quirky handling removed) Differences between mozilla-central and Chromium: * propagation of standards-mode text-decoration from blocks to tables (we underline) * propagation of quirks-mode text-decoration from blocks and inlines to tables (we underline) Differences between mozilla-central and Opera: * propagation of standards-mode text-decoration from inlines to tables (they underline) * propagation of quirks-mode text-decoration from blocks and inlines to tables (we underline) I'd like to test IE7, IE8, and IE9, but given what I know so far, I think Opera's behavior is the most sensible, and we should make our quirks-mode text-decoration handling more quirky and (perhaps not in this bug) our standards-mode text-decoration handling follow the spec.

It looks like IE8 matches Opera (I'd need to double-check the standards-mode case, but I'm pretty sure I'm remembering the quirks-mode case correctly).

However, the original URL for the testcases for this bug and bug 562039 are in standards mode. That said, the eBay forums (bug 601425) are in quirks mode; they're probably the most important to fix.

That said, I don't see any problems on the two pages that are in standards mode (http://www.netwerknotarissen.nl/page.asp?id=10722&menu=8&kid=2 and http://www.camino.pl/ ).

I have a patch, but want to write additional tests. I'm planning to make two changes: * don't propagate quirks-mode text-decoration across HTML table elements, to fix this bug (and compat with WebKit (mostly, though it also does the same for display:table) and IE) * don't propagate quirks-mode text-decoration across inline-block and inline-table, to match spec and WebKit I'm also setting up the quirks-mode code to be usable for all modes. This means that I make the first of the above two fixes conditioned on quirks mode but the second apply to all modes, and additionally add a standards-mode check for floating and abs-pos to match the spec's propagation rules.

See comment 27 for explanation.