WordPress 3.5 RC6 is out. Please, if you can (and earlier the better), hammer on TinyMCE with the most ridiculous object and embed tags you can find, and anything HTML5. (Here’s a zip for the nightly build.)
What’s important in RC6? After a good first attempt last week (#22790), we took another stab (#22842) at fixing TinyMCE’s handling of, well, ridiculous object and embed tags. Our goal right now is to ensure that nothing breaks in 3.5 that worked in 3.4.2. So, go find your best embed spaghetti* and make sure nothing breaks.**
- Very easy to test: Go to the “Text “tab, paste something in, head to the Visual tab, confirm things don’t look broken, head back to the “Text” tab, see if it looks mangled, head back to Visual, confirm things don’t look broken. Remember, we are looking for regressions, so also check 3.4.2 to see if it occurs there.
** Breaks means the embed disappears in whole or part, or there’s a JavaScript error, or your computer starts smoking. Whitespace and other HTML changes do/will happen (contents may settle during shipping). Of course, your content should never be damaged, as that’s just no fun.
What happened to the last two RCs? We generally try to do a “soft” or “silent” RC at the very end of a cycle. We’re confident we’ve gotten the testing it needs, but we’d like to enter a 24-hour period where there are no more changes to trunk. Having a cleaner version number provides for a good line in the sand, and can help in case some blocker bug report comes in. Of course this time, we’ve stuttered a few times. TinyMCE hell was RC4. A few final changes on Friday (after we decided to not release) resulted in RC5. And the second round of TinyMCE hell is RC6.
This also means our new target is Tuesday, December 11. We’ll again convene at 10 a.m. Eastern to see if the winds are blowing in our direction. (Even NASA needs good weather.)
And hey, on the bright side:
WordPress 3.5 now runs a custom TinyMCE schema tailored to the HTML5 living standard. We audited the whole spec to keep your content intact.
— Daryl Koopersmith (@koop) December 10, 2012
Nick 12:06 pm on December 10, 2012 Permalink | Log in to Reply
Just updated to WordPress 3.5 RC6 and conducted the above mentioned test… Nothing broken. Looks fine for me!
Xavier Borderie 12:40 pm on December 10, 2012 Permalink | Log in to Reply
Insane work you’ve all accomplished again, guys! Congratulations and thank you!
Ryan Markel 1:16 pm on December 10, 2012 Permalink | Log in to Reply
Casual thought: we need HTML-mangle unit tests so we aren’t caught off-guard by TinyMCE weirdness in future releases.
Andrew Nacin 1:43 pm on December 11, 2012 Permalink | Log in to Reply
Absolutely. We should start thinking about real unit-testing from this angle.
Ryan Markel 3:57 pm on December 12, 2012 Permalink | Log in to Reply
I would be happy to help with this in the 3.6 cycle.
Jane Wells 2:35 pm on December 10, 2012 Permalink | Log in to Reply
I’m thinking that @nacin and @koopersmith should submit a speaking proposal about this experience to the O’Reilly Fluent Conference. Proposals are due today, conference is May in San Francisco. http://fluentconf.com/fluent2013
mordauk 4:17 pm on December 10, 2012 Permalink | Log in to Reply
Works perfectly for me so far.
Ipstenu (Mika Epstein) 4:39 pm on December 10, 2012 Permalink | Log in to Reply
Banging on it without any errors so far. I’ve even gone to test my psycho code posts on halfelf. While it doesn’t work when I switch between posts that have a mondo amount of code (like I quote a whole MU plugin), it works better than it did on 3.4.2, so (1) not a regression (2) improvement!
(The tl;dr of “If you have lots and lots of code in posts, tinyMCE is not your friend” remains, and that ain’t us, it’s them.)
paolal 5:15 pm on December 10, 2012 Permalink | Log in to Reply
Tried to break it with many embeds, but it worked flawlessly. Thanks for all the work you are doing!
michaelha 5:27 pm on December 10, 2012 Permalink | Log in to Reply
@nacin thanks for the update
tjsix 6:26 pm on December 10, 2012 Permalink | Log in to Reply
Trying out various combinations I found that if a nested tag inside a parent with attributes will be removed if it is empty, even if it has attributes of it’s own. I’ve tried this with just about every tag I could think of that would potentially be an empty element, i.e. divs, spans, i’s (for icon fonts), anchors all of which could conceivably be an element element and styled with a background image/font via css.
Robert Chapin (miqrogroove) 8:01 pm on December 10, 2012 Permalink | Log in to Reply
If it also happens in 3.4.2 it is not relelvant to this RC6. Sounds like it was working as expected!
helen 8:07 pm on December 10, 2012 Permalink | Log in to Reply
Yep, happens in 3.4.2.
Eric Hoanshelt 8:00 pm on December 10, 2012 Permalink | Log in to Reply
Looking good so far!
memuller 1:08 am on December 11, 2012 Permalink | Log in to Reply
Tested with some crazy posts from members on my WP Network. There are issues, but I’m pretty sure there are no regressions.