After patch:

Looks like the reference wrapper is going breaking Parsoid's about grouping:

The easiest fix would be to move #mw-page-header-links above #firstHeading and then visually re-order then with CSS.

Not to do with translate. VE expects navigation tabs to come before firstHeading, in Timeless they are after so that get treated like a sub heading and get greyed-out and disabled.

Sounds like a slugging issue. Pinging @dchan

And what about the ellipsis approach? (which browsers use for multiple tabs...)

Do we have to support this? Button groups neither wrap nor scroll, you are just supposed to design your interface around that :)

Sure, whatever the services team is happy with (@ssastry, @GWicke?). Also https://www.mediawiki.org/wiki/Content_translation/cxserver/Setup should be updated.

The speed of the animation may also be a factor. Looking at that page with the fast pulsing makes me feel agitated :) The pulsing on the education popups is much more pedestrian.

Doesn't sound like a problem with VE, just the user doing what they shouldn't.

Looks like a regression from the preload patch. Pinging @DLynch

Wouldn't it be better to show labels and icons in the wide version, that way users can become familiar with the meaning of the icons should they then go on to use a smaller screen size?

The above patch fixes symptom but not the cause. Pinging @dchan

I think this is a dupe/side-effect of T53260

VE Mobile is currently using context-sensitive toolbars, similar to Google Docs mobile, such that when the document is not focused you see navigation and document-level tools:

I think TreeModifier could be to blame, emitting more events than we used to and causing the list to rebuild many more times. Pinging @dchan .

The above patch fixes the symptom if not the cause.

The problem is they have started using the <references> tag again, which is a good thing because it means the ref list updates dynamically. Unfortunately the large wikis have been using {{references}} templates for so long that we haven't noticed the performance issues.

Presently you can add a link to the source text

regression introduce in TreeModifier commit

What is the purpose of this lib? Why not use $(window).scrollTop()?

Chrome 61 is now released as stable, and this breaks MV badly.

We should separate rebaser server and client code, and ensure there is no ES6-ness in the client. ve.dm.TestRebaseClient.prototype.deliverOne also contains ES6 code. Pinging @dchan

In T57370#3601942, @Esanders wrote:

do our own memory management

In T57370#3416020, @Jdforrester-WMF wrote:

silently fail when the user's machine either has local storage disabled or is full?

Probably fixed at some point in the last 2.75 years

The patch will only work on the very latest versions of Chrome (>=60), and MobileSafari (10.0 with WebKit >603). On any other browser it is basically impossible to know that the user requested an 'undo'. So the best we can do (and we already do) is observe the change the system makes, and record the transaction.

Unfortunately this doesn't fire a reliable event. In Chrome it looks like you get event.inputType = 'historyUndo' / 'historyRedo', but in FF you get nothing (and I imagine IE).

Should we be reverting TreeModifier again before release then? @Deskana @Jdforrester-WMF

Pinging @DLynch

Basically a duplicate of T57370

This is T114531 or T78550

Interesting, the moz documentation doesn't mention the prefixed value: https://developer.mozilla.org/en-US/docs/Web/CSS/user-select

Bisect blames https://gerrit.wikimedia.org/r/#/c/363619/1

If putting display:block inside display:table is going to trigger mystery browser bugs we should definitely avoid it. So I'd suggest we revisit that decision on the figure DOM structure, and attach the zoom icon back to the figcaption. IIRC the only downside to that was that the icon wouldn't show if the caption was empty, but that could either be ignored or worked around.

Looks like this is a regression caused by https://gerrit.wikimedia.org/r/#/c/366898/ (pinging @Arlolra), which changed figcaptions to be display:block inside display:table, instead of display:table-caption. We discussed the invalidness of this at the time, but it didn't appear to be problematic. Now it does...

Looks like a browser bug, if we don't set -moz-user-select: none on ve-ce-focusableNode * and then reset it to text on ve-ce-activeNode it works fine. Strangely it works in standalone, where the CSS is very similar.

Interestingly it works fine in the standalone demo, so must just be some MW weirdness...

Can you link to a page? That looks a lot like a browser bug, so there may be nothing we can do...

The difficulty here is that changes to the fields update the model instantly, not just when 'done' is pressed, so there would need to be a significant refactoring to support a 'cancel' option.

It's provided by the TemplateData extension, written VE and former-VE members to support VE's template editing.

Pinging @Catrope.

Sounds like bad wikitext then, close as invalid?

There is a very old bug around redesigning the media dialog somewhere. There were also some technical reasons for keeping the caption if it can still appear in the wikitext (even if it doesn't render), but I agree it is odd.

The list group doesn't style properly when disabled.

I'm not sure it's possible to tell the parser to render the page as it was on a certain day. A related issue would be that templates used change independently of the article, as even parser logic/installed extensions may have changed, so any rendering of an old version is actually today's rendering of that day's wikitext, it isn't generally possible to recover the HTML on a specific day.

Here's mine:

[
	{"imeIdentifier":"","userAgent":"Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:55.0) Gecko/20100101 Firefox/55.0","startDom":""},
	{"seq":0,"time":2.434,"action":"sendEvent","args":["keydown",{"keyCode":17}]},
	{"seq":1,"time":2.437,"action":"endLoop","args":[]},
	{"seq":2,"time":2.443,"action":"sendEvent","args":["keydown",{"keyCode":16}]},
	{"seq":3,"time":2.444,"action":"endLoop","args":[]},
	{"seq":4,"time":2.895,"action":"sendEvent","args":["compositionstart",{}]},
	{"seq":5,"time":2.897,"action":"changeText","args":["u"]},
	{"seq":6,"time":2.897,"action":"changeSel","args":[1,1]},
	{"seq":7,"time":2.897,"action":"sendEvent","args":["input",{}]},
	{"seq":8,"time":2.898,"action":"endLoop","args":[]},
	{"seq":9,"time":3.738,"action":"changeText","args":["u0"]},
	{"seq":10,"time":3.738,"action":"changeSel","args":[2,2]},
	{"seq":11,"time":3.738,"action":"sendEvent","args":["input",{}]},
	{"seq":12,"time":3.74,"action":"endLoop","args":[]},
	{"seq":13,"time":3.922,"action":"changeText","args":["u00"]},
	{"seq":14,"time":3.922,"action":"changeSel","args":[3,3]},
	{"seq":15,"time":3.922,"action":"sendEvent","args":["input",{}]},
	{"seq":16,"time":3.927,"action":"endLoop","args":[]},
	{"seq":17,"time":4.218,"action":"changeText","args":["u00e"]},
	{"seq":18,"time":4.218,"action":"changeSel","args":[4,4]},
	{"seq":19,"time":4.218,"action":"sendEvent","args":["input",{}]},
	{"seq":20,"time":4.221,"action":"endLoop","args":[]},
	{"seq":21,"time":4.386,"action":"changeText","args":["u00e9"]},
	{"seq":22,"time":4.386,"action":"changeSel","args":[5,5]},
	{"seq":23,"time":4.386,"action":"sendEvent","args":["input",{}]},
	{"seq":24,"time":4.389,"action":"endLoop","args":[]},
	{"seq":25,"time":4.698,"action":"changeText","args":["<br>"]},
	{"seq":26,"time":4.698,"action":"changeSel","args":[0,0]},
	{"seq":27,"time":4.698,"action":"sendEvent","args":["input",{}]},
	{"seq":28,"time":4.701,"action":"sendEvent","args":["compositionend",{}]},
	{"seq":29,"time":4.702,"action":"sendEvent","args":["input",{}]},
	{"seq":30,"time":4.703,"action":"sendEvent","args":["compositionstart",{}]},
	{"seq":31,"time":4.704,"action":"changeText","args":["é<br>"]},
	{"seq":32,"time":4.704,"action":"changeSel","args":[1,1]},
	{"seq":33,"time":4.704,"action":"sendEvent","args":["compositionend",{}]},
	{"seq":34,"time":4.707,"action":"sendEvent","args":["input",{}]},
	{"seq":35,"time":4.711,"action":"endLoop","args":[]},
	{"seq":36,"time":4.809,"action":"sendEvent","args":["keyup",{"keyCode":13}]},
	{"seq":37,"time":4.811,"action":"endLoop","args":[]}
]

1. Can you go here: https://doc.wikimedia.org/VisualEditor/master/lib/ve/demos/ve/eventLogger.html
2. Click "Start Logging..."
3. Type your unicode sequence
4. Click in the giant text area and copy the contents, then paste here.

I can't reproduce in Ubuntu 16.04, Chrome/Firefox. Sounds like a broken IME implementation in that distro... :/

IME issue for @dchan

Yes, let's, before people starting using it!

Looks like the buttons got wider and so ran out of room on one line.

...and @Volker_E

Paste works fine for me - can you give more precise steps with the exact way you copied the text, if reproducible?

So it turns out CodeMirror has a CE mode as well (inputStyle: 'contenteditable'). They still disable spellchecking, and when I override that it seemed to work okay, although it moved the cursor to the wrong place after using it. It probably handles RTL and IMEs better, so it seems like a better system.

The VE surface was designed to use all the built in features of ContentEditable (such as spell check and RTL support) and @dchan has spent a lot of time making sure that when you use these features it synchronises directly to the model. I can't envision CodeMirror addressing these problems any time soon, but the way VE integration works is we use the VE surface as an invisible input, and only use CodeMirror for colouring, by rendering underneath VE.