This reverts the tests and amends them after the
application of commit 5da9eac08a.
Luckily nothing appears to be broken.
Change-Id: I67acfe5dc74ef750d5443dd619dbb114623ee233
The only exception is when there are new message notifications but no new
alert notifications.
Bug: 70461
Change-Id: I06daa3f7d526beeb878eb343c169e01acd49e71f
Also adds browser tests for the behaviour of the mark all as read button
to ensure it only clears message notifications.
Dependency: Ifb7b1b7b7feb4a5af65c79bb16b91a5a9c70166c
Change-Id: I46e1de229e32d705e67cebde678ecea3f3633906
IE8+ supports the pseudo element. The world will not explode if
IE6 and 7 do not have the chevron.
This simplifies code by dropping the use of the pokey element
Note: I know we can drop using the image too but I decided not to do
that in this patch.
Change-Id: Ia9ac434461e63bc2cf8554060126995ac65a3531
Danny noticed a bug where if both tabs have unread notifications,
then when opening the overlay and clicking on the alerts tab, the user
would be reverted back to the messages tab.
Test stops this from happening again.
Change-Id: I6bbbbf61251957336de8856201412fa3569ab22d
Don't mark messages as read until they have been acted on.
Show a mark as read button that marks entire list as read.
Change-Id: I4450a66cffd11c67b9a4ba9aac0fe958dc760e15
Note no design was defined so have taken this to mean use
mediawiki ui for consistency purposes.
* Use mw-ui-active and mw-ui-quiet for tabs
* Update tests
Change-Id: If7a51b2286cdfe6e839dacc2476c9a578bc7f1df
Shift to new API to support 2 tab view
When a new has no messages they will see the old style overlay with
Notifications heading. I have added tests to assure this is the case!
Later patches will:
1) Add the mark as read button only in message view
2) Note currently the tabs do not refresh when notifications is clear.
We need some kind of EventEmitter to make this sort of thing easier.
Change-Id: I62b590e81cd3fe867c4c13959cb43466aacfe2d5
Make various properties of the overlay properties of
EchoOverlay to reduce need for function parameters
Change-Id: I70ec0c650d58322aad93d71292a836d7cf7e7474
Not as scary as it looks. The change to _buildOverlay simply changes
indenting, and cleans up to reflect new parameter. Tests still pass ;-)
Kill noticationLimit concept
Only expose what's necessary on mw.echo.overlay
Change-Id: I2671a41af8188c14ad4c910396afa0d9000b6051
The badge is always close to the top of the window (if it's visible), so
this is not necessary. Also, this behavior is definitely not desired
for the upcoming fixed header beta feature (the badge will always be
visible as a part of the fixed header).
Change-Id: I66e8a50b1139f7bcd005cfb3c3576578efe6a653
Provide users a better user experience by catching and logging
errors with Echo notifications.
Bug: 60906
Change-Id: Iee93a05c6eed468af8bbfa60249df0819c49c45b
Apparently three wrongs make a right: hiding the overlay still worked.
* '#pt-notifications a' click event handler was not correctly checking
for events comig from inside the overlay due to a missing dot before
class name
* But if it were correctly checking, it would not be hiding the
overlay pokey (just its body) due to a copy-paste error
* Instead the clicks were handled by the 'body' click event handler,
which did not correctly check for events coming from
'#pt-notifications a'
I made things right again. The user-facing behavior is the same.
Change-Id: I02aba0e25ba4d81b234b327f120e0e2ff13d117c
Otherwise a new copy will be created every time the user opens and closes
the popup, and jQuery will hold on to each one forever (because it has
events bound), leaking a minuscule amount of memory.
Change-Id: I70c713be839f826fc27d07b04260c166f9052020
* In some languages like persian, the number 0 is represented as '.', we can't compare
'.' with either 0 or '0' to detect the no-notification status of the badge
* The markread API doesn't respect uselang param, it would return 0 instead of . in a url with uselang=fa
Note: we need to provide raw and formatted count in the API since client side javascript
doesn't provide fancy function like $wgLang->formatNum()
bug: 54575
Change-Id: I0a49828253ec346ed27c5b9a976f8bdff4e1fa90
We want the notifications in the flyout to behave just like links,
including standard middle-click and Ctrl-click behavior. The simplest
way to do that would be to actually make them links - but the area can
contain a few other links, so we can't do that and have to resort to
ugly hacks.
Or do we?
Turns out that while browsers won't accept HTML containing nested <a>
tags[1], such a structure is valid XHTML, and it's possible to create
such structure in HTML mode using DOM manipulation. It works like one
would expect: the entire thing is clickable, but inner <a> tags' hrefs
override outer ones.
Firefox even had a request to make that work[2] which was happily
fulfilled.
Tested the basic case [see below] on Firefox 22, Opera 12, Opera 15
(which uses the Blink engine like Chrome), IE 8 and IE 6 and it works
the same on all of them. Tested the XHTML variant [see below] on all
of the above except for the IEs which don't grok XHTML and it exhibits
the same behavior.
[1] Simple test: $('<div>1<a>2<a>3</a>4</a>5</div>').html() is
"1<a>2</a><a>3</a>45", not actually "1<a>2<a>3</a>4</a>5" like one
might expect.
[2] https://bugzilla.mozilla.org/show_bug.cgi?id=331959
----
The test cases used are below. When trying out the XHTML one make sure
that the browser uses application/xhtml+xml MIME type; saving the file
with .xhtml extension should be enough.
XHTML:
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<body>
<div>1<a href="http://google.com/">2<a href="http://example.com/">3</a>4</a>5</div>
</body>
</html>
HTML:
<!DOCTYPE html>
<html>
<body>
<script type="text/javascript">
var div = document.createElement('div');
var a1 = document.createElement('a');
a1.href = "http://google.com/";
var a2 = document.createElement('a');
a2.href = "http://example.com/";
div.appendChild( document.createTextNode('1') );
div.appendChild( a1 );
a1.appendChild( document.createTextNode('2') );
a1.appendChild( a2 );
a2.appendChild( document.createTextNode('3') );
a1.appendChild( document.createTextNode('4') );
div.appendChild( document.createTextNode('5') );
document.body.appendChild(div);
</script>
</body>
</html>
----
Bug: 52319
Change-Id: I311eca70f025ce92129c828cd88f96686b7cff72
