Invoking encodeURI on the imageFileTitle makes sure that a "%" character is properly URL-escaped.
Bug: T337540
Change-Id: I538b7f813b546134848e93cea4744de8f4da9e90
The latest update of 'svgo' dependency includes three optimizations on
converting path commands, which
- improves closing paths and how we determine if to use absolute or
relative commands.
- round arc or convert to lines based on the geometric sagitta
- convert cubic Bézier curves to quadratic Bézier curves where possible
Also
- removing obsolete 'grayscale.svg' which was introduced in
I0f2c2b24aece3b but seems to have not been in use. It itself is not
reasonable, as an SVG filter needs an element to be applied to, which
this SVG didn't feature.
- unifying npm command to qua standard notation `minify:svg` and add
“parent” command `minify` as well.
Bug: T354875
Change-Id: Ie4d8ef85b9b9d28943a7216f0b5ed6392592a43e
Thanks to T321527, MediaWiki makes it possible for default
values of user properties to vary by user.
Depends-On: I1549c3137e66801c85e03e46427e27da333d68e2
Bug: T354329
Change-Id: Ib6717be54d4f898b8b0433ee44b57f1e3e960545
Force a content model on the title used by HooksTest, so that
calls to getPageLanguage() won't end up hitting the DB
Only happens when the Flow extension is enabled
Change-Id: Ie4769f972c26e48a216e4546a584893a8b47eb19
* The file page has a button that does not match the font size of
the page. This is also the case for the overlay.
* The icons are too big on the file page. This reduces their size.
Bug: T345749
Change-Id: I4f4edc0aa7adc0c72773a939c43108019bc4295b
User-options related classes are being moved to the MediaWiki\User\Options namespace in MediaWiki Core; reflect that change here.
Bug: T352284
Depends-On: I9822eb1553870b876d0b8a927e4e86c27d83bd52
Change-Id: If63fb03b90f797a38a07f0d88f663e25cac75770
Use the eslint rule
"no-var": "error"
in resources/.eslintrc.json and in .eslintrc.json
to require the use of `const` and `let` instead of `var`.
Bug: T337102
Change-Id: I3b3d6d426966a97c13f62494443f62bd5b790920
Skip thumbs found by the newly added selector (from 4d4e86ec) to skip
when also matched for the parsoid part of the multimedia viewer.
Bug: T348275
Bug: T349448
Follow-Up: I5ec67accab4e1cbfa90544087750016e1baab85d
Change-Id: I0c45b877565fd54b3e1cc766e4ee745a48cdea27
This test was leaving behind the black overlay, obscuring the test
results interface. Upon closer inspection, it appears not one part
of this test is doing anything that actually works.
* Assertion 1: `viewer.isOpen === false`
This is the initial and default state. Not essential for the test,
but so far so good.
* Prep for Assertion 2 & 3:
We call loadImageByTitle. One might think that this will set isOpen
to true, but, this isn't asserted. If one did, the test would fail,
because loadImageByTitle returns early from its first branch
in `!this.thumbs.length` and is thus a no-op, and adds nothing to
the test.
We set `location.hash = 'Foo'`.
* Assertion 2: `location.hash === '#Foo'`
Assertion 3: `viewer.isOpen === false`
Assertion 2 is meaningless because handling of hash changes appears
to be asynchronous. Perhaps not always, but at least the indirection
used in this codebase makes it async. Hence, the fact that it is
correctly normalised to '#Foo' is meaningless as the popstate
and hashchange event handler haven't been processed yet. If they
were, it would still be meaningless since viewer.isOpen was never
true in the first place and MMV's event handlers for popstate/hashchange
are guarded by `isOpen`.
* Prep for Assertion 4: location.hash = Bar
* Assertion 4: `location.hash === '#Bar'`
Idem.
* Prep for Assertion 5: Replace the viewer.loadImageByTitle function
with a function that performs a nested assertion.
This function is never called and its assertion never reached.
This is a textbook example why assertions should never be
nested. If you rewrite this in following with the advice from
<https://api.qunitjs.com/assert/verifySteps/>, we find that
the array is in fact empty by the end of the test() function.
```js
var seen = [];
viewer.loadImageByTitle = function ( title ) {
// assert.strictEqual( … );
seen.push( title.getPrefixedText() );
};
location.hash = …;
location.hash = …;
location.hash = …;
location.hash = …;
assert.deepEqual( seen, [ 'File:' + imageSrc' } );
// Actual: []
```
== Misc notes ==
The test ends with "#/media/File:Foo bar.jpg" set on location.hash,
and the black overlay covering the page. It seems that the hashchange
are async, and that the first in a given event loop tick "wins", with
later ones being ignored. Observing this with debugger/breakpoints is
hard given that each pause results in the effect not being observable.
console.log(location.href) at the end of the test shows hash='#',
but placing a debugger at the end of the test results in the hash
being "#/media/…" instead.
Exprienced in Firefox 117 on macOS.
Change-Id: Ib37bec1b3e67fcab1da89d23381a3adbb6312827
