Commit graph

13 commits

Author SHA1 Message Date
Timo Tijhof 4ca4febd89 tests: Adopt private require() for skins.minerva.scripts (take 2)
Resubmission of Ib68f45d93a (548e94da98), which caused an error due
to `require('../../`)` not working outside the module base directory.
This is fixed with regression test in T373065.

Further confirmed in this repo by viewing
`/wiki/Special:CreateAccount?campaign=loginCTA&useskin=minerva`
with this patch applied. Without the T373065 fix, the console warns:
> Error: Cannot require undefined file …/ToggleList.js

-------

Support for private require() in tests is available since MW 1.41:
https://mediawiki.org/wiki/RL/PKG#Private_export_and_import
https://mediawiki.org/wiki/RL/PKG#Virtual_files_in_traditional_modules

This fixes a confusing assertion in page-issues/index.test.js,
where for "insertBannersOrNotice()" it was asserting that the HTML
contain "⧼skin-minerva-issue-learn-more⧽", where the ⧼ character
indicates the message is not found (i.e. an error).

The test had to be written this way in order to pass, because
the skins.minerva.scripts module was not actually loaded, and thus
its templates and messages are not present either. This lack was
filled in by index.js for mw.templates, but not mw.messages.

By adopting private require(), these workarounds can all be removed.

== Motivation ==

In change I3a4024ccf90e505581, I'm working on improving the testrunner
config to enforce uselang=qqx on all tests. This is passing except
for GrowthExperiments and Minerva, both of which have the above
workarounds in place that caused a message to be undefined, and then
kept in the assertion expectation. When using uselang=qqx, values are
returned as (key) instead of ⧼key⧽, which exposes these message
existence errors.

By removing this workaround, the test will simply import the module
in the test as normal, thus the messages will exist, and thus it
will expect (key), and thus it will continue to pass even after
enforcing uselang=qqx.

Bug: T373482
Depends-On: I777f2c12e845a738edeac00e19818d4c939a1ef1
Change-Id: I6348e5ae00776c2c01d91fa90a431b6ffe3da25d
2024-08-29 20:34:16 +00:00
Jdrewniak dbe79167e5 Revert "tests: Adopt private require() for skins.minerva.scripts"
This reverts commit 548e94da98.

Reason for revert: The patch was causing the following error on pageload

    Error: Cannot require undefined file includes/Skins/ToggleList/ToggleList.js
    require startup.js:1006

Bug: T373482
Change-Id: I22751109c124795af7d2e0ad6c4104745b2d6632
2024-08-27 20:06:24 +00:00
Timo Tijhof 548e94da98 tests: Adopt private require() for skins.minerva.scripts
Added to MediaWiki core last year with I9fca9fdf9b7623b1.

This fixes a confusing assertion in page-issues/index.test.js,
where for "insertBannersOrNotice()" it was asserting that the HTML
contain "⧼skin-minerva-issue-learn-more⧽", where the ⧼ character
indicates the message is not found (i.e. an error).

The test had to be written this way in order to pass, because
the skins.minerva.scripts module was not actually loaded, and thus
its templates and messages are not present either. This lack was
filled in by index.js for mw.templates, but not mw.messages.

By adopting private require(), these workarounds can all be removed.

== Motivation ==

In change I3a4024ccf90e505581, I'm working on improving the testrunner
config to enforce uselang=qqx on all tests. This is passing except
for GrowthExperiments and Minerva, both of which have the above
workarounds in place that caused a message to be undefined, and then
kept in the assertion expectation. When using uselang=qqx, values are
returned as (key) instead of ⧼key⧽, which exposes these message
existence errors.

By removing this workaround, the test will simply import the module
in the test as normal, thus the messages will exist, and thus it
will expect (key), and thus it will continue to pass even after
enforcing uselang=qqx.

Change-Id: Ib68f45d93a7054ed8bd35fc5644e2852f2f90248
2024-08-23 19:04:36 +00:00
Ed Sanders a6ab8d6da3 Prefer arrow callbacks (ESLint autofix)
Change-Id: I52b2feacd6216e99e04f193ba963e897b3e1a771
2024-06-04 08:01:54 -07:00
Fomafix c43cdce48c Some small JavaScript/jQuery improvements
Change-Id: I33feb8bccbd870f589a01dc187e6155b7498714c
2024-04-08 19:45:22 +00:00
Fomafix effd4fe133 Use eslint rule "no-var": "error" and replace all var
Change-Id: I67acf88e1b8de55054248d7cf8ca622d5772ea6f
2024-03-29 07:40:22 +00:00
Jdlrobson ca28efc9c7 Drop mw.mobileFrontend references
This restores the previously reverted patchset
If5b76245bf60bfa9cf977cdbf37ee0d6bb65f9d9

Changes since original:
* Added Depends-On to MobileFrontend
* Uses OOUI classes for page issues rather than es6 classes - ES6
classes do not support modifications to class prior to running
super so MobileFrontend's View class is not compatible without
significant refactors.

Depends-On: I24ad75adf8519102ca356d64d99d765ab69180cc
Bug: T348807
Change-Id: I4ff82af0251254c846f2caee330af5af738f6029
2023-11-10 01:30:52 +00:00
Jdlrobson 2ee3e27406 Revert "Drop mw.mobileFrontend references"
This reverts commit 19ea6328b0.

Reason for revert: Breaks page issues and image overlay. I
will break this up into smaller less risky patches.

Change-Id: If5b76245bf60bfa9cf977cdbf37ee0d6bb65f9d9
2023-11-03 19:26:16 +00:00
Jon Robson 19ea6328b0 Drop mw.mobileFrontend references
Bug: T348807
Change-Id: Ib71b43005e5788d0e29917a165281faa84926414
2023-10-31 21:54:26 +00:00
Jon Robson d407d1054c Remove or document remaining mw-ui-icon usages
Most of these are captured in other bugs
The anon talk message style doesn't do anything.

Bug: T344022
Change-Id: Iba1fce98ea05931fe07b2b20f07c62d198039d1b
2023-09-13 00:03:38 +00:00
Timo Tijhof 5a65bd5200 tests: Minor clean up of in TitleUtil/UriUtil tests
* Declare variables inline, as per the current code conventions.

* Use built-in assert.true() and assert.false() in a few places.

* Use built-in QUnit.test.each() to remove need for ad-hoc loops and
  inline composing of assertion messages with common prefixes. This
  also creates clearer and more detailed test reports, and more granular
  ability to re-run specific test cases.

* Remove unneeded use of `QUnit.newMwEnvironment()`.

* Simplify restoring of mw.Uri by storing the original reference
  once outside the test, and then re-using that each time.

* Use mw.config.set() for improved familiarity and rely on natural
  restoration instead of the extra 'config' key abstraction which is
  another thing to learn and understand.

Change-Id: I796e034854203d2e0e78e510458f4b34603e9901
2022-05-15 00:46:28 +00:00
jdlrobson c0f08790ea Remove the mw-ui-icon hacks and overrides
Note: this agitates T230232 again - when merging this please
make sure a merge for I929090848f3e04647a97f4979ec78682623fa070
is pending.

In various places we try to override the default mw-ui-icon behaviours
The hacks need to be removed as part of addressing the core problem.

Changes:
* Wherever we use mw-ui-icon-before in PHP - wrap the label with a span
so that label font-size is altered where needed - not the icon
* Where a small icon is needed us isSmall parameter for the Icon component
* Apply font-size to labels of mw-ui-icon-before elements
* The browser tests need a slight update to access the span element inside
a menu item - in the case of the logout button the label is always hidden,
so we need to check the visibility of the parent element (secondary_action)

Bug: T229440
Depends-On:  I3f803ec4c9068b30aa93b803391aa4d65d8310ff
Change-Id: I07e4ae233979636b739f1117dd7703571e0a9366
2019-09-10 10:53:20 -07:00
jdlrobson 72df451bd3 Embrace packageFiles
Help with readability by using module.exports and require rather than the MobileFrontend
provided mw.mobileFrontend module manager (and avoid adopting webpack at this time)

Replace usages of mw.mobileFrontend.require with local require and module.exports
(compatible with RL or Node implementation)

Changes:
* Notifications modules are merged into skins.minerva.scripts and initialised
via a client side check.
* new file overlayManager for exporting an overlayManager singleton
rather than being hidden inside resources/skins.minerva.scripts/init.js
* All M.define/M.requires swapped out for require where possible
The `define` method is now forbidden in the repo.

Bug: T212944
Change-Id: I44790dd3fc6fe42bb502d79c39c4081c223bf2b1
2019-07-16 18:04:10 +00:00