The used phpcs has a bug, so the version 0.9.0 could not be enforced at the moment.
Will be fixed in next version, see T167168
Changed:
- Remove duplicate newline at end of file
- Add space between function and ( for closures
- and -> &&, or -> ||
Change-Id: I4172fb08861729bccd55aecbd07e029e2638d311
This moves attributes out of the top level, and namespaces them under
each extension. If the extension that it belongs to is not installed,
the attribute is not exported and dropped.
The full name of the attribute is the name of the extension plus the
name of the attribute key. This enforces the recommendation that the
attribute name start with the extension's name.
Add test coverage for attributes under manifest_version 1 and 2.
Bug: T133627
Depends-On: I5a148763f68989c8da313a4fb1d0213658ee4495
Depends-On: I5a148763f68989c8da313a4fb1d0213658ee4459
Change-Id: I8613a027c56e2c9d2c6a83ca14749eb1c8fc23be
Emit a wfDeprecated() warning for each extension that does not have
manifest_version set.
Because we don't have any mechanism to cache warnings, just disable the
cache if any deprecation warnings are emitted. This ensures that the
warnings are shown on every page load instead of probably once a day.
Bug: T155610
Change-Id: I163a84ae1d381ca5bd67b4b317c04f9f51c066ea
This allows extension namespaces to be assigned a custom ID, e.g. in case
the namespace pre-defined by the extension is already taken on the local
wiki.
This is done by defining the respective namespace constant in
LocalSettings.php.
Bug: T160462
Change-Id: If648d6e218847e6632d643ea724cd3da3945db70
I was bored. What? Don't look at me that way.
I mostly targetted mixed tabs and spaces, but others were not spared.
Note that some of the whitespace changes are inside HTML output,
extended regexps or SQL snippets.
Change-Id: Ie206cc946459f6befcfc2d520e35ad3ea3c0f1e0
Previously, MediaWikiServices would be initialized on the first
wfLoadExtension() call, which is super early and before it should be.
There's actually no reason we need to create the BagOStuff object in the
constructor, so let's defer it to loadFromQueue(). It now gets called
from the top of Setup.php, which is still a little early, but better
than before.
Bug: T154960
Change-Id: I3feef3b974ba1ba3afec0d453e1899cd476e72fb
Instead of having the current version of the schema live at
extension.schema.json, always include a version number in it. This
simplifies both the code, and makes merging, cherry-picks, rebasing,
etc. easier when the current schema version changes.
Change-Id: I1b5ef4ced1eb34b8bdda7ba9b3f086b1997b704c
There are some extensoins that depend upon another extension or skin,
usually in different ways:
* A constant that is added in the dependency extension, and the
existence of is checked for. This is problematic because it requires a
specific load order.
* Checking whether a specific class exists. This is problematic because
it is extremely fragile, and breaks whenever the class is renamed.
* Checking ExtensionRegistry::isLoaded(). This is mostly there, but it
only checks at runtime, and doesn't provide any machine readable data.
Furthermore, developers implement each one differently, with very little
standardization.
With this, extensions may now specify what other extensions they depend
on. This is for explicit *hard* dependencies that must be installed.
For example:
"requires": {
"MediaWiki": ">= 1.25.0",
"extensions": {
"FakeExtension": "*"
},
"skins": {
"FakeSkin": "*"
}
}
This would add a minimum requirement on MediaWiki 1.25.0+ (already
implemented), as well as the requirement that the FakeExtension extension
needs to be installed, as well as the FakeSkin skin. A wildcard (*) is
used instead of an explicit version requirement as many extensions do
not actually version themselves, and there is no consistent versioning
scheme yet.
Bug: T117277
Change-Id: If1cccee1a16a867a71bb0285691c400443d8a30a
This allows us to put other requirements more easily into extension
registration, such as skins and/or extensions.
Bug: T117277
Change-Id: I3ec1b28b6af380621585cd61b38e5ebb8be9f9c7
We want all extensions to be queued together and load at the same time
so in the future we can properly evaluate dependencies as a whole. If
extensions load late, they would bypass this, potentially causing
issues.
Bug: T117277
Change-Id: I09b306bd6f6ccf4210f36be0118e7f17f2c3d264
Registration callbacks now provide basic credits information (name,
path, type, authors, license-name, version, etc.) as the first argument.
The main use case right now for this is to support extension VERSION
constants for backwards-compatibility.
In addition, callbacks now run *after* attributes are exposed, so
callbacks could use data from them if they wanted to.
Bug: T151136
Change-Id: Ic5965dd4e259e1f46222ac92b8e78750e67b51d6
Previously, logic to validate extension.json files was in two places:
validateRegistrationFile.php maintenance script, and the
ExtensionJsonValidationTest.php structure test. This caused duplication
as validation became more complex (e.g. usage of spdx-licenses library).
A generic ExtensionJsonValidator class now handles most of the
validation work, while the maintenance script and test case just wrap
around it for their output formats.
Change-Id: I47062a4ae19c58ee1b1f2bb4877913259bf19c8b
For extension registry, add array_replace_recursive merge strategy,
as some extensions/configuration may prefer that to array_merge_recursive.
In some cases, configuration is merged from multiple extensions,
such as JsonConfig's $wgJsonConfigs configuration: ZeroBanner defines
"JsonZeroConfig": {
"namespace": 480,
"nsName": "Zero",
"isLocal": false,
"cacheKey": "1"
}
and mobile.php overrides it with
$wgJsonConfigs['JsonZeroConfig']['isLocal'] = false;
$wgJsonConfigs['JsonZeroConfig']['remote'] = [
'url' => 'https://zero.wikimedia.org/w/api.php',
'username' => $wmgZeroPortalApiUserName,
'password' => $wmgZeroPortalApiPassword,
];
Having identical value 'isLocal' would be converted into an array
if array_merge_recursive is used, but the replace strategy fixes it.
Change-Id: Ica6ddd0ae76f23e60de9b6235c6e2a3f2754a95d
* Use services container in more places.
* Undeprecated getLocalServerInstance() since $fallback is not
handled elsewhere.
Change-Id: Id1fcd1c465d2d92653357523f4225f1c4d1ace2f
Some extensions (e.g. EventLogging) conditionally register a namespace
depending upon the wiki's configuration using the CanonicalNamespaces
hook. Since the "namespaces" key automatically registers the namespace,
we need an option to not register it and let the extension take care of
that manually using the hook.
All of the other namespace related settings are safe to unconditionally
set since they won't do anything if the namespace doesn't actually
exist.
Bug: T141604
Change-Id: Ie8c217fdc8bd783b30f98210309ea56fef39c0da
Add support for configuration settings that are relative paths to the
extension directory. The most common use-case for this is where the
extension ships a generic default file, but sysadmins or packagers may
want to override it. Setting `"path": true` on the config definition
will make the value be interpreted as a relative path.
And add tests!
Bug: T100956
Change-Id: Ia87ced25b35be7a314ee2937e0ccc63f9acc5bb5
merge_strategy is set in $data, which is the definition of the config
setting, not the default value of it ($value).
Change-Id: I12ee95892b2c157bb0beb97cf2cd5ef920b7d52e
To add extra metadata for "config" options without constantly adding
hacky underscore prefixed keys, convert "config" items into an object,
where the value is under a "value" key.
"_merge_strategy" is now just "merge_strategy", but still has the
underscore prefix for the cache structure, to avoid changing it.
Since this is a fully breaking change, it only applies to files with
manifest_version: 2 set. A conversion script has been added to assist
with automated conversion.
Bug: T133626
Change-Id: Id1071fc0647892438e5cd0e3ee621fbdaaa64014
It never worked to begin with since it did not prefix paths, and the
documentation for it says it should not be used by extensions, so
let's stop pretending it works.
Bug: T108271
Change-Id: I665d4439d14406038f2ac767ee5ec6eb5b67fca6
This implements the AuthManager class and its needed interfaces and
subclasses, and integrates them into the backend portion of MediaWiki.
Integration with frontend portions of MediaWiki (e.g. ApiLogin,
Special:Login) is left for a followup.
Bug: T91699
Bug: T71589
Bug: T111299
Co-Authored-By: Gergő Tisza <gtisza@wikimedia.org>
Change-Id: If89d24838e326fe25fe867d02181eebcfbb0e196
Instead of forcing an object even for single-hook-listeners, allow
string values, too (one hook listener for one hook, only). Also:
use it as default for the conversion script, if only one listener
is added to a hook (which is usually the case). This leads into a
much cleaner output of the Hooks section of extension.json.
Bug: T133628
Change-Id: Ie9e54f0931c41706eb8d82d00256698992ec41cc
Callbacks should be able to safely assume that anything (or at least mostly)
defined in extension registration is already loaded and processed when
the callback itself is executed. To make sure, that this applies, callbacks
should be executed after all extra autoload paths are loaded.
Bug: T131978
Change-Id: I2c6624423957a8a00523b126fa7209d9c283aa9e
Noticed it in APC usage for ResourceLoader that various keys from
FileContentsHasher contain double slashes. All from extensions that
use the `"localBasePath": ""` trick in extension.json (e.g. Citoid).
Change-Id: I5bac1e2e05e063aa7ff251ce7ffaa965a3451db9
...but don't remove the code. Calling isset( $bar['foo'] ) without checking
that $bar is an array seems not very nice to me.
Change-Id: I822c925b6f36bf34902f8075e54f71fe4f6d2566
