| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Not needed (any more) because the code already contains
enough @return docs and such.
Change-Id: If461fa035d2c646878540565b3087a74c6628552
|
| | |_|/
|/| |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This forces SqlBagOStuff to instead of sharding keys, write to n out of
m servers instead. It also reads from those servers as well and in case of
incosistency, picks the value with the highest exptime.
This is mostly for mainstash and allows us to provide stronger
consistency guarantees while allowing for a section to be depooled and
put to maintenance. It basically implements the logic already used by
NoSQL database systems such as Cassandra (There are two types to solve
conflict, quorum or timestamp, Cassandra is using quorum while we are
using timestamp).
There will be some edge cases that it might still pick the wrong value:
- if TTL is set to INDEF
- if the TTL gets shortened for various reasons.
- If we go with two clusters, value is set, one gets depooled, a new
value is set, the depooled one gets pooled and the other depooled and
then read happens.
But all of these are extremely rare edge cases and we should be fine.
This also means if data redundancy is set, locking means all sections
will be locked and removal means all sections must allow the unblock.
Otherwise, the lock will be kept.
Bug: T383327
Change-Id: I80da12396858ee4fc58ae257f6c154b3050df696
|
| | | |
| | |
| | |
| | |
| | |
| | | |
Follows-up Ia9ddb1b83a6f20d3f7ce.
Change-Id: I685699d905c94e3b5af99a73dd3ba1d8def2d779
|
| | |/
|/|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Why:
- We want to replace the PageDeletionComplete hook with an event
- We want to validate our design for the hierarchical event type system
and the modeling of page events.
What:
- Introduce PageDeletedEvent
- Make DeletePage emit PageDeletedEvent
- Move update triggers from DeletePage into ingress objects:
- search index
- message cache
- module cache
Bug: T379932
Change-Id: I01490fbaf33118eba109aa91908783117ba5aa20
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* Add a batch version of CentralIdLookup::isOwned(), to support the
linked bug.
* Add test.
* Fix error in LocalIdLookupTest::provideIsAttachedShared() which
failed to provide the shared case due to an incorrect loop bound.
This test has been non-functional since it was introduced in 2015.
* Fix error in LocalIdLookupTest::newLookup() which failed to override
the config.
Bug: T386584
Depends-On: Ie63ed1f14c9ba4f9cefcceef05d5585a6676f111
Depends-On: I83716f7930bd0d2de4761cc5b04d9d3ff4172da2
Change-Id: I99dd733547f46e635c2f363166ce111e76e06a5e
|
| | |
| |
| |
| | |
Change-Id: Ic5f178b202574b937169a47faf1566bd62bf6ad3
|
| |\ \ |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Why:
- When a transaction lifecycle callback such as
onTransactionPreCommitOrIdle() is invoked by rdbms, rdbms passes in
the DB connection handle the callback was scheduled on.
- However, the DB connection handle may have since been reset to a
different DB domain on the same physical cluster due to connection
reuse.
- We could implement bookkeeping to keep track of the expected DB domain
for each pending callback and ensure the passed-in DB handle matches
that before invoking the callback, but nobody seems to be making use
of the passed-in DB handle except for two call sites in core.
- As such, it seems more prudent to remove this functionality.
What:
- Stop passing the DB connection handle to rdbms transaction lifecycle
callbacks.
Bug: T386190
Change-Id: I9125bfb8b5e2cac4aab3525ffd229ea49beccc17
|
| |\ \ \ |
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Why:
- We only want to support a single invocation mode for now.
What:
- Remove handling of invocation modes from EventSubscriberBase and
EventDispatchEngine.
Bug: T387012
Change-Id: Ie1df95f50cdb763fceda5751d6defa665fd114ff
|
| |\ \ \ \ |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Why:
- OutputPage::addModules() and OutputPage::addModuleStyles() currently
store added module names in two lists, mModules and mModuleStyles.
- The contents of these lists then get passed to array_unique() when
OutputPage::getModules() is invoked, to filter out any duplicate
module names.
- In I4f70ff15becbc4991c4f1b0307a14d5354c79dc1, we would like to modify
UserLinkRenderer::userLink() to automatically add the requisite
ResourceLoader modules to the output, instead of every page where
userLink() is called needing to do so manually.
This can cause these modules to be added thousands of times when a
large list with many user links is rendered, e.g. on RecentChanges.
- Running `php maintenance/run.php benchmarks/benchmarkEval output.php`,
where output.php holds:
```
$out = ( new RequestContext() )->getOutput();
for ( $j = 1; $j <= 2000; $j++ ) {
$out->addModules( 'ext.foo' );
}
$out->getModules();
```
takes about 4-5ms on my system.
- Turning the properties into associative arrays and getting rid of the
array_unique() call drops this to < 0.2ms.
- These properties are exposed via public property accessors that have
been deprecated since MW 1.38 and have no usages in codesearch.[1]
Since we're changing how the properties work, this is a good time to
remove the deprecated public property accessors.
What:
- Use associative arrays to store module names in OutputPage.
- Document that addModules() and addModuleStyles() take a
string|string[] rather than a string or an arbitrary array.
- Add unit tests for the behavior.
- Drop the deprecated public property accessors for mModules and
mModuleStyles.
[1] https://codesearch.wmcloud.org/search/?q=mModule%5BSs%5D
Bug: T358469
Change-Id: I86c557a4ce7207359d100538c0d4b1ffa75fcbf9
|
| |\| | | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Why:
- After I4f70ff15becbc4991c4f1b0307a14d5354c79dc1, it will become
necessary to prefetch the expiration status of temporary accounts when
rendering user links via UserLinkRenderer for a large list of users.
- We would ideally like to accomplish this without requiring every user
of UserLinkRenderer to take a direct dependency on
TempUserDetailsLookup, which exposes this functionality.
- Since most users of UserLinkRenderer / Linker::userLink() already use
LinkBatch to prefetch page existence for the user and user talk pages
that UserLinkRenderer will link to, encapsulating the
TempUserDetailsLookup interaction within LinkBatch seems like a decent
middle ground.
What:
- Add LinkBatch::addUser(), which takes a UserIdentity and adds its user
and user talk pages to the batch while also triggering a batch lookup
of expiration status for users added this way when execute() is called.
Bug: T358469
Change-Id: Ic837961296cc4bf166dde79c7f073cc50ce925da
|
| |\ \ \ \ \ |
|
| | | |_|/ /
| |/| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Just a few modernizations, utilizing more recent PHP features.
Notable:
* Much more strict language-level types. E.g. it's not allowed to
call newFromNames with a single string any more. (Was never
allowed and doesn't make any sense anyway.)
* Drop unnecessary @covers tags in the test for the reasons
explained in I2678992.
Motivated by I15d9134.
Depends-On: Ia88a8e724bb35781286528fa6403b2fe8ee3cfab
Change-Id: Id90b6838429eca4bd662546e89fb1a91a8ae7c4f
|
| |\ \ \ \ \
| | |/ / /
| |/| | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Why:
- The TSP team would like to adjust the rendering of user links
associated with temporary accounts, depending on whether the temporary
account is expired.
- This requires a mechanism to efficiently retrieve the expiration
status of temporary accounts, even when rendering large lists with
many user links, e.g. Special:RecentChanges or page histories.
What:
- Introduce the TempUserDetailsLookup class and service that holds the
expiration status of temporary accounts in an in-memory LRU cache.
This uses the new UserRegistrationLookup::getFirstRegistrationBatch
API.
- Provide a preloadExpirationStatus() method, similar to GenderCache or
LinkBatch, that allows callers to prefetch expiration status for a
batch of users. A followup patchset will add these to pagers as
needed.
Bug: T358469
Change-Id: I98ec17046c0863878eba3768d83f083bc2753e6f
|
| |\ \ \ \ \ |
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Generate the error message in the getLegacyStatus function in the constraint
class instead of EditPage. The submit button label has to be passed to the edit
constraint since the logic for determining it it resides in EditPage.
Bug: T384399
Change-Id: Id2a9ab57198dfa4df07f91b067944e9f9fefb81a
|
| | |/ / / /
|/| | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
MessageCache has enough to do without also being a parser. Split a
MessageParser class out of MessageCache and make it a service.
* MessageCache::parseWithPostprocessing() becomes MessageParser::parse()
because every caller wants postprocessing so it makes sense to use the
shorter name for this.
* MessageCache::parse() becomes MessageParser
::parseWithoutPostprocessing(). I changed the return type from
ParserOutput|string to ParserOutput, which is a followup to
I8a1fa84e650d920d07d74722d8059d5afeedec6b. Narrowing the return type
does not break b/c so it is possible to make this change for both
variants.
* In the new methods, a null title is always a convenience alias for a
placeholder title (Special:Badtitle). This reflects the convention in
Parser::setPage(). The old MessageCache::parse() retains its b/c
fallback to $wgTitle. MessageCache::transform() had the potential to
fall back to the title used in the previous call, a fragile mechanism
which I removed without deprecation.
* For consistency, allow a string language in all new methods.
* In EmailNotification, clean up an early attempt at global state
avoidance.
Change-Id: I05ab21508d5f8394189fd41ac6a2254ac0e0d785
|
| |\ \ \ \ \
| |/ / / /
|/| | | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Why:
- MediaWikiPropagator is currently untested.
What:
- Add unit tests for this class.
Change-Id: I8278b3f49bdd838be6fba9264d666f3d3ba3423d
|
| |\ \ \ \ \ |
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
When there are no handlers system shouldn't fail with an exception.
Bug: T383992
Change-Id: Ia33524af2574d10375f2cb0586bbdabd9d0adb48
|
| | |/ / / /
|/| | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Why:
- The TSP team would like to change the way expired temporary account
user links are displayed, which requires an efficient way to fetch
their registration timestamps.
- On WMF wikis, which use CentralAuth, this requires fetching the first
(i.e. global) registration timestamp of the account, rather than the
naïve approach of using the registration timestamp from the local user
table.
- MediaWiki provides the UserRegistrationLookup facade to transparently
fetch the earliest registration timestamp for a single user, but
offers no batch interface to do the same.
- Since user links are often rendered in large pagers, a batch interface
is needed.
What:
- Add IUserRegistrationProvider::fetchRegistrationBatch(), which takes
an iterable of UserIdentities and returns a map of their registration
timestamps (or null if not available), keyed by user ID. Although this
interface is marked as stable to implement, its sole non-core
implementor according to codesearch is CentralAuth.
- Add UserRegistrationLookup::getFirstRegistrationBatch(), which
delegates to fetchRegistrationBatch() on configured registration
providers and returns the earliest registration timestamp for each user
in the batch.
- To avoid potential interface incompatibility in WMF production, this
depends on CentralAuth implementing the new IUserRegistrationProvider
method first.
Bug: T358469
Depends-On: Ibe28163e962161567d486607e36d999a36a1e604
Change-Id: I1f6af2693a8f0c5c854b8a6b04edd1eb21934007
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Main change:
* Add a class hierarchy representing block targets, representing a
target and type.
* Add BlockTargetFactory, replacing BlockUtils.
* Add CrossWikiBlockTargetFactory, replacing BlockUtilsFactory.
* Construct a BlockTarget object early in the request flow and pass it
down through the layers, instead of having every layer interpret
UserIdentity|string target specifications.
Also:
* Remove Block::TYPE_ID. Nothing uses it in code search, so there's
no point in porting it to the new system.
* Stop using the type constants as specificity scores. Add
BlockTarget::getSpecificity().
* Add DatabaseBlockStore::newUnsaved() to replace direct construction of
DatabaseBlock in insertBlock() callers. There are many such callers in
tests. This is part of the effort to remove the service container
usage in DatabaseBlock::__construct().
* Make DatabaseBlock::getRangeStart() and getRangeEnd() return null if
the block is not a range, since that is convenient for their only
caller following the resolution of T51504.
* Add DatabaseBlock::getIpHex() which similarly maps to a DB field in
the new schema.
* In ApiBlock and ApiUnblock, have ParamValidator provide UserIdentity
objects instead of converting to a string and back to a UserIdentity
again.
Bug: T382106
Change-Id: I2ce1a82f3fbb3cf18aa2d17986d46dbdcc70c761
|
| |\ \ \ \ \ |
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
The doc comment of newBlockPermissionChecker() describes the $target
parameter as being optional. The returned object does not need or use a
target when checkBasePermissions() or checkEmailPermissions() are
called. But failing to pass a target when calling
checkBlockPermissions() is incorrect. This was the subject of the
linked bug. If an admin is performing a block, self-unblock permissions
need to be checked, this is not optional.
This could be enforced at runtime, but it seems safer and simpler to
enforce it statically.
So, move $target from being a constructor parameter to being a formal
parameter of checkBlockPermissions(). This formal parameter will be
statically required after a deprecation period.
Deprecate newBlockPermissionChecker() and introduce newChecker(), using
a name with a slightly less than conventional verbosity to allow us to
change the parameter order. There is no $target parameter to
newChecker().
Backwards compatibility is supported by adding an internal mutator
method setTarget(). This can easily be removed after the deprecation
period is over.
In Special:Block, checkBlockPermissions() was called with $target=null
on form entry, meaning that the user could unblock themselves if the
target was specified in the URL, but could not unblock themselves by
searching with the form. This seems inconsistent. So allow blocked
admins to see the search form, but show an error when they try to block
or unblock someone other than themselves.
Bug: T384716
Change-Id: I8c26cdcc9b87b74bc458fe731cf7f170a2607150
|
| |\ \ \ \ \ \
| |_|/ / / /
|/| | | | | |
|
| | | |_|/ /
| |/| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Why:
- $wgHooks should work like a regular configuration variable which
cannot be modified after bootstrap.
- FauxGlobalHooksArray was in place to trigger deprecation warnings for
write access to $wgHooks since 1.40.
What:
- Remove FauxGlobalHooksArray and supporting code
Bug: T331602
Change-Id: I8d881857850a906b893cf3066dcfbe277f9b0b14
|
| |\ \ \ \ \
| |/ / / /
|/| | | | |
|
| | | |/ /
| |/| |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Create a provider pattern for sending notifications (T383992).
There is no base notifications handler in core; they can only
be provided by extensions.
A handler in the Echo extension, compatible with the existing
Echo notifications system, will be implemented in T383993.
Co-Authored-By: Piotr Miazga <pmiazga@wikimedia.org>
Bug: T383992
Change-Id: I16b309935c3d29c3cde4459b5e36abce063a8534
|
| |\ \ \ \ |
|
| | |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Added in I7074b65d07c5c7d and unused ever since. It was used until patch
set 6, then ExtensionProcessor was refactored to support batching in PS7
and the tested method, `setToGlobal`, was removed together with its
test but not this data provider.
Change-Id: I31a07ec8ea320cd5dc946f881f380f603632619a
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
content and file pages
This patch reworks RestrictionStore::getCascadeProtectionSourcesInternal
to return a third and fourth array:
* One for cascading restrictions originating from templatelinks
* Another for those originating from imagelinks
They are used in PermissionManager::checkCascadingSourcesRestrictions
to differentiate cascading protection of file content and file page,
but could also be used in the future by action=info and other callers.
Bug: T24521
Bug: T62109
Bug: T140010
Change-Id: Ia5863f418538106f4fd657c672298ff6ac835805
|
| |\ \ \ \ |
|
| | |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
It is valid to use TS_MW for chronological compare as the timestamp
stored in a string is ordered from the most significant datetime
information (year) to the lowest (seconds).
It is not needed to convert to unix.
It would not be valid to use TS_MW format to calculate time differents.
This is how it is used in the database schema from mediawiki as well.
Also replace array + min() with variable + if.
Change-Id: Id41e660f3952876e20ac4e88502d7e2757299573
|
| |\ \ \ \
| |_|/ /
|/| | | |
|
| | |/ /
| | |
| | |
| | | |
Change-Id: I23ce3b4d26bbe58257b449993a54c45093007a3e
|
| |/ /
| |
| |
| |
| |
| |
| | |
Also avoid null as array item,
as that is not allowed according to the return type
Change-Id: I4083c55a69d6186448a13f35f18d96bfe9ffd23c
|
| |\ \ |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* Many tests failed if $wgAutoCreateTempUser['enabled'] was not set.
This is set in DevelopmentSettings.php. Set it in TestSetup.php.
* The Asia/Calcutta timezone, described as deprecated, is missing from
my system. Remove the test case.
* Three stub object tests called error_reporting(-1) on setUp() and
restored it on tearDown(), but our check for preserved error level
now runs as a postCondition hook, before tearDown(). This worked
anyway for most people because DevelopmentSettings.php also has
error_reporting(-1). Remove the error_reporting(-1) calls from the
tests since it's not actually needed to pass -- we intercept
deprecation calls at the application level. I confirmed that this fix
works with php -d error_reporting=0.
Change-Id: Ia43ce26345a0a8432c41b46907de2dfbffa00670
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This allows footer icons to declare different images at different
media queries via the picture element
e.g.
$wgFooterIcons['poweredby']['mediawiki']['sources'] = [
[
"media" => "(max-width: 720px)",
"srcset" => "https://en.wikipedia.org/static/images/mobile/mediawiki.png",
];
Bug: T384619
Change-Id: Ib325690906bc7fd9c0668cf2f77f0690d826d98e
|
| |\ \ \
| |/ /
|/| | |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Why: RFC9110 requires 304 "not modified" responses to contain the "date"
and "etag" headers. Failing to include them caused etags to be missing
from responses that were returned from cache after checking the backend
using if-modified-since.
What:
- Make Handler::checkPreconditions() call
applyConditionalResponseHeaders() to add the conditional headers
if the precondition check fails.
Bug: T357603
Change-Id: Ic2e288c971774a561395b3df04bab7256ca62014
|
| |\ \ \ |
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Why:
- Callers of the dispatch() method should be able to be sure that
event listeners cannot interfere with the current transaction.
What:
- remove support for the "before commit" invocation mode
Bug: T379775
Change-Id: I04ae5281228e71e82a1fe514d2d7a5e53b510cc8
|
| |\ \ \ \ |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Bug: T368740
Change-Id: I1b9435dcdacd952b2bc703b3d5f7083406ebbed6
|
