Add Database::queryMulti(), which will execute an array of
queries as a batch with minimal roundtrips.
SQLite fallbacks to looping through each statement and
invoking doQuery().
Add QueryStatus class to reduce complexity in Database.
Rewrite doQuery() as doSingleStatementQuery().
Change-Id: I3d51083e36ab06fcc1d94558e51b38e106f71bb9
When JSON support was introduced into ParserCache in 1.36, it was
controlled by a feature flag, $wgParserCacheUseJson. The feature flag
was "born deprecated" in 1.36. It can now be removed.
This means that ParserCache will always store entries as JSON.
Support for reading old non-JSON entries remains intact.
This is needed when updating wikis from a version older than 1.36
to the current version.
Change-Id: Id04e42bfb458d98414bac50e0d6c505e8878e5c0
In case two revision have the same timestamp, the higher rev_id is the
later revision and needs to be patrolled as well.
This also fix the tests, where the clock was faked to get higher
timestamps on the revisions
The WHERE changed from:
rc_cur_id = 2 AND (rc_timestamp > '20220603122124') AND rc_actor = 10
to
rc_cur_id = 2 AND ((rc_timestamp > '20220603122124') OR (rc_timestamp =
'20220603122124' AND (rc_this_oldid > 2))) AND rc_actor = 10
Bug: T309817
Bug: T304766
Bug: T260798
Follow-Up: I66bb6575dad4a970339e93b91fe486727111092a
Change-Id: I3cfcca4f0405f1d701c2883109c53b15854d6e64
When performing a rollback, we marks the reverted edits as
patrolled (on UseRCPatrol wikis) and also pretend they were by a bot
(discussion at T304766).
The way it is marked as patrolled is by setting the patrol status to
autopatrolled. But, if (one of) the edits was already patrolled before
the rollback, the manual patrol status should not be overwritten.
* Fix this.
* Also improve query performance by setting via rc_id instead of
through more complex conditional clauses in the update query
(should help with T260798).
* Test as much of this patrolling stuff as possible.
Bug: T304766
Bug: T260798
Change-Id: I66bb6575dad4a970339e93b91fe486727111092a
As a means of understanding the usage of the stash FEAT for
/page/html & /revision/html endpoints used by VE extension,
this patch introduces the collection of stats using the
StatsDataFactory.
Bug: T309017
Change-Id: I4e17d50e79da263637bdd55ab62e993df441fe38
This patch enables the response from PageHTMLHandler and
RevisionHTMLHandler to have different eTags for different
output modes and varying flavors.
Before, the only difference we got was when the stashing
option is set or not, but we need more flavors.
Bug: T308744
Change-Id: I2e9679e46a31955a2106a52af4eb612b32799c8c
Add stash option to /page/html & /revision/html endpoints.
When this option is set, the PageBundle returned by Parsoid is
stashed and an etag is returned that can later be used to
make use of the stashed PageBundle.
The stash is for now backed by the BagOStuff returned by
ObjectCache::getLocalClusterInstance().
This patch adds additional data to the ParserOutput stored in ParserCache.
Old entries lacking that data will be ignored.
Bug: T267990
Co-Authored-by: Nikki <nnikkhoui@wikimedia.org>
Change-Id: Id35f1423a69e3ff63e4f9883b3f7e3f9521d81d5
NOTE: This changes the HTML returned by the endpoint!
It will now include the id="mwXYZ" attributes needed to
later map to data-parsoid entries.
Bug: T268205
Change-Id: I0a29434b996cc289eb67083e62bd6f1ad750cb4d
The underscores are not strictly necessary, but appear to be good
practice. The vast majority of special page aliases we have already
use underscores (94% according to my test).
I made this a warning in the test for now, not an error.
Change-Id: I18827eb795f84bdb7895793d00bc3c895abebc7a
Even the service does not long stay in that classes,
it should be injected to avoid global state
Bug: T304780
Change-Id: Ib488037f5a6966ab61042ed3cd889ddc50f1ba8e
As required by the Iterator interface. In PHP 8.1 we can suppress the
warning with #[\ReturnTypeWillChange], but it's wrong and should change
anyway.
Depends-On: I576109808755b054e7876556f244ee8eafd12f9d
Change-Id: Ibaa7b4c81d8f114783db45c9100200ba14547c8f
It calls the Shell class, which uses MediaWikiServices. Why was this not
causing test failures on CI?
Change-Id: Id953c1425659ede158034988dd652b59a0e9811b
I would like to argue that it's useful to know if an error message is
returned, and if it's the correct one.
Bug: T306568
Change-Id: I865fa4c8a5b2d2d88d26c190352c852e63913a56
All revision related classes are namespaced MediaWiki\Revision
instead of MediaWiki\Storage since 1.32. The old namespaced
class names are deprecated and only kept for backwards-compatibility.
Bug: T305784
Change-Id: I34e492d84d9fc4bc78481667202716d93b3c43cb
Rename canRecoverFromDisconnect() in order to better describe
its function. Make it use the transaction ID and query walltime
as arguments and return an ERR_* class constant instead of a bool.
Avoid retries of slow queries that yield lost connection errors.
Track session state errors caused by the loss of named locks or
temp tables (e.g. during connection loss). Such errors will prevent
further queries except for rollback() and flushSession(), which must
be issued to resolve the error.
Add flushPrimarySessions() methods to LBFactory/LoadBalancer
and use it in places where session state loss is meant to be
safely aknowledged.
Change-Id: I60532f86e629c83b357d4832d1963eca17752944
We really don't need this complexity and it prevents us from improving
connection management.
MaintainableDatabase should stay but the connection ref shouldn't.
Bug: T255493
Change-Id: I867301dc7fa07cac298f8faba9cf82ca4617f50e
Add services and utilities for automatic creation of temporary user
accounts on page save, in order to avoid exposing the user's IP
address.
* Add $wgAutoCreateTempUser, for configuring the system
* Add TempUserConfig service, which interprets the config.
* Add TempUserCreator service, which creates users during page save as
requested by EditPage. With proxy methods to TempUserConfig for
convenience.
* Add table user_autocreate_serial. Table creation is necessary before
the feature is enabled but is not necessary before deployment of this
commit.
Bug: T300263
Change-Id: Ib14a352490fc42039106523118e8d021844e3dfb
These three endpoints have been experimental for many months:
revision/{id}
revision/{id}/html
revision/{id}/with_html
Promote them to officially released. This completes the
basic "revision" endpoint support, and helps clear out
the coreDevelopmentRoutes.json file for unrelated
experiments.
This also modifies the existing revision/{id}/bare endpoint
response, which previously pointed callers to the experimental
endpoint for html. It now points callers to the official one.
Bug: T305506
Change-Id: Iee8d1723e98dd3e3e389a0514dde28799914b2fd
All revision related classes are namespaced MediaWiki\Revision
instead of MediaWiki\Storage since 1.32. The old namespaced
class names are deprecated and only kept for backwards-compatibility.
Bug: T305784
Change-Id: Ia0030814ce2176d06e2898acffe533d31633fccb
When newAnonymous() was called with a specified IP address, it called
newFromName(), leading to a DB query on User::load() as if the IP
address could be in the database.
Instead, use User::setName() which is documented to work for this
purpose and indeed appears to do so.
Change-Id: I8e8be719c89b1d1e4ae2cf9063db68b929bb3651
Use multiple error boxes instead of a wikitext list of error messages,
as the latter has several issues (as reported on phab). Also, remove the
edit-error-* messages as they don't seem to add any value. Also, don't
use the passed-in context, inject what the class actually needs instead
(but keep the context for the hook).
Bug: T303060
Bug: T230229
Change-Id: I2806e57826e7ab062b45dd1d9972b05fd6baeafb
Deferring newtalk updates caused the skin to show the notification on
the same user talk page view request in which it is cleared. Even if the
user refreshed the page, it would still show the notification since a
304 response was given.
So, move the newtalk part of WatchlistManager
::clearTitleUserNotifications down to TalkPageNotificationManager, so
that the in-process updates can be done immediately, with the DB updates
being deferred. TalkPageNotificationManager is now responsible for
deferring the relevant parts.
Originally (2002) the newtalk status was set by calling
User::saveSettings(), which had the side-effect of updating
user_touched. Restore this behaviour, which is fast now due to the
"quick touched" cache. Also update the touched timestamp when the
notification is cleared.
Change-Id: Ic3c14bca7bed04f46afffa971fa0a57224c666ce
This reverts commit eed58f2f61.
Reason for revert: Blocking the train, going to revert the whole chain
Change-Id: I64f9e5a9dde106671783f958a686ca697182077b
Use the LoadBalancer id in flushPrimarySessions(), not the LBFactory one,
and use assertOwnership() to check $owner, similar to other methods.
In DatabaseMysqlBase::doFlushSession(), change RELEASE_ALL_LOCKS() query
to use RELEASE_LOCK(), since only newer MariaDB versions (>=10.5.2) support
it. No errors were thrown in the method since they are suppressed, but the
syntax error would cause the transaction to be placed in an error state.
Add assertion to testTransactionCallbackChains() that would otherwise fail.
Randomize lock names in lock() tests to avoid contention.
Bug: T292239
Bug: T303887
Follow-Up: ee3c65d541
Follow-Up: 4cac31de4e
Change-Id: I414d737028338cfd5369eee24576df4aa26a2f6f
This ensures that assertions work in a uniform way,
and provides meaningful messages in cause of failure.
Change-Id: Ic01715b9a55444d3df6b5d4097e78cb8ac082b3e
The policy allows this and since 1.39 is going to be the next LTS
release, I think it is fine to do this now.
Change-Id: If426e0ee349252ccc0ba9c4222c7d6865ab57fa2
The testUpgrades() method is slow and is easier to
parallelize via paratest if it lies in a separate class/file
Also make the test use a data provider
Bug: T50217
Change-Id: I5c8df6ae0ff34941f4cf7275680b309e79565925
