A bit of a clean up. Switching out of lbFactory to ICP is not possible
at the moment because there is $this->lbFactory->getLocalDomainID()
for cache prefix. Fixable later.
Bug: T330641
Change-Id: I2d4537f75f5877552c6d9fd2b76c5be1959ea400
Just methods where adding "static" to the declaration was enough, I
didn't do anything with providers that used $this.
Initially by search and replace. There were many mistakes which I
found mostly by running the PHPStorm inspection which searches for
$this usage in a static method. Later I used the PHPStorm "make static"
action which avoids the more obvious mistakes.
Bug: T332865
Change-Id: I47ed6692945607dfa5c139d42edbd934fa4f3a36
We always wrap the local cluster cache, and there are no subclasses
of WANObjectCache. It was never documented or recommended how these
would be used. It is a left-over from the original 2015 Multi-DC plan
in which WANObjectCache would work differently. See task for details.
Note that this requires no configuration changes, even in the
theoretical case of these variables being used, as the only
option is to use the main cache, and that's also the default.
* Update WAN overrides to override the underlying main cache
instead.
* Fix EditPageTest which was previously implicitly using a 'hash'
as main cache but also relying on wan cache to be 'none'.
The part that it actually needs is the 'none'. When WAN cache is
enabled, testUpdateNoMinor fails due to an edit conflict because
one of the edits it makes is made with a current timestamp whereas
it expects to simulate wpEdittime in the year 2012 which, when
caching is enabled, is ignored and becomes the current time instead.
I don't understand exactly why, but I'm going to conserve that
behaviour for now.
* Fix TemplateCategoriesTest, which was failing due to an unexpected
cache hit:
> [objectcache] fetchOrRegenerate(…:page:10:…): volatile hit
This could be solved in a more realistic way by splitting the test,
or by explicitly resetting services half-way the test to clear
WikiPageFactory, PageStore and WANCache process state.
For now, keep the prior behaviour of no cache in this test.
Bug: T305093
Bug: T329680
Depends-On: If890622eed0d0f8b4bd73d36ba1815a3d760ea05
Depends-On: Ie1def75208822bdf19bb2cfd7e6edf32c2000e6b
Depends-On: I35cce61dc3ee90dcee3dd6f0b36f84133be029ed
Change-Id: I53781a8c06ebb2583f6ca83dd91bbfe8a5c88b13
Introduced in 2017 with I7f14b9ca2533032 (2e5eb693) but remains
unused at WMF, and disabled by default.
Follows-up I62107789fa (9e49260fc9) which added reap to LinkCache
test cases in 2021.
Change-Id: I0654c29a671467dd6b366f462d1c09b90a273413
The day has gone. Still keeping the code as the schema changes are not
done in production but the data migration has been finished.
Bug: T299417
Change-Id: I906e069a63d1dae14924c72318b22b16244371d6
createMock() is still relatively new. This code was using more
complicated mock builders before. This was changed just recently.
createMock() is now so short, the extra helper methods don't make
the code more readable, I would argue.
Change-Id: Ia7e24827157d5f49fc7da102418c79ae33c8e053
- schema change to allow tl_namespace and tl_title being empty
This is done by removing them from primary key. They don't need to be
nullable as they have default value.
- Make sure with WRITE_NEW, updater avoids writing to the old columns
Bug: T306674
Change-Id: I2b8a29043e952060e7a79b6a7a3d647d48cd16fb
Even the service does not long stay in that classes,
it should be injected to avoid global state
Bug: T304780
Change-Id: Ib488037f5a6966ab61042ed3cd889ddc50f1ba8e
Follows-up I361fde0de7f4406bce6ed075ed397effa5be3359.
Per T253461, not mass-changing source code, but the use of the native
error silencing operator (@) is especially useful in tests because:
1. It requires any/all statements to be explicitly marked. The
suppressWarnings/restoreWarnings sections encourage developers to
be "lazy" and thus encapsulate more than needed if there are multiple
ones near each other, which would ignore potentially important
warnings in a test case, which is generally exactly the time when
it is really useful to get warnings etc.
2. It avoids leaking state, for example in LBFactoryTest the
assertFalse call would throw a PHPUnit assertion error (not meant
to be caught by the local catch), and thus won't reach
AtEase::restoreWarnings. This then causes later code to end up
in a mismatching state and creates a confusing error_reporting
state.
See .phpcs.xml, where the at operator is allowed for all test code.
Change-Id: I68d1725d685e0a7586468bc9de6dc29ceea31b8a
Expected value is the first parameter to assertSame() or assertEquals().
And turn to use assertCount() for some assertions aginst count of array.
Based on code search `assert(?:Same|Equals)\(.+,.+expected` and I look
through files roughly, so some assertions that don't contains 'expected'
are also fixed. In the meantime, some assertions that I am not clear
about are not touched.
Change-Id: I75798b60d29fd19b33f4fdf34ed3c788db420d01
The remaining test cases (MessageCacheTest-*) should already test
everything that testing for sunday does. This change is to ensure that
changes to the translation strings made on translatewiki do not cause
unnecessary test failures.
Bug: T298392
Change-Id: Ie9632da6eee505ab77108b74aa126e479d99be4f
These are needed PageStoreRecord::REQUIRED_FIELDS and that's why putting
PageStore on top of LinkCache caused a lot of increase in db reads.
Bug: T297669
Change-Id: If77c2f9879d7bae71eb59944efd8b3798d16aa46
The code that was previously in LinkCache::getGoodLinkRow
may return a cached row object (either from the in-process
cache, or memcached). In some cases, this caused
LinkCache::addGoodLinkObjFromRow to throw, because a field
was missing (although the field is missing only from
the cached copy, not from the database).
To avoid this, try to invalidate the cache
in LinkCache::getGoodLinkRow and retry, before
letting the exception propagate.
Bug: T205349
Change-Id: Ie9e90bf32964047c1831f575cc260d7d62e9e848
Some methods in the PageUpdater's class implements the fluent interface
design pattern. Use the fluent interface where need be.
Change-Id: If76a4b8c5070c20ed40038a4ee78e2d677de5180
addGoodLinkObj() has many optional arguments, but omitting them actually
means corrupting the cache.
Nearly all existing callers are in tests.
So LinkCacheTestTrait::addGoodLinkObject() was created only
for testing. It is better to have this method in the
trait, because building the row directly in each test
would make these tests brittle against schema changes.
The only usage in WMF production code was in WikiPage and has been
fixed.
Bug: T284955
Change-Id: I03a2bd9ed64fcc0281ee29a286c8db395a9e03d9
This makes the data stored by LinkCache compatible with PageStoreRecord,
so we can use LinkCache inside PageStore.
This causes PageStore to make use of local caching as well as WANObjectCache.
Note that getPageById() does not yet benefit from cache, but does
populate the cache.
Bug: T278940
Change-Id: Icc27a0d9299a3e4ce45521daef87ad06ec06f064
This patch does several things to LinkCache to make its behavior more
consistent and predictable:
* Methods that set a "good" link now clear the "bad link" flag, and vice
versa.
* invalidateTitle() now also clears the local cache, not just the
persistent cache.
* Attempts to set data for LinkTargets that are not proper local pages
are ignored.
* All methods now accept LinkTarget|PageRecord as the key.
The ones that previously accepted a string still allow that as well.
* addLinkObject() now consistently uses the local cache if possible,
and consistently bypasses it if the forUpdate() flag is set.
This is all done in preparation for LinkCache being used inside PageStore.
Bug: T278940
Change-Id: I62107789fa185606a81be20ffa8f0be48297c08f
MediaWikiIntegrationTestCase::overrideMwServices() is not static
and should not be called statically
Change-Id: I9d1f07b5bca8a39b82017e59978f2b28c399803e
