Loading all messages with getSubitemList() takes about 10ms per
language and loads an array with ~20k elements. When messages in many
languages are requested, this causes an OOM.
So, use getSubitemList() only when isMainCacheable() is called from
loadFromDB(). Remove the second parameter in that case, since it was
always the same.
In the getMsgFromNamespace() case, use getSubitem() to check the
specific message for existence. Have the caller specify the language, in
order to share a subitem cache entry with usual previous
getMessageForLang() call.
Bug: T247223
Change-Id: I6369f307b6bf74bd4aeb1d6e4c41d6e59e403703
Add public, protected or private to function missing a visibility
Enable the tests folder for the phpcs sniff
Change-Id: Ibefce76ea9984c47e08c94889ea2eafca7565e2c
assertSame() is guaranteed to not do any type conversion. This can be
critical when acciden tially comparing, for example, 0 to 0.0.
Change-Id: Iffcc9bda69573623ba14af655dcd697d0fcce525
The way isMainCacheable() was used, it always returned false in
non-content languages, because it would try to find strings like
'hidetoc/fr' in the array of message keys (which contains strings like
'hidetoc').
The consequence of this was that MessageCache would check the database
for a MediaWiki:hidetoc/fr page even if it already knew that that page
didn't exist. This is a substantial performance hit when requesting lots
of messages, like when building version hashes for ResourceLoader's
startup module.
Follows-up 4fc5ba8bf8.
Bug: T228555
Change-Id: I20433175ca919acc1c995f4a9cd50ca53afcdd02
Probably good to start testing this in CI sometime soon.
The unit tests being changed here are:
* Hacky importing of a revision in MessageCacheTest, changed to use an
interwiki username.
* WikiPageDbTestBase will get null rather than 0 for ar_user.
* DumpTestCase was using a custom IP $wgUser when creating its test
revisions, which was winding up referring to an actor_id that had been
deleted from the database. Fixed by having it use a registered user
instead.
Also this has MediaWikiTestCase reset the $wgUser User object when the
actor table is truncated, as various tests were otherwise failing because
it was caching a later-deleted actor ID for User:127.0.0.1.
Bug: T188327
Change-Id: Ic483d0fda931810ac639b9de1cc5e5a4ef8161cc
Depends-On: I1bd0288cc132627d75b4001219522ec5e952eda7
This re-enables the "flaky" testLoadFromDB_fetchLatestRevision which
was disabled because it failed on random patches for no apparent
reason, as documented in T219042.
The reason was an invalid timestamp in the test data, which caused
the timestamp to default to "now", which was "sometimes" newer than
the timestamp of the revision immediately created before.
Bug: T219042
Bug: T218918
Change-Id: I38f4816b2dc735a7698626c7e3fe4a93ec95a9d1
In Id044b8dcd7c, we lost a condition that ensured that the cache would
be populated with the latest revision. Now it was being populated with
all revisions, with a random one winning.
Bug: T218918
Change-Id: I1a47356ea35f0abf35bb1a3489d0d3442a3400a5
Follow up to a3d6c1411d.
This avoids extra queries for messages that have a software defined value.
Bug: T193271
Change-Id: I25aa0e27200a0b417721cf1fbd34a82095405b89
Prior to I462554b30, MessageCache::replace() did just that: it took the
existing cache and updated the one entry.
In I462554b30, the rearrangement of work into a DeferredUpdate
introduced a bug: the in-process cache was updated, but when the shared
cache was loaded later the entry was never updated in there so the
shared caches kept the old value. This was found in code review and
worked around by reloading all the messages from the database instead of
updating the existing cache.
But all that extra work reloading everything from the database causes
major slowness saving any MediaWiki-namespace page when the wiki has
many such small pages. Let's go back and fix the bug so replace() can
again replace instead of reloading everything.
Bug: T203925
Change-Id: Ife8e1bd6f143f480eb8da09b817c85aadf33a923
It's unreasonable to expect newbies to know that "bug 12345" means "Task T14345"
except where it doesn't, so let's just standardise on the real numbers.
Change-Id: I46261416f7603558dceb76ebe695a5cac274e417
Do the process cache update immediately (as before) but push
the shared cache updates to a deferred update. This update
will thus start with a clear transaction snapshot, so it can
acquire the lock before the first SELECT as is proper.
Also added some missing method visibilities.
Bug: T144952
Change-Id: I462554b300d4688b09ab80cd1bb8a4340ffaa786
This reduces the runtime of database-bound tests by about 40%
(on my system, from 4:55 to 2:47; results from Jenkins are
inconclusive).
The basic idea is to call addCoreDBData() only once, and have
a addDBDataOnce() that is called once per test class, not for
every test method lie addDBData() is. Most tests could be
trivially be changed to implement addDBDataOnce() instead of
addDBData(). The ones for which this did not work immediately
were left out for now. A closer look at the tests that still
implement addDBData() may reveal additional potential for
improvement.
TODO: Once this is merged, try to change addDBData() to
addDBDataOnce() where possible in extensions.
Change-Id: Iec4ed4c8419fb4ad87e6710de808863ede9998b7
Removed redundant set up in these classes (same as their paren
class MediaWikiLangTestCase does already).
* BlockTest
* ExportTest
* MWTimestampTest
* TitlePermissionTest
Change-Id: I28d18cb797bb249981727b02dffce4f0d8682b02
* The cache has to reload and *after* locking to avoid
losing any concurrent changes.
* Also fixed incorrect assumption in MessageCacheTest.
Message overrides for the content language do not use
the language suffix.
Change-Id: I98ff158a1575330bc59efe6badb27f8de8717951
- Swap "$variable type" to "type $variable"
- Fixed spacing inside docs
- Makes beginning of @param/@var/@throws in capital
- Changed some types to match the more common spelling
Change-Id: Ia041964250d8b7c0349d79dc9b131c5b8696e795
The core function behind wfMessage() (MessageCache->get()) did not
apply the language fallback chain to on-wiki messages.
This patch has changed the behavior to iterate over all possible
languages, first checking on-wiki and then checking the CDB cache,
until it finds the message. Note that fallback languages never
take precedence over the actual requested language.
This patch was taken from the following changes and then
adjusted to fix issues that caused bug 46579.
* Change-Id: Iaaf6ccebd8c40c9602748c58c3a5c73c29e7aa4d
- Author: Matthew Walker <mwalker@wikimedia.org>
- (cherry picked from commit d434bfcf3b)
* Change-Id: Ib607a446d3499a3c042dce408db5cbaf12fa9e3d
- Author: Mormegil <mormegil@centrum.cz>
- (cherry picked from commit 1b8cb8dc31)
Bug: 1495
Bug: 46579
Change-Id: I420457863eeb79824698d06abc7784032b267af2
This patch set reverts the following:
* Iaaf6ccebd8c40c9602748c58c3a5c73c29e7aa4d
* Ib607a446d3499a3c042dce408db5cbaf12fa9e3d
* Ic59fd20856eb0489d70f3469a56ebce0efb3db13 (partially)
Bug 46579 comment 17 describes a desired solution. In
If88923119179924a5ec091394ccab000ade16b3e we are working on a fix, but it is
taking longer than we anticipated. There was a deployment window planned
about now, but we didn't make it. It makes sense to revert for now, and commit
a proper solution somewhere next week.
Bug: 46579
Bug: 1495
Change-Id: Iac7ac4357dd80e8cdb238a6a207393f0712b3fe5
Also update some previous inconsistencies pointed out by Krinkle in change IDs:
* Ide20743a2e84ff68549286120e6cff9d9f396f54
* I811ca957b6588085d67606ebc0cd4033a1e53839
Change-Id: Ife33b931870d0d7e04fcb40974997436d27f528f
In Iaaf6ccebd8c40c9602748c58c3a5c73c29e7aa4d, the behavior of MessageCache
was changed so that it iterates over all possible on-wiki fallbacks including
the site language fallbacks. However, it omitted the site language itself.
Obviously, the problem appears only when the site language is not English,
because English is included always as the fallback language of last resort.
But in other cases, we need to add it explicitly.
Bug: 1495
Change-Id: Ib607a446d3499a3c042dce408db5cbaf12fa9e3d
MessageCacheTest::addDBData is called before setUp, so wgLanguageCode
and wgContLang are not set to 'en', which result in mediawiki pages
created as $key/en. When the test runs, the language is set correctly
and mediawiki is searching under $key, and does not find anything.
Set wgLanguageCode and wgContLang at the begin of addDBData
1) MessageCacheTest::testMessageFallbacks with data set #0
('FallbackLanguageTest-Full', 'ab', 'ab')
Message fallback failed.
Failed asserting that false matches expected 'ab'.
Change-Id: I8e70b7b424df6a9ad98bfb6c228809f4b4437d1e
The core function behind wfMessage() (MessageCache->get()) did not
apply the language fallback chain to on-wiki messages.
This patch has changed the behavior to iterate over all possible
on-wiki fallbacks (starting with the user's language) before
using the built-in language cache (CDB files). Previously we only looked
for the existence of an on-wiki message in the users's language.
Performance wise, using the 'ab' language ('ru', 'en' fallbacks)
MessageCache::get (Averaged over runs and calls)
New Code: ~8.5% TET (110us/call)
Old Code: ~6.5% TET ( 90us/call)
TET: Total Execution Time
Change-Id: Iaaf6ccebd8c40c9602748c58c3a5c73c29e7aa4d
