In SqlBagOStuff:
* Add modifyMulti() helper method to reduce code duplication
* Improve atomicity of add(), cas(), and changeTTL() queries
* Avoid integer serialization and improve atomicity of incr()
* Optimize new BagOStuff::changeTTLMulti() method
In BagOStuff:
* Add changeTTLMulti() method for subclasses to optimize
* Make set() ignore WRITE_ALLOW_SEGMENTS for integers so incr() works
* Strip WRITE_ALLOW_SEGMENTS flag from the setMulti() call in set() to
avoid triggering bogus sanity check exceptions
* Fix BagOStuffTest::testSetSegmentable failures via the above changes
* Enforce WRITE_ALLOW_SEGMENTS sanity check in setMulti() for all the
subclasses by using a final wrapper method
* Add WRITE_ALLOW_SEGMENTS sanity check to deleteMulti()
Bug: T113916
Change-Id: I25d1790fa9b0d1837643efccfa94a12043cfbf42
Add and use IDatabase::getServerConnection() method to avoid loops caused
caused by pickReaderIndex() calling getConnection() for the master server.
That lead to getReadOnlyReason() triggering pickReaderIndex() again.
Make getLaggedReplicaMode() apply when the master has non-zero load and
the replicas are all lagged.
Remove "allReplicasDownMode" in favor of checking getExistingReaderIndex()
instead. This reduces the amount of state to keep track of a bit.
Follow-up to 95e2c99094
Bug: T226678
Bug: T226770
Change-Id: Id932c3fcc00625e3960f76d054d38d9679d25ecc
* Unset globals to avoid tests that look like unit tests but actually rely on
globals
* move some tests out of unit directory so that the test suite will pass.
* Assert that tests which extend MediaWikiUnitTestCase are in a directory with
"/unit/" in its path name
Depends-On: I67b37b1bde94eaa3d4298d9bd98ac57995ce93b9
Depends-On: I90921679518ee95fe393f8b1bbd9134daf0ba032
Bug: T87781
Change-Id: I16691fc8ac063705ba0c2bc63b96c4534ca8660b
Mediawiki does not recognizes the mime type of an OpenOffice / Libre
Office Database Frontend Document and thus uses application/zip.
Our MimeAnalyzer::detectZipType() looks at the Zip files to detect known
signatures, however it is based on a reference from 2005 which did not
have the Database type yet.
OASIS Open Document Format v1.2 specify non normative MIME types, the
reference supposedly being the ones registered with the IANA, but we
have some dispredancy: ODF recommends 'base' while the IANA got
'database' registered: https://www.iana.org/assignments/media-types/
I supposed the format got approved as is, the Database mime type being
'recommended' probably because the IANA registration was pending.
Add the 'base' type which is being used by OpenOffice 3.x and later as
well as at least Libre Office 5.x.
Document my findings in code comments.
Add a test.
Bug: T35515
Change-Id: If0210a87067358612ecb8b3edd001fb05d01653d
Protect private wikis by providing basic read restrictions,
closely following the example of the action API.
The BasicAccess module provides a narrow interface for this
functionality, without exposing the whole session/user concept to the
router.
Also, add RouterTest and fix a bug in Router::getRelativePath() thus
discovered.
Change-Id: I82319d56f08b2eec4a585ff6dbd348ccdbadc5b5
MSCompoundFileReader::readFile uses iconv to convert information
given in UTF-16 character set with little-endian byte order to
the UTF-8 character set. The input string has no BOM and the byte order
is not explicitly given, causing iconv to try to guess the byte order
based on the host operating system. This causes the method to return
different results for the same file in different environments.
This patch explicitly provides the byte order for the input to be
converted (UTF-16LE) to ensure portability and predictability.
As part of this, move MSCompoundFileReaderTest into the unit test tree.
Bug: T225019
Change-Id: I62154897d303b28c288c3a4f2f5456bedcc81852
SpecialPageFatalTest:testSpecialPageDoesNotFatal is one of the slowest
tests (specially running Special:Version) due to the fact that it needs
to translate so many message keys.
3206ms to run SpecialPageFatalTest:testSpecialPageDoesNotFatal with data set "Version"
Running with lang=qqx would ensure that the special doesn't fatal but also avoid
unnecessary message translations
Change-Id: I8ff715ac539e93915c98f7209523df1b3ea3a7e8
Out of 150 tests of SanitizerTest.php, 100 of them are pure unit tests
they are moved to the new file in the new structure, the rest stay
Change-Id: I366d37607abff4bcd624a56fb8b2299729fbc088
Out of 140 tests of this file, 131 one of them are pure unit test
Let's keep the 9 in the original file and move the rest
Bug: T87781
Change-Id: I86dfe17f794c615048b3c20487b0e84d38d13b93
Make flushSnapshot() logging more detailed and check for explicit transaction
rounds. Also removing periods and trailing newlines from log/exception messages.
Change-Id: I0f6520f563680ab3a65b6338ced59ba25a2ec7b5
This fixes parameters returned by AbstractBlock::getBlockErrorParams,
but not those from ApiBlockInfoTrait.
Change-Id: I122017808766de1e9a9035f2f39a7b08607e56c1
Split out private resolveGroups() method for normalizing query
groups. Move some server index validation to getConnectionIndex()
and make it stricter.
Make getConnectionIndex() group fallback logic aware of the generic
group. The main use case for custom default groups is heavy scripts
or jobs that run in the background. It is probably better to have
good DB server redundancy for the query group and rely on it rather
than possibly fallback onto all of the main servers used for normal
requests. The later behavior could spread slowness or outages.
Also make getAnyOpenConnection() more robust and readable by
splitting out a private pickAnyOpenConnection() method that
checks IDatabase::trxLevel().
In addition, remove redundant is_int() check from isOpen()
method as it will return false in that case even without it.
Remove bogus getConnection() parameter in testCopyTestData().
Change-Id: Ica619c5487c761c724791d151db7388e4b41b0aa
In mediawiki.js, this marker has always been optional, falling back to
appending to <head>. When no stylesheets need to be after the marker
(e.g. no site styles on the wiki, and user is not logged-in), then
there is no need for the marker to exist.
In a previous refactor, I was going to do this and created an
"$append" variable in the function to do what this commit does,
but I forgot to actually use it for anything.
Test Plan:
* Local wiki, with no MediaWiki:Common/{Skinname}.css pages existing.
* When logged-out, before this change, there is a marker, now there is not.
* When creating "MediaWiki:Group-user.css" and logging in, there is still
a marker, and it is still above the <link> for that user styles request
in the <head>.
Bug: T219342
Change-Id: I2e9657f318088860916823efeb96ae4f1532974c
The SHA1 should be computed automatically if empty, not just if it is
missing entirely. This is needed since rev_sha1 may contain an empty
string for old revisions.
Bug: T200653
Bug: T219816
Change-Id: Ia6870a828bc9661fb05085e36315a86483ec48c4
The value of "all" can be completely arbitrary and in practice
it's usually an empty string. Avoid attempting to filter it out
and format anyhow.
Bug: T227202
Change-Id: I371466407bd038914faa4dc3b0cae9547cf427eb
This makes LogFormatter more robust against bad user names coming from
broken log records.
Bug: T224368
Change-Id: Ifce9e86bddba274182253cd25338766e8093d6ad
This causes Title::getTalkPage and NamespaceInfo::getTitle() to throw
an MWException when called on a LinkTarget that is an interwiki link
or a relative section link. These methods were already throwing
MWException when called on a link to a Special page.
Bug: T224814
Change-Id: I525c186a5b8b8fc22bca195da48afead3bfbd402
