Make the 'interwiki' option temporarily accept null as well as true/false
(and default to null instead of false) so form fields which disallow
interwiki titles can avoid deprecation warnings. This also includes
implementing the new error message.
The idea is that for the duration of the deprecation process, form fields
can be set explicitly to interwiki => false (which will eventually become
the default). It would be super annoying to have to do this for all
title fields, but this way it's only needed for fields where users actually
submit interwiki titles with some frequency.
Also improve the logic for legacy fields (which do not set the option
either way): since these (while emitting a warning) allow an interwiki
title, it does not make sense to apply namespace etc. checks to that
title.
Bug: T288155
Change-Id: Ic00f4a0f27747b5ff0893b4c01f42f68a99771ab
This seems to be a regression from 166ed5c.
DatabaseBlock::newFromTarget may return an autoblock for an IP
unless there is an existing manual block. If so, ignore it
because it is allowed to have an IP address both manually
blocked and autoblocked (see ipb_address_unique index).
Also add an integration test case.
Bug: T287798
Change-Id: I7c9a66ba0ffe759f43f4d0821c30fb94649d3dea
Make IResultWrapper inherit from SeekableIterator instead of Iterator.
As suggested by the interface documentation, throw an
OutOfBoundsException when the seek is out of bounds, instead of failing
in subtle ways.
Change-Id: I60427bc9057677254f1b06177ec525f1ae917721
Before MovePageTest was skipped if the move was valid,
claiming we can't test actual moves. Now we can.
Additionally, use MediaTestTrait for file and repo
mocking.
Change-Id: Ie8a1edbdb2f22432919f03a60c2dacc5d4528615
Additionally it switches the query from DB_PRIMARY to DB_REPLICA.
I understand the idea with a quick revert, but I do not think
it can be that quick - to revert a newest revision of a page,
an editor or a bot needs to actually read it first, and reads
come from a replica. So we know at least some replicas already
had the latest revision showing to the user. Very likely by the
time revert is made, we'd have it in all replicas. If not - oh well,
we can't be perfect. But we shouldn't really do such a query on
primary - it's too heavy.
Change-Id: I2fae8dbe5f19635f4d99e26242e3b08ddad8f8af
Cleanup after the switch of Database::query() to return ResultWrapper
instead of resource.
* Soft-deprecate the IResultWrapper accessors in IDatabase.
* Move relevant DBMS-specific functionality to ResultWrapper subclasses.
The deprecated methods in IResultWrapper become short and simple.
ResultWrapper is now abstract (b/c break).
* Move the implementation of fieldName(), numFields() and one of the
fieldInfo() implementations to the ResultWrapper subclass in order to
avoid ResultWrapper::unwrap() calls.
* Make Database::doQuery() return a ResultWrapper subclass instead of
underlying result data, so that the Database parent class does not
need to be aware of wrapper construction.
* Hard-deprecate ResultWrapper::unwrap(),
DatabaseMysqlBase::fieldType(), DatabasePostgres::fieldType().
* Fix the inefficient seeking method in SQLite.
* Make FakeResultWrapper extend ResultWrapper with an implementation
similar to the SQLite one. This is possible because ResultWrapper does
not depend on IDatabase anymore.
* Resolve fixme in DatabasePostgres: from studying the source,
neither pg_fetch_object() nor pg_num_rows() can set an error
retrievable with pg_last_error(). Removed unnecessary warning
suppression.
* ResultWrapperTest didn't make sense as a unit test anymore, so I
adapted it as an integration test against the current DBMS.
This change also means that ResultWrapper::key() always gives the
correct offset, even if Iterator methods are not being used.
Bug: T286694
Change-Id: I935835316c0bd7d3d061bd8fde9c9ce99ce756ec
Previously, BlockUser::placeBlockInternal attempted to insert
the new block with possible failure due to an existing block.
Regardless of the result, it would then retrieve the current
block which could be the just now inserted block. This one
was treated as the prior block and incorrectly provided to
the hook as such.
Refactor the method a little by retrieving the prior block
before attempting to insert a new one to make it more
efficient and to ensure the hook will have the correct value
for the prior block.
Change-Id: I70480ce32545f28f8317bb58f9c724d64cc54ccf
Returns an actual working ReadOnlyMode object
based on a ConfiguredReadOnlyMode and a mock
load balancer
Change-Id: I429c8e81343c24c536718b3fc112e797afbc05a2
Results in passing a user where previously the fallback
to $wgUser was being used, mostly in tests.
Bug: T255507
Change-Id: Iabe24315b23c0ad1272353186425e71974528d23
Replace all uses of AbstractBlock::getTarget with
Block ::getTargetName and ::getTargetUserIdentity.
Create AbstractBlockTest and two test cases for
AbstractBlock::getTarget and ::getTargetAndType.
It tests triggering of the deprecation warning.
Bug: T282247
Depends-On: I0543f363af66c57f5763b91320d87a69f23f9466
Change-Id: Iaeca824cac30172178de72f3cf7b7ae4cdd6f880
Image metadata is usually a serialized string representing an array.
Passing the string around internally and having everything unserialize
it is an awkward convention.
Also, many image handlers were reading the file twice: once for
getMetadata() and again for getImageSize(). Often getMetadata()
would actually read the width and height and then throw it away.
So, in filerepo:
* Add File::getMetadataItem(), which promises to allow partial
loading of metadata per my proposal on T275268 in a future commit.
* Add File::getMetadataArray(), which returns the unserialized array.
Some file handlers were returning non-serializable strings from
getMetadata(), so I gave them a legacy array form ['_error' => ...]
* Changed MWFileProps to return the array form of metadata.
* Deprecate the weird File::getImageSize(). It was apparently not
called by anything, but was overridden by UnregisteredLocalFile.
* Wrap serialize/unserialize with File::getMetadataForDb() and
File::loadMetadataFromDb() in preparation for T275268.
In MediaHandler:
* Merged MediaHandler::getImageSize() and MediaHandler::getMetadata()
into getSizeAndMetadata(). Deprecated the old methods.
* Instead of isMetadataValid() we now have isFileMetadataValid(), which
only gets a File object, so it can decide what data it needs to load.
* Simplified getPageDimensions() by having it return false for non-paged
media. It was not called in that case, but was implemented anyway.
In specific handlers:
* Rename DjVuHandler::getUnserializedMetadata() and
extractTreesFromMetadata() for clarity. "Metadata" in these function
names meant an XML string.
* Updated DjVuImage::getImageSize() to provide image sizes in the new
style.
* In ExifBitmapHandler, getRotationForExif() now takes just the
Orientation tag, rather than a serialized string. Also renamed for
clarity.
* In GIFMetadataExtractor, return the width, height and bits per channel
instead of throwing them away. There was some conflation in
decodeBPP() which I picked apart. Refer to GIF89a section 18.
* In JpegMetadataExtractor, process the SOF0/SOF2 segment to extract
bits per channel, width, height and components (channel count). This
is essentially a port of PHP's getimagesize(), so should be bugwards
compatible.
* In PNGMetadataExtractor, return the width and height, which were
previously assigned to unused local variables. I verified the
implementation by referring to the specification.
* In SvgHandler, retain the version validation from unpackMetadata(),
but rename the function since it now takes an array as input.
In tests:
* In ExifBitmapTest, refactored some tests by using a provider.
* In GIFHandlerTest and PNGHandlerTest, I removed the tests in which
getMetadata() returns null, since it doesn't make sense when ported to
getMetadataArray(). I added tests for empty arrays instead.
* In tests, I retained serialization of input data since I figure it's
useful to confirm that existing database rows will continue to be read
correctly. I removed serialization of expected values, replacing them
with plain data.
* In tests, I replaced access to private class constants like
BROKEN_FILE with string literals, since stability is essential. If
the class constant changes, the test should fail.
Elsewhere:
* In maintenance/refreshImageMetadata.php, I removed the check for
shrinking image metadata, since it's not easy to implement and is
not future compatible. Image metadata is expected to shrink in
future.
Bug: T275268
Change-Id: I039785d5b6439d71dcc21dcb972177dba5c3a67d
When an username cannot be a valid actor,
ActorStore::newUserIdentityByName() should return null instead of
throwing an Exception.
Bug: T283167
Change-Id: Id04978cf2a417f2354d5e336ed1b0b67462603e9
Make getServer() always return a string, as documented, even with new
Database::NEW_UNCONNECTED handles that have yet to call open(). If the
'host' parameter to __construct() is ''/null, getServer() now returns
'localhost' instead of null. This avoids problems like fatal errors in
calls to TransactionProfiler::recordConnection().
Use Database constants for "connectionParams" field keys for better
static analysis.
Also:
* Add Database::getServerName() method that returns "readable" server
names in the style of LoadBalancer::getServerName(). Note that the
"hostName" field is already passed in from LoadBalancer.
* Migrate most getServer() callers to getServerName() for easier
debugging and more readable logging.
* Also, normalize Database/LoadBalancer SPI logging context to use
"db_server" and reduce logging code duplication in LoadBalancer.
Bug: T277056
Change-Id: I00ed4049ebb45edab1ea07561c47e226a423ea3b
DatabaseBlockStore is capable of inserting cross-wiki
DatabaseBlock, which works via passing non-local IDatabase.
Ideally, we'd want to make Block a WikiAwareEntity, have
cross-wiki DatabaseBlockStore, etc. But as a quick-fix for
a regression we can just fetch correct ActorNormalization
Bug: T281972
Change-Id: I796f54b7b7303b8c442d5b9ba1926b76d3805fb5
JobRunner catches all exceptions and hides them in the status array,
meaning that it is not obvious when a job fails during a test case.
So, introduce MediaWikiIntegrationTestCase::runJobs(), which runs jobs
and asserts various things about the returned status array.
Depends-On: I4f4790c5d16a0767790eeff202e0be8fcdaeda93
Depends-On: I118f9e3f8950fd82d7b02baed6705b29fd6ab7d5
Change-Id: I63603aa158f77df4b40add096cb248f3b24979f4
Move MockTitleTrait::makeMockTitleCodec to DummyServicesTrait, and
replace the two existing uses, which are in core. Add some new
uses instead of mocking each time.
Unfortunately, we cannot use an actual MediaWikiTitleCodec
for the tests in BadFileLookup, because those tests are unit tests
and a MalformedTitleException cannot be created in the context
of a unit test. BadFileLookupTest gets around this by using
a mock that throws a mock exception - add a comment inline
explaining why we cannot use a real MediaWikiTitleCodec.
Paired with adding of NamespaceInfo to make mocking the language
methods related to namespaces easier by matching the real
logic in the Language class to the extend possible. Update a few
tests to use the DummyServicesTrait for their NamespaceInfo services.
Change-Id: Ibd691ccf0e632e1bf0bc1f7e9ddc0c660d5cad32
We don't actually know where is the given user name coming from
and whether it's been normalized before hitting the storage layer,
so always check if it's a valid user name.
Bug: T273933
Change-Id: Ie9848188c81c95bedff9d18e6398bd3511c459e4
This is an extremely dangerous method, but it's needed
for the temporary users magic in translate extension.
At least going via ActorStore we can keep our in-process
caches consistent and get rid of some static methods
and caching in user object.
Change-Id: I8157f7ccee7d72aee405e9b6109dfc1838e1f380
Most of the code in WatchlistManager does not depend
on integration logic, with the exception of
- ::clearTitleUserNotifications adding a DeferredUpdate
- ::addWatch and ::removeWatch using
User::newFatalPermissionDeniedStatus, which uses
MediaWikiServices
Move the majority of the rest to the unit tests, except
for an integration test of the overall basic functionality
that we leave to make sure that the integration with the
actual services works properly.
Also take this opportunity to document why the integration
tests left behind need to be integration test, and fix
the name of some tests to addWach/removeWatch instead of
the old doWatch/doUnwatch.
Explicitely add 'watchlist' to the tables used by the
integration test.
Change-Id: I899ef1618c59c179ed2591149b93969fd935276d
