Don’t fail the test immediately when createMultiClient() or
createGuzzleClient() is called – instead, return service instances that
will fail the test as soon as any method is called on them. This is a
bit friendlier for dependency injection.
Bug: T272998
Change-Id: I79f90ef028867920f6a228991eda59ed4aeadb62
This allows separating notifications from personal tools.
The notifications are still inserted into the personal tools,
after the userpage, for skins using BaseTemplate, skins that
call buildPersonalUrls without the argument, skins that call
either getStructuredPersonalTools, makePersonalToolsList without
providing it personal tools or getPersonalToolsList.
Mustache skins that use data-personal are unaffected, and can
retrieve personal tools without notifications from data-user-menu
and notifications from data-notifications, both of which are in
the data-portlets array.
Notifications are manually inserted in both SkinTemplate and
SkinMustache to prevent calling buildPersonalUrls multiple times.
For backwards compatibility with user code and gadgets, the new
user-menu portlet uses the same id and classes as the personal
tools, allowing it to serve as a drop-in replacement.
Skins shouldn't output both.
Bug: T266613
Depends-On: Ib4112364c173952eb363e52756f03693a2e03512
Change-Id: Ia1451e3e802441162eecfc5b7f6a7ba2ae72f377
Update SQL, REST, and redis subclasses to emit call count and
payload size metrics for cache key operations. These metrics
are bucketed by cache key collection (similar to WANCache).
Bug: T235705
Change-Id: Icaa3fa1ae9c8b0f664c26ce70b7e1c4fc5f92767
Fix "break" statement in getAnyOpenConnection() so that the method no
longer returns false in cases where an IDatabase should be returned.
The prior bug caused doWait() to return false and set "lagged replica
mode" in some cases.
Remove $open parameter from doWait() to avoid subtle bugs. All callers
either passed true or checked if the index already had a connection in
the pool themselves (in which case true would not hurt). If somehow, no
connection exists, it is better to connect and log rather than
Additional cleanups:
* Add more "lagged replica mode" debug logging
* Add READER_INDEX_NONE constant to improve readability
* Add getStreamingReplicaIndexes() method to reduce code duplication
and to avoid replication waits on static clone servers
* Rename "connectionAttempted" field to "chronologyCallbackTriggered"
* Improve various bits of documentation
Bug: T252564
Change-Id: If24736d34c2755ca69ae10b4cc1967e7311fe400
This reverts commit 066e77735e.
Reason for revert: Objections brought up on ticket by Esanders are in alignmenet with own objections, that arose after initially providing this patch. We need clarification that is better addressed in new patch, reverting this.
Bug: T270796
Change-Id: Id873b3e3b7efb3003daa4986225b1fb3e7e218c8
The alphabets are based on what is used in
"Dictionnaire Général de la Langue Amazighe Informatisé"
by IRCAM (https://tal.ircam.ma/dglai/lexieam.php, DGLAi).
This was also requested by the community in the Tashelhit
Wikipedia Incubator.
Changes:
* Tests are enhanced to cover the whole alphabet.
* The Latin letter š is replaced with the letter c as
the equivalent of Tifinagh letter ⵛ.
* The Tifinagh letters ⵠ and ⵒ are eliminated
because they are not used in the Moroccan version
of Tifinagh, as presented in the DGLAi.
* The Latin letters O, P, V are converted to the
Tifinagh letters ⵓ, ⴱ, ⴼ, which are the same letters
that correspond to the Latin letters U, B, and F.
* Greek Gamma and Epsilon characters are replaced with
the corresponding characters from the Latin range.
* All the non-ASCII Latin letters are added to
the uppercase-lowercase Latin conversion.
This is the first patch in a series to fix the most important
issues listed above. It's kept minimal to make reviewing easier.
There will be more patches to fix readability and public/private
members, and to add more tests.
Change-Id: I7134216457b12018fd187ca7200e45c1b5a67471
For MySQL/Sqlite:
- Drop default values of ipb_timestamp and ipb_expiry as part of
standardizing timestamp columns
For Postgres:
- Drop foreign key on two columns as approved by the RFC
- Set default of ipb_user
- Make three columns not nullable to sync with MySQL
- Change data type of two columns to BIGINT to be the same as MySQL
Bug: T230428
Bug: T164898
Bug: T42626
Change-Id: I2c5303d76c6ce059d7fef324a4521c6336c5b1f3
At the time, Parsoid only emitted the figcaption if a caption was
present. So, having the border set on the figcaption was leading to
some incomplete borders on figures.
However, since I856ee962b70cef1f8d49652396ea5264e11a8ade,
that's no longer the case. We can now switch back to displaying the
figcaption as a table-caption and don't have to rely on the hacky
`width: 1px` to have the figcaption width match the media in the figure
(see https://stackoverflow.com/a/6536025).
One problem with the 1px hack is that in T241392, TimedMediaHandler sets
the max-width of a video to 100%, which shrinks it down to that 1px.
The change in display for the figcaption necessitates a bit of trickery
to get the magnify clip in the right place. At the time of the reverted
commit, css for the magnify clip wasn't present.
Bug: T272186
Bug: T266149
Change-Id: If283c1ab200ca1049343cdac759d7a2b78306227
