This was here from a year ago when we added CHROMIUM_FLAGS, which
has since rolled out.
Also make the assert-mw-env less fragile by checking the
env variables in one place each instead of two places. Before this,
it was relatively easy to accidentally add a warning for one, without
the shell exit code changing.
Change-Id: I6b85483981b415fe251b3760d5f282433e7fc4e9
* Get rid of a long regexp that had to be maintained and
was broken anyway, resulting in a false negative.
* Fix that false negative.
* Make the failures array a bit more readable.
Bug: T248075
Change-Id: I4e4e5d6487d23b0d64f29c113d84bddce758e516
This has come up a few times already with users (T248624#6008494,
T246827#5944100) so it would be useful to document the solution here.
Change-Id: I7c2fa16fd3f129e03c053935f32ebda3b589b9fe
This allows full configuration of the LoadMonitor, not just setting
the class name like loadMonitorClass allows. Deprecate the later.
Also fix obsolete LoadBalancer constructor documentation and
standardize the style of LBFactoryMulti::__construct() comments.
Change-Id: Icfdc0d6bbf227dce56e65ad679b93ce3f604e9f7
Old hook PageContentSave only received the main slot of the article.
In order for extensions to intercept changes in all slots, they should
receive a RevisionSlots object with new content.
PageContentSave hook is still preserved for backward compatibility.
Note: parameters $iswatch and $section (from PageContentSave) are
not added into the new hook, because they are obsolete in the old hook.
Note: new hook also doesn't unnecessarily pass-by-reference (T193950),
which was an issue with the old hook.
Bug: T205982
Change-Id: I55455639c5ff6ccbf3917d85c1ab0dac847ec853
UDPTransport by default will truncate the payload to 64k.
In order to allow a user of UDPTransport to format the payload
length to fit, the configured max payload length must be
accessible.
Change-Id: I5406eda588d2ee748872aecd8c64eb20afacea74
The code is appropriately logging when it encounters an exception.
However, the CI interprets that as an actual exception being thrown.
Setting the logger to null allows the CI to complete.
Bug: T248189
Change-Id: I9ce7c451b9951892e509afa048213e7691498e47
Ensure that the value returned is always also added to the cache, since
we (eventually) want to remove the hook's access to
$parser->mVarCache. Also don't allow monkeying with the value passed
as $magicWordId, even though it's a reference.
Code search:
https://codesearch.wmflabs.org/deployed/?q=ParserGetVariableValueSwitch&i=nope&files=&repos=
Bug: T236813
Depends-On: Ia12faefada7e4cf04f1a6b12b3ed1703bf28e437
Depends-On: Ia9e2e00c6b7d0e62cbe80c3b124165b691b1ff3d
Depends-On: I1e48fa47b8723958d543a69aaf9b62b872200dc5
Depends-On: If9a0885a8664c22473ade712364c2dd0a5c04e0b
Change-Id: I3d6b281f8e4e0bf68eefbf9767047527b4573b79
The existing magic word API (including hooks) gives implementers direct
access to $parser->mVarCache and full control of caching. Simplify the
implementation (prior to moving the magic word implementations out of
Parser.php) by always caching built-in magic words in $parser->mVarCache.
The cases which were skipping the cache don't vary anyway.
This behavior change is split from the rest of the refactoring in
I68fb1e786374e445b7df047934c532d7e10b8e94 to allow it to be deployed
separately, just in case there was some strange reason this would
cause regressions in practice.
Bug: T236813
Change-Id: I34a0d9a37e41c854c801d546427fc8cd0fe39dc4
CSRF tokens should only be required (and only be allowed) if
the current session isn't already inherently safe against
CSRF due to the way the authentication mechanism works.
This allows (and requires) tokens to be omitted for requests
that use an OAuth Authorization header.
Bug: T230843
Bug: T230842
Bug: T237852
Change-Id: Ib2922d556ff2470d4bf8c386c18986ca9f37d1b5
