I have tried to have the tests skipped from the before hook using:

describe( 'Community Configuration Example Page', () => {

Quibble clones Vector unconditionally, that is hardwired in the code, similar to how we always clone mediawiki/vendor when Quibble is invoked with --packages-source=vendor: https://gerrit.wikimedia.org/g/integration/quibble/+/refs/heads/master/quibble/cmd.py#131

def _repos_to_clone(self, projects, zuul_project, clone_vendor):
    """
    Find repos to clone basedon passed arguments and environment
    """
    dependencies = set()
    dependencies.add('mediawiki/skins/Vector')  # <---------[ HERE ]-----------
    if clone_vendor:
        log.info('Adding mediawiki/vendor')
        dependencies.add('mediawiki/vendor')

That was done early on via, https://gerrit.wikimedia.org/r/c/integration/quibble/+/422553 because Selenium tests require a skin, other tests require or expect it as well.

I have tried to have the tests to be skipped in the before hook without using @cypress/skip-test since it is deprecated. I could not figure it out :-\ Thus I guess I should just add @cypress/skip-test. That will be for tomorrow.

In my original run I did not take in account browser tests. The Cypress testsuite in CommunitConfiguration requires ConfigurationExample, they should be skipped when it is not present.

Success!

Thanks @Lucas_Werkmeister_WMDE and @Reedy , it turns out on my copy of MediaWiki New Errors ECS had some extra filters which hiding the spam of logs :(

It is mostly quiet. I have left a comment as follow up to T411585 since another warning has surfaced: T411585#11431939

After deploying the patch there is another issue surfacing, I am adding it on the same task rather than filing one cause that sounds more or less related or at least under the same context. That is only appearing on ruwiki and we got only 160 of them over 90 minutes:

Thank you for the fix, I have deployed the backport before running the train.

As to not loose a couple of remarks that were made on the change and its backport, I am pasting them here:

When doing the backport to REL1_39 I noticed it had this.sandbox.stub( step.ui, 'load' ); which sidestep load.

There is a single line of code matching offset.top:

/**
 * Initialize this step.
 *
 * @param {mw.UploadWizardUpload[]} uploads
 */
uw.ui.Step.prototype.load = function ( uploads ) {
    // eslint-disable-next-line no-jquery/no-global-selector
    const offset = $( 'h1' ).first().offset();

If I look at https://ru.wikinews.org/wiki/Special:RecentChanges , at 13:56 @Ladsgroup has deleted a bunch of pages with the reason: Deleting obsolete stat pages ( https://ru.wikinews.org/w/index.php?title=%D0%92%D0%B8%D0%BA%D0%B8%D0%BD%D0%BE%D0%B2%D0%BE%D1%81%D1%82%D0%B8:%D0%A4%D0%BE%D1%80%D1%83%D0%BC/%D0%9E%D0%B1%D1%89%D0%B8%D0%B9&oldid=16808456#Are_these_pages_needed? )

The error has vanished from ruwikinews at 13:56 UTC:

I have filed:
T411580: SitemapFileHandler: PHP Deprecated: Implicit conversion from float 33333.333333333336 to int loses precision
T411585: PHP Warning: Stats: (action_api_modules_hit_total): Stats: (action_api_modules_hit_total) Cannot associate label keys with label values - Not all initialized labels have an assigned value.
Those do not seem to be worth blocking the train. I triaged them with Daniel this morning and he routed those tasks to the appropriate MediaWiki teams.

The Page::SitemapGenerator() requires int and it passed a float by SitemapGenerator which get a float on construction:

$this->sitemapSize = ( $apiConf['pagesPerSitemap'] ?? 10_000 ) / ( count( $variants ) + 1 );

I have dig into last week but missed updating this task. The reason is the composer-merge-plugin silently ignore the error, reset the state of ./vendor and let composer continue the execution.

Note from @colewhite, the scap logs are available at (requires NDA login) https://grafana.wikimedia.org/goto/XO75YxZDR?orgId=1

Scap has:

def increment_stat(self, stat, all_stat=True, value=1):
    """Increment a stat in deploy.*

In T406762#11410132, @jcrespo wrote:

These are the top files by size:

cumin2024@db1213.eqiad.wmnet[bacula9]> select Name, lstat_size(LStat) FROM File JOIN Filename USING(FilenameId) where JobId=666670 ORDER BY lstat_size(LStat) DESC LIMIT 15;
+-----------------------------+-------------------+
| Name                        | lstat_size(LStat) |
+-----------------------------+-------------------+
| git_file_diff.h2.db         |        2925682688 |
| comment_context.h2.db       |        1925754880 |
| account_patch_reviews.h2.db |        1427044352 |
| gerrit_file_diff.h2.db      |        1319897088 |
...

I have deployed the CI configuration which enabled phan static analysis for the repository. I have send an empty change and surely it failed to pass Phan:

src/CommunityConfigurationSchema.php:7 PhanUndeclaredExtendedClass Class extends undeclared class \MediaWiki\Extension\CommunityConfiguration\Schema\JsonSchema
src/CommunityConfigurationSchema.php:9 PhanUndeclaredConstantOfClass Reference to undeclared class constant \MediaWiki\Extension\WP25EasterEggs\CommunityConfigurationSchema::TYPE
src/CommunityConfigurationSchema.php:9 PhanUndeclaredConstantOfClass Reference to undeclared class constant \MediaWiki\Extension\WP25EasterEggs\CommunityConfigurationSchema::TYPE_STRING
src/CommunityConfigurationSchema.php:10 PhanUndeclaredConstantOfClass Reference to undeclared class constant \MediaWiki\Extension\WP25EasterEggs\CommunityConfigurationSchema::ENUM
src/CommunityConfigurationSchema.php:11 PhanUndeclaredConstantOfClass Reference to undeclared class constant \MediaWiki\Extension\WP25EasterEggs\CommunityConfigurationSchema::DEFAULT
src/Hooks.php:58 PhanTypeMismatchReturn Returning ['key'=>$key,'value'=>$value,'isEnabled'=>$isEnabled] of type array{key:'wp25eastereggs-enable',value:mixed|null,isEnabled:'0'|'1'} but getUserPrefEnabled() is declared to return array{key:string,value:mixed,isEnabled:bool}
src/Hooks.php:85 PhanTypeSuspiciousStringExpression Suspicious type bool of a variable or expression $userPrefEnabled['isEnabled'] used to build a string. (Expected type to be able to cast to a string)

In T411339#11419231, @Jdforrester-WMF wrote:

@hashar: Did you plan to back-port your fixes for release branches?

The Run buttons triggers a callback named recheckActionCallback() and defined at https://gerrit.wikimedia.org/g/operations/software/gerrit/+/refs/heads/deploy/wmf/stable-3.10/plugins/wm-checks-api.js#790

Found it! The JsonConfig extension has LUA integration tests which require Scribunto. To avoid failing when Scribunto is not loaded, the php class has an early return:

I am removing Math from the list of projects since that is a general issue.

Fixed by adding ParserFunctions as an explicit dependency of ProofreadPage. Thanks for the report!

Indeed I broke it when I disabled recursion on VisualEditor ( https://gerrit.wikimedia.org/r/c/integration/config/+/1211189 ) which removed a lot of dependencies. Previously the long list included ParserFunctions and was:

AbuseFilter
AntiSpoof
BetaFeatures
CentralAuth
CheckUser
CirrusSearch
Cite
cldr
CodeEditor
CommunityConfiguration
CommunityConfigurationExample
ConfirmEdit
DiscussionTools
Echo
Elastica
EmailAuth
EventBus
EventLogging
EventStreamConfig
FlaggedRevs
Flow
Gadgets
GeoData
GlobalBlocking
GlobalPreferences
GrowthExperiments
GuidedTour
IPInfo
IPReputation
LabeledSectionTransclusion
Linter
LoginNotify
MetricsPlatform
MobileApp
MobileFrontend
OATHAuth
PageImages
PageViewInfo
ParserFunctions
PdfHandler
Popups
Renameuser
Scribunto
SecurePoll
SiteMatrix
TemplateData
TextExtracts
Thanks
TimedMediaHandler
TorBlock
VisualEditor
WebAuthn
WikiEditor
WikimediaEvents
WikimediaMessages

I have updated the helm-lint jenkins job.

Thank you very much for your assistance @Dreamy_Jazz 🎉

In T409318#11412422, @Dreamy_Jazz wrote:

AFAICS they should be skipped when CheckUser is not installed

This task was for gerrit2003 which had a 55GB backup because, as part of a migration, the git repositories have been entirely rewritten. That specific cause has been resolved for now (but will trigger again next time the repos are wiped and regenerated via parent T387833).

It is stalled because that is merely a which list and it is definitely not a priority.

I am reopening this in order to address the root cause: the composer merge plugin does not abort / fail when it fails to install, continues execution which eventually leads to class not being found

Update to apply merge settings failed, reverting ./composer.lock to its original content.

CheckUser did not receive the update in REL1_39 which eventually led to Composer to refuse to install firebase/php-jwt due to the security issue.T411006

I have encountered that issue with ProofreadPage on https://gerrit.wikimedia.org/r/c/mediawiki/extensions/ProofreadPage/+/1193873 and ended up pushing the change.

Same issue on REL1_39 of ProofreadPage and spotted on https://gerrit.wikimedia.org/r/c/mediawiki/extensions/ProofreadPage/+/1193873

For the migration I think I will do it in small steps:

With word diff, the added BlueSpiceSocialTopics dependency stands out in green:

I think git diff --unified=0 --color-words will do

I have added a mapping for the MediaWiki installer with https://gerrit.wikimedia.org/r/c/mediawiki/core/+/1205175 , made my WikimediaBadges patch to depend on it and we will see what happens. I will then follow up on T258822. My use case is to have install.php --with-extensions to work :-]

And CI failed! Even though Wikibase has been successfully cloned, the installer can't find WikibaseClient or WikibaseRepository:

$ php maintenance/install.php --with-extensions
Error: A dependency error was encountered while installing the extension "WikimediaBadges": Could not find the registration file for the extension "WikibaseClient"

Thanks! Then I guess we can make WikibaseClient a requirement in extension.json. And indeed when I look at mediawiki-config:

// Load Client extensions
if ( !empty( $wmgUseWikibaseClient ) ) {
    if ( !empty( $wmgUseWikibaseWikimediaBadges ) ) {
        wfLoadExtension( 'WikimediaBadges' );
    }

For the projects that are on WMF production, 44 projects failing so far. I have filed some as subtasks, some others are due to Resource Loader dependencies hard wiring VisualEditor even when it is optional, I need to file another task for that 😊