User Since

Sep 19 2014, 7:05 PM (633 w, 5 d)

Roles

Disabled

IRC Nick

^demon

LDAP User

Chad

MediaWiki User

😂 [ Global Accounts ]

I need someone to review the linked patch :)

Easy fix is to just set the repo to private on Github (just done). Then nobody will see it and send us warnings, and Gerrit replication doesn't need any special-casing to skip it.

From the looks of it, we can add OIDC as a second omniauth provider. We might need to use omniauth_auto_link_user to make sure the users match properly. @jbond could you have a look at this anytime soon? We're running up against a deadline of the new Gitlab due to be released in May which removes the CAS-based login entirely.

Per T327583

Per T327583

Per T327583

Per T327583

Per T327583

Per T327583

Per T327583

Per T327583

Per T327583

One subtask is still open, but closing this as the train was able to move forward with a partial fix.

We aren't creating new repos in Phab anymore.

In T309731#8045089, @Milimetric wrote:

• Delete https://github.com/wikimedia/mediawiki-services-aqs-deploy

(I don't have rights to do this)

• Disable rGAQD

(I also don't have rights to do this)

In T165631#8485123, @Dzahn wrote:

@ayounsi It would mean a considerable effort to recreate an entire LVS service, which we just recently shut down for Phabricator in a lenghty decom process (git-ssh.wm.org), just to have a relatively small improvement, that users can use the "nice" port 22 instead of 29418.

What if we used the provider-agnostic public-cloud?

This would involve us continuing to add Change-Ids onto commits even on Gitlab? No no, we don't want to recommend that. They're ugly and pointless.

In T99268#7830522, @Krinkle wrote:

I noticed today that one of the first commits in the Maintenance system actually had a class-based registry, and while we lost that at some point, it connected the dots for me to something we still do today: The maintenance scripts are in the autoloader and each script has logic to avoid execution if loaded outside their own entry point. This was I believe designed to be run via a wrapper one day.

From the original commit:

$wgMaintenanceScripts = array();

Register their maintenance scripts [class name] with the system

$wgMaintenanceScripts for extensions to add their scripts to the default list.

public static function getMaintenanceScripts() {
      global $wgMaintenanceScripts;
      return $wgMaintenanceScripts + self::getCoreScripts();
  }

http://mediawiki.org/wiki/Special:Code/MediaWiki/54225
https://gerrit.wikimedia.org/r/plugins/gitiles/mediawiki/core/+/a1c51e18af85a9ac464c5b555921e58ec422cd11%5E%21/

Weird! Thanks for spotting that.