Correction: Almost all pages are redirects

Change 752634 had a related patch set uploaded (by 4nn1l2; author: 4nn1l2):

[operations/mediawiki-config@master] hewikisource: remove \"\u05E7\u05D8\u05E2\" namespace and its talk page

https://gerrit.wikimedia.org/r/752634

Hello!

delete the pages there

Just to confirm: deleting the namespace will make its pages (both currently existing and deleted) permanently unaccessible without sysadmin intervention, meaning that usual tools like Special:Undelete don't work with them. Is this fine with you?

In T298430#7608745, @Majavah wrote:

Hello!

delete the pages there

Just to confirm: deleting the namespace will make its pages (both currently existing and deleted) permanently unaccessible without sysadmin intervention, meaning that usual tools like Special:Undelete don't work with them. Is this fine with you?

All the pages in the namespace have already been moved to their proper place. So yes, please proceed.

Change 752634 merged by jenkins-bot:

[operations/mediawiki-config@master] hewikisource: remove \"\u05E7\u05D8\u05E2\" namespace and its talk page

https://gerrit.wikimedia.org/r/752634

Mentioned in SAL (#wikimedia-operations) [2022-01-10T12:34:39Z] <taavi@deploy1002> Synchronized wmf-config/InitialiseSettings.php: Config: [[gerrit:752634|hewikisource: remove "קטע" namespace and its talk page (T298430)]] (duration: 00m 58s)

Hey everyone,
Since we deleted this namespace we have small issue.
There are redirects from the transfer of pages between namespaces that already appear on the site. They are still on Special:DoubleRedirects.

Removing the namespace has caused T356024: TypeError: Argument 4 passed to Wikimedia\Parsoid\Utils\Title::__construct() must be of the type string, null given, called in /srv/mediawiki/php-1.42.0-wmf.15/vendor/wikimedia/parsoid/src/Utils/Title.php on line 392. It seems like the namespace is still referenced in some places in the database. Unclear why it is showing up now, two years later...

@Nahum @taavi: There seem to a A LOT of pages still in that namespace, which are now inaccessible:

MariaDB [hewikisource_p]> select count(*) from page where page_namespace = 100;
+----------+
| count(*) |
+----------+
|    42922 |
+----------+

They are all redirects, but they are still pages, and mediawiki may find them in the table and try to do things with them, which will fail because the namespace is not defined...

We could just manually delete them and their associated revisions from the database. That's a bit hard, but it would work. Alternatively, we could make them show up in a different namespace.

If you make them show up in a different namespace, will there be an easy way of mass-deleting them with a bot? (we already have a bot or two with admin access, operated by one or two of our trusted admins). Or will they be lost in the namespace which they show up in? I guess we could designate a seldom-used namespace for them...

In T298430#9503312, @Nahum wrote:

If you make them show up in a different namespace, will there be an easy way of mass-deleting them with a bot? (we already have a bot or two with admin access, operated by one or two of our trusted admins). Or will they be lost in the namespace which they show up in? I guess we could designate a seldom-used namespace for them...

We could also temporarily restore the old namespace, then delete the pages, then remove it again.
But in a way, it only shifts the issue. We will then have entries in the archive table that references a non-existing namespace...

I don't really know a clean solution. Retiring a namespace is not something that our database schema supports.

Those references are the problem here, not the targets they reference. We need to list those "entries in the archive table that reference the non-existing namesapce" and have them fixed or deleted.