Note that this would introduce bypassing Lockdown without any checks that a user can access the undoafter revision.

In T307322#7909405, @Dylsss wrote:

Note that this would introduce bypassing Lockdown without any checks that a user can access the undoafter revision.

May be we could add a config like $wgThirdWayMergeWhiteNamespaces or $wgMcrUndoWhiteTitleRegex? like:

$wgThirdWayMergeWhiteNamespaces[ NS_TEMPLATE ] = [ NS_USER, NS_TEMPLATE, NS_DRAFT ];

allow "3-way merge into template namespace" which source from user namespace, template namespace, or draft namespace.

In T307322#7912878, @SunAfterRain wrote:

In T307322#7909405, @Dylsss wrote:

Note that this would introduce bypassing Lockdown without any checks that a user can access the undoafter revision.

May be we could add a config like $wgThirdWayMergeWhiteNamespaces or $wgMcrUndoWhiteTitleRegex? like:

$wgThirdWayMergeWhiteNamespaces[ NS_TEMPLATE ] = [ NS_USER, NS_TEMPLATE, NS_DRAFT ];

allow "3-way merge into template namespace" which source from user namespace, template namespace, or draft namespace.

I don't think this is really a good solution. Instead one of the following should be considered:

1. Properly check whether user have right to view pages related to the revision specified by undoafter.
2. (This does not require any change in MediaWiki) Use a gadget to copy a revision to another page.
3. (Even if there are proper permission control in MergeHistory, a feature may be useful to merge interleaved revisions) Use Special:MergeHistory to merge a revision to another page.

I think the 3-way merge workflow is cool, but it should be a dedicated special page with concrete security checks that verify each revision's visibility by the current user instead of subtly hacking it into action=undo in EditPage.