Page MenuHomePhabricator

WikiLabels doesn't handle well revdeleted edits
Closed, ResolvedPublic

Description

@Halfak reporting an issue mentioned on the itwiki IRC channel:

19:06 <Vito> Elitre: I'd like to report that the labeling interface doesn't handle well revdeleted revisions
19:06 <Vito> it gives an empty page when the page is deleted
19:06 <Euphydryas> yes, I found 3 such edits too

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptMar 17 2016, 6:21 PM

@DarTar, do you have an example of screenshot and rev_id that I can test with?

@DarTar, do you have an example of screenshot and rev_id that I can test with?

Any random deleted rev works. Unfortunately I have no screenshots but well, the diff section was blank while buttons, title and rev id were ok.

Gotcha. Will review. Thanks for reporting.

@DarTar: Hey, I want to work on this, what do you want to happen if parent revision of an edit is deleted?

We should likely not even display a diff if the parent revision is unavailable. If we wanted to be a bit more intelligent though, we could compare to the most recent non-deleted parent. If we're doing that in Wikilabels, we'll want to do that in revscoring too.

In the short term, I think we should just notify the user that a diff is unavailable.

In the long term, I think we should file a task to consider the implications of diffs-across-deletions.

We should likely not even display a diff if the parent revision is unavailable. If we wanted to be a bit more intelligent though, we could compare to the most recent non-deleted parent. If we're doing that in Wikilabels, we'll want to do that in revscoring too.

IMHO that would be misleading, the diff should be showed to people with the relevant rights to see it.

+1 @Vituzzu I believe that will continue to work right now.

@Ladsgroup is working on implementing task abandoning T105521. That will allow labelers who lack the rights to "abandon" tasks on diffs they can't view and others with the applicable rights to pick them up. It isn't perfect, but this could work with a little coordination. Maybe we can set up constraints that check a users' permissions before a task is assigned to make sure they can complete it.

So, I think we should call this done when (1) tasks can be abandoned and (2) our API error messages are translated (i18n).

Halfak closed this task as Resolved.May 3 2016, 3:04 PM