Last modified: 2012-10-29 16:39:58 UTC
Easiest described by example: Oversighter #1 sees a load of vandalistic edits in some page history with possible defamation in the edit summary. So he/she checkboxes those edits and suppresses the edit summary for 20 revisions. Oversighter #2 reviewing a bit later notices that a couple of them don't have vandalistic edit summaries, but one of them did have vandalistic revision text. Modifies the suppression for 2 of those 20 revisions and passes it on for discussion. The consensus is to suppress the revision text on all of the edits. At this point, the "change visibility" option for the 20 revisions breaks. RevDelete recognizes some of the 20 already have that setting and results in "failure". What's happening is RevDelete needs all 20 revisions to be in the same state prior to an action on multiple revisions. if some of those revisions have been manually changed already, it can't cope, even if the requested change is to bring other revisions into line with them. What I think should happen is, when RevDelete is reached by selecting multiple revisions (either by using checkboxes, or clicking "change visibility" in a log), the handling needs to be changed slightly. 1/ perhaps use tri-state buttons at [[Special:RevisionDelete]], to specify for each field either "forcibly show this field, for all revisions listed", "forcibly hide this field, for all revisions listed", or "don't change this field for all revisions listed". Hopefully that would be a more intuitive result. A warning "These revisions are in an inconsistent state" + <usual summary of the revisions> + "do you want to force <field = value> for these revisions?" would be better than just "failure".
Looks like another regression from the refactoring tim made.
A further aspect to this bug report, posted for DerHexer: When RevDelete is used to set visibility or suppression for multiple revisions, it also needs tri-state boxes. It may be for example that 20 revisions need suppression added, but 6 of them have combinations of edit summary or revision text already hidden by admins before this. It would be unnecessarily annoying to not be able to set suppression on all 20 just because some already had RevDelete used on them. But tri-state buttons could fix that, also.
(To clarify: You can set suppression on all 20 in one action, but lacking tri-state buttons, this will also modify the visibility of all other fields on all 20 revisions to the specified value, which is unwanted.)
Done in r59949
