Page MenuHomePhabricator

Confusing reply tool dirty diff (caused by history merge)
Closed, ResolvedPublic

Assigned To
Authored By
matmarex
Apr 24 2021, 8:30 PM
Referenced Files
F34563953: image.png
Jul 26 2021, 12:58 PM
F34564053: image.png
Jul 26 2021, 12:58 PM
F34422624: image.png
Apr 24 2021, 8:30 PM
F34422627: image.png
Apr 24 2021, 8:30 PM
Subscribers

Description

This reply tool diff looks really bad: https://en.wikipedia.org/w/index.php?diff=1019597647&oldid=1019597422

image.png (2×1 px, 399 KB)

But this is misleading. The actual changes made in that edit can be seen in this diff: https://en.wikipedia.org/w/index.php?diff=1019597647&oldid=1019596517

image.png (1×1 px, 167 KB)

The diff, and the page history, are all messed up because the history of two pages was merged into one:

See the logs here: https://en.wikipedia.org/w/index.php?title=Special:Log&page=Wikipedia%3ASockpuppet+investigations%2FDenizgezmis557761.

The original diff is actually a comparison of two completely different pages.

Event Timeline

(I filed this so that a link to this explanation shows up on https://dtcheck.toolforge.org/dtcheck-2021-04-24.html, there's no other action needed.)

ppelberg claimed this task.

This diff found in dtcheck is caused by the same scenario: https://en.wikipedia.org/?diff=1029083485

If it keeps occurring, maybe at some point I'll try to find a way to stop detecting these. Not now though.

matmarex moved this task from Ready for Sign Off to Doing on the Editing-team (Kanban Board) board.

Actually this seems easy to fix, the recent changes data includes the ID of the previous revision at the time the edit was made, and we should be using that ID instead of the ID of the previous revision right now.

Fixed by this change: https://github.com/MatmaRex/dtcheck/commit/38e6a36b3f21bf70310582aca9ad2b72a6fd4149

After re-running everything, https://dtcheck.toolforge.org/dtcheck-2021-07-23.html no longer includes the false positive:

BeforeAfter
image.png (2×3 px, 408 KB)
image.png (2×3 px, 164 KB)

The stats for 2021-07-23 are also corrected, and no other days changed unexpectedly.

Older occurrences in dtcheck will not be updated, because the recent changes data is only available for the last 30 days.