Here Diannaa is listed as having the "Max. text added", that's because she revdel'd a series of edits, and xtools compared the last visible diff with the first live diff after it, which is this. She in fact did not add any content, see the revision history. If there's some way to tell that revisions have been revdel'd, we should not take into account the surrounding diffs. I'm guessing the revdel'd edits are not in the revision table, even though it's only the content (and not the author, edit summary, etc) that's been scrubbed from view.
- Mentioned In
- rXTRMRBae26583752ec: Articleinfo: Use diff size of 0 if prev edit is missing rev_len
rXTae26583752ec: Articleinfo: Use diff size of 0 if prev edit is missing rev_len
rXTRae26583752ec: Articleinfo: Use diff size of 0 if prev edit is missing rev_len
- Mentioned Here
- T101631: rev_len should be available also for deleted revisions in database replicas
The real issue is that rev_len is not being replicated, even though the diff sizes of revdel'd edits are visible in production. See T101631 for more.
As a workaround, all we're doing here is checking if the previous revision has a rev_len of null, and if so, treat target revision as having a diff size of 0. This means if User:Foo made two edits but only the first one was revdel'd, the second still has a diff size of 0 even though it actually made a change. Until T101631 is resolved I don't see a way around this.