Gerrit does weird stuff when cherry-picking a change from a branch to itself (to get rid of unnecessary unmerged dependencies). And if this is an unsupported use case, then it shouldn't be possible to do.
https://gerrit.wikimedia.org/r/#/c/74401/ used to depend on https://gerrit.wikimedia.org/r/#/c/74270/ (in PS1).
I used the "Cherry Pick To" function to cherry-pick it to master (it was already submitted to master), removing the "cherry picked from" line from the commit message.
Now PS2 it based directly on master's HEAD, but:
- There is no mention of PS2 being submitted in the comments section (only about PS1 being cherry-picked).
- It appears as if gerrit considered PS2's commiter to be TheDJ (PS1's commiter), not me. (Yuvi's grrrit-wm reported: "[16:56] <grrrit-wm> (PS2) TheDJ: Tablesorter: Add testcase for unsortable column [core] - https://gerrit.wikimedia.org/r/74401" - ordinarily it reports actual authors of patchsets; I also vaguely recall seeing this in gerritmail send from some other changeset, but I don't have it handy right now).
Version: wmf-deployment
Severity: normal