You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Make ownership filter consistent between relink and rimport #32
As of #31, rimport walks directories to find files to publish with no regard for who owns the file. relink.py, on the other hand, still filters to just the owner. This is necessary, I think, because the rimport makes it so commands are run as the staging owner, and thus we would need special handling to check that ownership by the calling user is checked.
It may be acceptable to go either way on this: Both check ownership or neither does. It feels weird to have this obvious failure case possibly lead to situations where someone publishes someone else's file but then can't relink it, so I lean towards both checking.
To investigate: Is it even possible for one CSEG user to delete another's file in the input data dir? Yes. In fact, this is what rimport does when relinking! It's only a manual call of relink.py that skips files the user doesn't own.
I think the best thing to do would be to add a --include-files-i-dont-own flag or similar as an argument to both scripts. This makes it harder for someone to accidentally process someone else's files, while still keeping it possible.
Activity
samsrabin commentedon Sep 23, 2026
I think the best thing to do would be to add a
--include-files-i-dont-ownflag or similar as an argument to both scripts. This makes it harder for someone to accidentally process someone else's files, while still keeping it possible.