Versions: a file renamed while it is being saved carries the in-flight save into the new name's history #942

Open
opened 2026-10-16 21:15:52 +00:00 by Cordy · 1 comment
Owner

Found in the fix-922 Task 6 review (2026-10-16).

What happens: a save to a.txt has recorded its content but not yet published it. If a.txt is renamed to b.txt during that window, the rename moves the version chain, including the in-flight record, to b.txt. Then the save publishes to a.txt again. As a result:

  • b.txt's history lists a version that was never the content of b.txt;
  • the save shows up as a new a.txt whose own history is missing that record.

Nothing is lost, because every content stays readable somewhere, but the histories are confusing. Spec 2026-10-16-fix-922 §14 records this only for folder renames.

Expected (to decide when worked on): pick one of:

  • the rename waits for in-flight saves of the source to finish (same instance: the serializer; several instances: #742);
  • the save notices the move and records under the new name;
  • the save is refused with a conflict.

To do: decide the rule, add a twin test, and document it.

Found in the fix-922 Task 6 review (2026-10-16). **What happens:** a save to `a.txt` has recorded its content but not yet published it. If `a.txt` is renamed to `b.txt` during that window, the rename moves the version chain, including the in-flight record, to `b.txt`. Then the save publishes to `a.txt` again. As a result: - `b.txt`'s history lists a version that was never the content of `b.txt`; - the save shows up as a new `a.txt` whose own history is missing that record. Nothing is lost, because every content stays readable somewhere, but the histories are confusing. Spec 2026-10-16-fix-922 §14 records this only for folder renames. **Expected (to decide when worked on):** pick one of: - the rename waits for in-flight saves of the source to finish (same instance: the serializer; several instances: #742); - the save notices the move and records under the new name; - the save is refused with a conflict. **To do:** decide the rule, add a twin test, and document it.
Author
Owner

The same happens one level up (fix-922 spec §14): renaming a folder while a save inside it is in flight moves the chain together with that save's in-flight record. The save then publishes under the old path. The moved chain holds a record of content that never became current there.

This is harmless and nothing is lost. Whatever rule is chosen here should cover folder renames too.

The same happens one level up (fix-922 spec §14): renaming a **folder** while a save inside it is in flight moves the chain together with that save's in-flight record. The save then publishes under the old path. The moved chain holds a record of content that never became current there. This is harmless and nothing is lost. Whatever rule is chosen here should cover folder renames too.
Sign in to join this conversation.
No labels
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: Cordy/Cairn#942
No description provided.