Bad tests in COPY and MOVE commands

[shikharid]

These following tests have expectations nowhere defined in RFCs:

Test copy command 17/24 (line 32)

• This tests expect that UIDs assigned in the destination mailbox will be exactly in-order of msg UIDs copied from source mailbox
• For example: If msgs with UIDs 1,2,4 are moved from A->B, equivalent UIDs in B will be 1->1, 2->2, 4->3
• RFC3501 nowhere asks servers to behave this way and server can choose to copy these msgs in any arbitrary order
• A mapping of 1->2, 2->1. 3->2 is also valid
• This test can either be disabled as broken
• Or simplified by Just copying 1 msg at a time and doing the assertions instead of copying 3 at once

---

Test move command 12/13 (line 44)

• This test moves a msg in the same mailbox
• The test expects that a new UID will be assigned to the message
• Although the RFC 6851 does allow for moving msgs in the same mailbox
• But it does not specify that the server cannot optimise and consider it a NOOP
• So the assertion here is not reasonable
• The test should probably just look for a successful completion of the MOVE command

Again, happy to raise a PR fixing these. Let me know.

[sirainen]

COPY: I guess it's not prevented by RFCs, but I've a feeling it's not going to be nice client/user behavior. I wouldn't be surprised if it breaks some clients.

MOVE: Here I think even the RFC disagrees. It says MOVE is the same behavior as:

1. [UID] COPY
2. [UID] STORE +FLAGS.SILENT \DELETED
3. UID EXPUNGE

So the 1. step already tells it to assign a new UID, because two message instances can't have the same UID.

[shikharid]

For 2, you are right I guess.

For 1, I can't comprehend why. The simplest implementation for a client is:

• COPYUID <src-uids> <dest-uids>
• Each element in src uid msg set maps with corresponding entry in dest uid set

Any other way will be very convoluted, in fact I can't see how a client may do it any other way.

[sirainen]

For 1 I'm mainly thinking that if there are no other IMAP servers copying out-of-order, the existing client implementationa may never have tested whether it works. Anyway, I guess I could do some change in that test, but first I'd want to ask from imap-protocol mailing list whether someone else can think of a reason why the order would actually have to be preserved. IMAP RFC is sometimes hiding information in a bit random places.

[shikharid]

Makes sense, thanks! And I have an IMAP server copying out of order, hence the ask : )