There have been reports of mails still not being sent out which was fixed by reopening the mail and sending it again (that is with the version that doesn't remove the stuck messages from the outbox)

One potential root cause could be a race condition during retrieveItems in the maildir resource (the item has been moved in the database, but the change has not been replayed, so the item sync ends up recreating the item that is still found on disk). This scenario aligns with the id of the item in the outbox being greater than the one in the sent mail folder and the missing attributes. It doesn't seem to align with the revision though.

item in local outbox: rev3, SEEN flag, size 1656, valid remoteid, no attributes
item in imap sent-mail: rev2, SEEN flag, SENT flag, size 870, no remoteid, has attributes

We just reproduced the problem on a local system (and that's the first time that I've seen that). The mail has been sent though, and simply reappeared in the outbox without attributes.

I confirmed that the affected items don't have any attributes at all in the database. Trying to send the item again worked flawlessly, so the question is how those items ended up in the outbox without any of the necessary attributes.

This problem could be a race condition (the mysql server is external, not on the local machine).

This scenario is not possible at the moment and because SeaFile has it's own WebDAV service I'm not sure there's a need to use iRony for that. You can configure iRony to provide CalDAV and CardDAV only if needed.

Assigning to @seigo for review of D108.

I still can't get guam work with Cyrus Murder, although it is not clear if that depends on if guam is talking to Murder or standalone server. Guam configuration is:

So it looks like the '*' wildcard doesn't work in a mixed version murder environment (backend 2.5.6 and frontend 2.5.7) doesn't work. When all cyrus-imapd instances are of the same version - no more segfaults.

All the tests which are in the iRony repository used to run through without errors. So if that changed on Winterfell, it should be looked at.

Thx, it seems to have 'fixed' the issue for now..
While i am at it:
Would it be beneficial for you guys to find the offending piece of client software?

I recently run iRony's tests on Kolab Winterfell, and it gave me 14 failed tests. It looks like we don't really use these for CI, are we?

Also, adding the TYPE attribute in v3 seems a reasonable change to me.

Yes, Apple clients also use "b" when exporting and synching contact photos although they can handle upper-cased values as well. But the way to go definitely is lower-cased.

In T1062#16492, @vanmeeuwen wrote:

In T1062#16491, @adomaitis wrote:

If I run

. GETMETADATA (DEPTH infinity) "*" (/shared/vendor/kolab/folder-type)

Against 2.5.7:

Protection error: Connection reset by peer

messages log file shows:

kernel: proxyd[17138]: segfault at 7ffffff8 ip 00007f0ef1d5dab4 sp 00007ffca2cf2aa0 error 4 in libc-2.17.so[7f0ef1d15000+1b6000]

I cannot reproduce this, but I'm running this on Santiago, and you are running Maipo.

In T1062#16491, @adomaitis wrote:

I can't confirm the issue with guam still displaying groupware folders from Cyrus Murder is fixed.
Detailed description:
I have murder cluster running with cyrus-imapd-2.5.6-8.1. Frontends and Backends have commented suppress_capabilities: METADATA config option.

I can't confirm the issue with guam still displaying groupware folders from Cyrus Murder is fixed.
Detailed description:
I have murder cluster running with cyrus-imapd-2.5.6-8.1. Frontends and Backends have commented suppress_capabilities: METADATA config option.
I have added yet another frontend cyrus-imapd-2.5.7-8.1 to the cluster (from Winterfel repositories). Guam service is running on port 9143 on this new machine as well.

The fix for this is in https://git.cyrus.foundation/D86

It is the frontend that proxies a GETANNOTATION command to the backend, even though the original command is GETMETADATA.

Investigating where the issue is exactly.

In T1082#16053, @machniak wrote:

"B" and "b" both work with iOS7 Contacts app.

Can you disable the use of compression in the client, or revert rG534d19b37ec57efe9754c6cf8a22a59cbdcd7b6f?

"B" and "b" both work with iOS7 Contacts app.

Updating the ticket to reflect the new semantics.

Backported, tagged, pushed, built and distributed.

In T1082#15730, @vanmeeuwen wrote:

Perhaps it has come from section 2.4.1?

Does the Guam proxy or the Cyrus IMAP frontend have anything to say on the matter? Guam should be requesting the metadata for each of the folders in the list. Telemetry can tell you if it indeed does.

Perhaps it has come from section 2.4.1?

I don't recall where that change came from and I can't find an according test/sample case. We should comply with RCF here.

As this contradicts with rId20afc78222279, would be nice to have feedback from Thomas.

Not sure if that is enough.

We need a more accurately detailed transcript of the IMAP sessions.