On 14.11.2011, at 11.51, Kai Arif wrote:
attached you'll find a patch for cmd-thread.c which addresses a problem with certain clients in conjunction with the use of the UID THREAD REFS command:
TAG5 UID THREAD REFS us-ascii SINCE 14-May-2011
- THREAD (0)(246)(247)(248)(249)(250)(252)(253)(254)(255)(256)(257)(258)(259)(260)(261)(262)(263)(264)(265)(266)(267)(268) TAG5 OK Thread completed.
The first UID delivered (0) is invalid, some clients - @mail for instance - then try to fetch the invalid UID 0 subsequently and run into problems.
Yes, it is. It should never happen.
The patch actually disables the return of (0) and seems to work for us. Please review and - if OK - incorporate it in the next version.
Could you instead send me such a mailbox where you can reproduce this problem? Probably sending dovecot.index, dovecot.index.log and dovecot.index.thread files would be enough. None of those contain any sensitive information.