Le 15 janv. 2014 à 15:33, Axel Luttgens a écrit : > [...] > > I'll try with 2.2.10 as soon as possible.
I anyway had to compile 2.2.10. ;-) No more dubious header removal. But but but... Tried again, but now without a leading space at the beginning of the "boundary=" line, i.e. with: Content-Type: multipart/alternative; boundary=20cf302234d5b8063c04efcd4318 I received this one into my mailbox: [...] Content-Type: multipart/alternative; Message-Id: <20140115150454.5A9E28FA2D3@ALMba.local> Date: Wed, 15 Jan 2014 16:04:49 +0100 (CET) From: me@some.domain X-UID: 287477 Status: X-Keywords: Content-Length: 357 boundary=20cf302234d5b8063c04efcd4318 Subject: sieve test 6 --20cf302234d5b8063c04efcd4318 Content-Type: text/plain; charset=ISO-8859-1 instead of: [...] Content-Type: multipart/alternative; boundary=20cf302234d5b8063c04efcd4318 Subject: sieve test 5 Message-Id: <20140115144803.1C9FB8FA17A@ALMba.local> Date: Wed, 15 Jan 2014 15:47:51 +0100 (CET) From: me@some.domain X-UID: 287476 Status: X-Keywords: Content-Length: 296 --20cf302234d5b8063c04efcd4318 Content-Type: text/plain; charset=ISO-8859-1 [...] So, is could well be that you really are receiving messages without the line-continuation character (the leading white space before "boundary="). On the other hand, I guess the dovecot/pigeonhole behavior isn't the most appropriate one when facing such a malformed message... After removal of the single-line sieve script, thus allowing for direct delivery into the recipient's mailbox, I get a similarly corrupted message. One could thus infer that the source of the message's massaging is on dovecot's side. HTH, Axel