[Dovecot] (BUG REPORT) sieve 0.1.14: vacation encodes Subject twice

[Christian Rohmann]

Hey folks,

allow me to bump this notice about a possible plain bug in the sieve
filter conversion.


Regards


Christian

On 12/30/2009 01:37 PM, Roland Rosenfeld wrote:
> Hi!
> 
> I noticed the following problem in dovecot-1.2-sieve-0.1.14:
> The following vacation causes trouble with the Subject line:
> 
> if not exists ["List-Id", "List-Post"]  { 
>     vacation :days 7 :addresses ["test at netcologne.de"] :subject "=?utf-8?b?w4TDlsOc?= sadasd" "Bin gerade nicht hier.";
> } 
> 
> As you can see, the Subject is already MIME encoded according to
> RFC2047.
> 
> Starting with sieve 0.1.14 the subject is now encoded by sieve again,
> so I get the following in my vacation mail:
> 
> Subject: =?utf-8?q?=3D=3Futf-8=3Fb=3Fw4TDlsOc=3F=3D?= sadasd
> 
> As you can see, the RFC2047 encoding is RFC2047 encoded, which results
> in an unreadable subject.
> 
> According to RFC5230 section 4.3 this shouldn't happen:
> 
>    The ":subject" parameter specifies a subject line to attach to any
>    vacation response that is generated.  UTF-8 characters can be used in
>    the string argument; implementations MUST convert the string to
>    [RFC2047] encoded words if and only if non-ASCII characters are
>    present.  Implementations MUST generate an appropriate default
>    subject line as specified below if no :subject parameter is
>    specified.
> 
> If I understand this right, sieve should do the RFC2047 encoding only
> if the string is non-ASCII.  But the RFC2047 encoded string is already
> ASCII, so it should not be encoded a second time.
> 
> 
> I just compared this with sieve 0.1.13, which behaves different: It
> does not encode the Subject header at all.  As long as the managesieve
> client (in our situation Horde/IMP) encodes the subject, everything
> works correct, while we run into trouble with 0.1.14.
> 
> I think that there needs to some mechanism in sieve that checks
> whether the subject contains non-ASCII and does the encoding only in
> this case.
> 
> I just checked out, that 0.1.14 does not encode a non RFC2047 encoded
> 7bit subject, while a 7bit subject starting with =?utf-8?... is
> encoded twice.
> 
> Tschoeeee
> 
>         Roland
>