Changes have been made in Git for this ticket:

MFB: Correctly match negative header values with numeric comparison in 
Sieve driver (bb@apc.ag, Bug #7308).

http://git.horde.org/diff.php/ingo/docs/CHANGES?rt=horde-git&r1=a50332ef341fa29c7acda0ed3e449a207ef0584d&r2=bade9b19e5062957438ae2935ef95cb653250b1a
http://git.horde.org/diff.php/ingo/lib/Script/Sieve.php?rt=horde-git&r1=9b22223f642b6ab65a71694ff34f709f06dfce9d&r2=bade9b19e5062957438ae2935ef95cb653250b1a
http://git.horde.org/diff.php/ingo/locale/ca_ES/help.xml?rt=horde-git&r1=34fde92a2cd38c530ede281396c13937d32f8554&r2=bade9b19e5062957438ae2935ef95cb653250b1a
http://git.horde.org/diff.php/ingo/locale/de_DE/help.xml?rt=horde-git&r1=34fde92a2cd38c530ede281396c13937d32f8554&r2=bade9b19e5062957438ae2935ef95cb653250b1a
http://git.horde.org/diff.php/ingo/locale/en_US/help.xml?rt=horde-git&r1=34fde92a2cd38c530ede281396c13937d32f8554&r2=bade9b19e5062957438ae2935ef95cb653250b1a
http://git.horde.org/diff.php/ingo/locale/es_ES/help.xml?rt=horde-git&r1=34fde92a2cd38c530ede281396c13937d32f8554&r2=bade9b19e5062957438ae2935ef95cb653250b1a
http://git.horde.org/diff.php/ingo/locale/fi_FI/help.xml?rt=horde-git&r1=34fde92a2cd38c530ede281396c13937d32f8554&r2=bade9b19e5062957438ae2935ef95cb653250b1a
http://git.horde.org/diff.php/ingo/locale/fr_FR/help.xml?rt=horde-git&r1=34fde92a2cd38c530ede281396c13937d32f8554&r2=bade9b19e5062957438ae2935ef95cb653250b1a
http://git.horde.org/diff.php/ingo/locale/lv_LV/help.xml?rt=horde-git&r1=34fde92a2cd38c530ede281396c13937d32f8554&r2=bade9b19e5062957438ae2935ef95cb653250b1a

Tweaked and committed, thanks!

Changes have been made in CVS for this ticket:

http://cvs.horde.org/diff.php/ingo/docs/CHANGES?rt=horde&r1=1.55.2.122&r2=1.55.2.123&ty=u
http://cvs.horde.org/diff.php/ingo/lib/Script/sieve.php?rt=horde&r1=1.63.10.37&r2=1.63.10.38&ty=u
http://cvs.horde.org/diff.php/ingo/locale/ca_ES/help.xml?rt=horde&r1=1.1.2.2&r2=1.1.2.3&ty=u
http://cvs.horde.org/diff.php/ingo/locale/de_DE/help.xml?rt=horde&r1=1.1.4.3&r2=1.1.4.4&ty=u
http://cvs.horde.org/diff.php/ingo/locale/en_US/help.xml?rt=horde&r1=1.9.4.3&r2=1.9.4.4&ty=u
http://cvs.horde.org/diff.php/ingo/locale/es_ES/help.xml?rt=horde&r1=1.1.2.5&r2=1.1.2.6&ty=u
http://cvs.horde.org/diff.php/ingo/locale/eu_ES/help.xml?rt=horde&r1=1.1.2.1&r2=1.1.2.2&ty=u
http://cvs.horde.org/diff.php/ingo/locale/fi_FI/help.xml?rt=horde&r1=1.1.2.7&r2=1.1.2.8&ty=u
http://cvs.horde.org/diff.php/ingo/locale/fr_FR/help.xml?rt=horde&r1=1.1.2.2&r2=1.1.2.3&ty=u
http://cvs.horde.org/diff.php/ingo/locale/lv_LV/help.xml?rt=horde&r1=1.1.8.3&r2=1.1.8.4&ty=u

This is a patch adding a warning to the Ingo help for the English and 
German help.xml files

OK, here is a new version of the patch, this time it's the Relational 
function that's changed. The change in the spam function was reverted 
because creating the Relational match has the same effect.



For comparisons <=, < and !=, the normal comparison is ORed with the 
check for the minus sign so negative numbers in the mail header match, 
too.



For comparisons >=, > and =, the minus sign in the mail header causes 
a non-match.



It's still not possible to use negative numbers for the comparison 
(i.e. enter them in Horde and hope that they'll work). But at least 
all cases of negative numbers in the mail header are supported as long 
as the entered comparison value is positive.

We can still check for negative values in the headers and ignore the 
rules in this case, like you do it now with the spam rule patch.

That's right. The problem is, as Sieve simply doesn't compare negative 
numbers correctly, there's no way to really fix it. If the header or 
comparison value are negative, the result is undefined.



See RFC (from January 2008): http://tools.ietf.org/html/rfc5231

It's the standard that says that the comparison doesn't support 
negative numbers, so probably this won't be fixed soon. Maybe Horde 
should display a warning when using the Sieve backend.



It works for the spam case because the reasonable range of 
X-Spam-Score is positive and we can assume that mails with a negative 
score won't be spam regardless of the score.

The patch looks fine, but only takes care about the spam filter. If 
you made this work for the Relational test instead, all numeric 
comparisons would take advantage of the fix.

When using the Sieve backend in Ingo and setting numeric comparison of 
X-Spam-Score, Ingo generates a rule that compares the X-Spam-Score 
numerically with the entered value (e.g. 5). The problem is that this 
comparison fails with negative spam scores (which can be quite common) 
and the e-mail is moved into the spam folder. See, e.g.:

http://tomster.org/blog/archive/2004/12/15/spamfiltering-using-spamassassin-and-sieve



The rule will be correct if it checks for the minus sign.



Wrong:

if header :value "ge" :comparator "i;ascii-numeric" "X-Spam-Score" ["6"] {



Correct:

if allof ( not header :comparator "i;ascii-casemap" :contains 
"X-Spam-Score" "-", header :value "ge" :comparator "i;ascii-numeric" 
"X-Spam-Score" ["6"] ) {



Working patch attached.