dovecot - Jan 2009 - Sieve regex match problem

I'm trying to make a regex to match common mailing list addresses and file
messages to
corresponding folders.

I'm using sieve-test to try and understand what is happening. The sieve
script is:

require [ "fileinto", "regex", "variables" ];
if header :regex ["Sender"]
   ["(.*>[ \\t]*,?[ \\t]*)?([^-@]*)-([^-@]*)(-bounces)?@antlr.org"]
{

   fileinto "${0} :: 1:${1} 2:${2} 3:${3} 4:${4} 5:${5} 6:${6}
7:${7}";
}

The email contains:

Sender: antlr-interest-bounces at antlr.org

I'm trying to form a folder from the first two parts of the sender. In the
real script,
that will do a "fileto antlr.interest;"

Running sieve-test, I get:

  * store message in folder: antlr-interest-bounces at antlr.org :: 1: 2:antlr
3: 4:interest
5: 6:-bounces 7:

I don't understand why parts 3 and 5 are empty. From the regex grouping, I
expected

Part 2 "antlr"
Part 3 "interest"
Part 4 "-bounces"

The regex groups break done as:
1: (.*>[ \\t]*,?[ \\t]*)?    - match leading name, comma or spaces
2: ([^-@]*)                  - match first part: 'antlr' in his case
-                            - Hyphen between parts
3: ([^-@]*)                  - match second part: interest in this case
4: (-bounces)?               - When cc'd, -bounces part isn't used

The same regex tested with perl works as I expected.

Without the first group of the regex '(.*>[ \\t]*,?[ \\t]*)?', it
works correctly. That
group isn't getting matched anyway, so it shouldn't matter.

This is with 1.2alpha5.

What am I missing?

Thanks,

	Andy

Andy Howell wrote:
> 
> Running sieve-test, I get:
> 
>  * store message in folder: antlr-interest-bounces at antlr.org :: 1: 
> 2:antlr 3: 4:interest 5: 6:-bounces 7:
Interesting use of the fileinto extension. ;) I'd better define a 
vnd.dovecot.debug extension to provide some sort of debug printing 
command. This would be an extension only provided by the sieve-test tool 
to provide such functionality more cleanly. This will print the values 
instantly in stead of waiting until the Sieve result is printed.
> Without the first group of the regex '(.*>[ \\t]*,?[ \\t]*)?',
it works
> correctly. That group isn't getting matched anyway, so it shouldn't
matter.
> 
> This is with 1.2alpha5.
> 
> What am I missing?
Nothing, this is a bug. If a non-matched match value is encountered, the 
indexes of the match values after that are mangled.

Fixed:

http://hg.rename-it.nl/dovecot-libsieve/rev/de6db6757418

As it turns out, I forgot to add the :regex match value handling to the 
test suite. I've now added a couple of simple tests based on your example.

Regards,

-- 
Stephan Bosch
stephan at rename-it.nl