Start the enablement of bounce detection.

* Move VERP header extraction to mailman.app.bounces, and ports the
  implementation to the get_verp() function.
* Restore the [mta] verp_regexp value which got lost along the way.
* Move mailman.email.utils to mailman.utilities.email and add a test for
  split_email()
* Move specialized_message_from_string() from test_documentation to
  mailman.testing.helpers.

1 files changed, 9 insertions, 1 deletions

diff --git a/src/mailman/config/schema.cfg b/src/mailman/config/schema.cfg
index 030e5fc2c..a0f4baaf7 100644
--- a/src/mailman/config/schema.cfg
+++ b/src/mailman/config/schema.cfg
@@ -423,7 +423,7 @@ delivery_retry_period: 5d
 # Note that your MTA /must/ be configured to deliver such an addressed message
 # to mylist-bounces!
 verp_delimiter: +
-verp_format: ${bounces}+${local}=${domain}
+verp_format: ${bounces}${verp_delimiter}${local}=${domain}
 
 # For nicer confirmation emails, use a VERP-like format which encodes the
 # confirmation cookie in the reply address.  This lets us put a more user
@@ -434,6 +434,14 @@ verp_format: ${bounces}+${local}=${domain}
 # $cookie   -- the confirmation cookie
 verp_confirm_format: $address+$cookie
 
+# This regular expression unambiguously decodes VERP addresses, which will be
+# placed in the To: (or other, depending on the MTA) header of the bounce
+# message by the bouncing MTA.  Getting this right is critical -- and tricky.
+# Learn your Python regular expressions.  It must define exactly three named
+# groups, `bounces`, `local` and `domain`, with the same definition as above.
+# It will be compiled case-insensitively.
+verp_regexp: ^(?P<bounces>[^+]+?)\+(?P<local>[^=]+)=(?P<domain>[^@]+)@.*$
+
 # This is analogous to verp_regexp, but for splitting apart the
 # verp_confirm_format.  MUAs have been observed that mung
 #