When I call a perl module via rlm_perl and don't undef %RAD_CHECK and %RAD_REPLY before exiting, rlm_perl duplicates some attributes contained within the hashes. For instance: At entry to rlm_perl instance: $RAD_CHECK{'Ldap-Group'} is an ARRAY: (GroupA, GroupB) After exiting the script, "Added pair Ldap-Group..." messages appear in debug output. If I call another perl script to dump the %RAD_CHECK hash, it shows: $RAD_CHECK{'Ldap-Group' } is an ARRAY: (GroupA, GroupB, GroupA, GroupB) If I "undef %RAD_CHECK" before exiting from the first perl module, the values are not duplicated. I did some analysis of the sequence of events and I believe this is what's happening: - rlm_ldap creates the Ldap-Group attributes on the check list with operator T_OP_CMP_EQ during authorize (Ldap-Group is a checkItem in my ldap.attrmap) - upon return from the perl script, rlm_perl calls pairmove to move the attributes from the RAD_CHECK, RAD_REPLY and RAD_PROXY_REPLY hashes back to the respective pairlist. - pairmove adds attributes to the destination list for operator T_OP_CMP_EQ (takes default case) which creates duplicates Is this expected behaviour of rlm_perl? If so, can it be put on the "to do" list for rlm_perl documentation updates (or is it there already and I missed it)? Also, this seems to imply that it's not possible to change or remove, at least, some types of check or reply attributes from within rlm_perl? Also, the wiki for rlm_perl states that it passes configuration pairs in %RAD_CONFIG. I don't believe this is true (the hash is empty and I checked the source for 1.1.2, 1.1.3 and the latest snapshot and it doesn't create that hash). Is this a feature that is "in the works" or is the wiki incorrect? I can supply debug output, radiusd.conf and scripts if necessary.. Thanks.