Since shipping our most recent release using pam_unix instead of pam_pwdb, we've gotten reports from users having problems authenticating using NIS servers running on HP/UX. It looks like the the pw_passwd or sp_pwdp fields have additional data in them, which is marked by appending a comma to the real data (similar to the way the gecos information is formatted). Given that md5 crypt() allows commas in the salt, would the right way to work around this be to use strncmp() instead of strcmp() to compare the hashed values for md5 passwords, and to terminate the copy read from the file at the first comma for other (crypt/bigcrypt) cases? Nalin
