Bug#362966: more occurences of nscd assert() failure

[Colm MacCarthaigh <colm@stdlib.net>]

We appear to have also been hit by this bug Firstly, the bug is
seemingly only triggered by GETPWBYNAME cached entities whose cached key
(ie what the process tried to getpwnam()) does not match pw->pw_name.

Looking at pwdcache.c shows that for records of this type, cache_addpw
creates a copy of the key for use with cache_add(), however cache_add()
itself (in cache.c) uses pointer arithmatic which makes assumptions
about the "key" argument and table->data coming from the same memory
allocation which I'm not certain are valid. 

The code is pretty poorly commented and arcane. But I'll try to pick
more up when I get time.

-- 
Colm MacCárthaigh                        Public Key: colm+pgp@stdlib.net