svn commit: samba r18875 - in branches/SAMBA_3_0/source/groupdb: .

[tridge at samba.org]

Simo,

 > > It also will not affect anything we do as an ldap server.
 > 
 > We must return an error as LDAP server when someone sends us a message
 > with an empty attribute.

yes, and with my patch we will still do that. The ldap server is
completely unaffected (it never calls this function).

 > > how could the ldb_msg_add_string() change hide a real bug? Do you
 > > really think we have code that relies on getting back an error when an
 > > empty string is added to a ldb message ?
 > 
 > Nope, but if you don't set the value by mistake, and not by choice you
 > may not notice it (unless using an ldap server and the attribute is
 > mandatory).

Maybe you can give me an example of how someone could construct a bug
around this? 

I certainly have an example of an existing bug that happened because
of the current behaviour. I'm pretty sure there are more. I can't
however see how to construct a bug with the patch I proposed :-)

Cheers, Tridge