Proposal for modifying Get_Pwnam() [Re: couple of getpwnam() questions]

Gerald Carter gcarter at
Sun Nov 26 01:32:42 GMT 2000

Steve Langasek wrote:
> I think Gerald's question is still relevant: even 
> if there is a more efficient way to do what the 
> code currently does, *why* is it being done at all?

ok.  Here is my proposal for a change.  Psuedo code to follow...

struct passwd* Get_Pwnam(char* username, BOOL flag)

    1. attempt to lookup the username in all lower case
       if success

    2. attempt to lookup the username as transmitted (if 
       transmitted in mixed case).
       if success

    3. attempt to lookup the username in all caps
       if success

    4. perform number case permutations defined by 
       'username level' (defaults to 0)

    return failure;


This will catch non-NT systems than convert the username to
all caps.  In the event of a bad username, we will be 
saving 3 getpwnam() calls.  In the event of a good username,
and if the host systems follows the standard convention of 
usernames in all lower case (by far the most common case 
I think), we end up with only one getpwnam().   

What I really want to prevent is 5 lookups when a name
does not exist.  Sound fair?

As this is a big change in the semantics of how samba does 
a name lookup (is many sections of the code), I want to
give a fair chance for input from people before I go breaking
things :-)

Cheers, jerry
   /\  Gerald (Jerry) Carter                     Professional Services
 \/  VA Linux Systems   gcarter at       SAMBA Team          jerry at                     jerry at

       "...a hundred billion castaways looking for a home."
                                - Sting "Message in a Bottle" ( 1979 )

More information about the samba-technical mailing list