How unstable is the CVS source?: smbpasswd

[Lonnie J. Borntreger]

I just got the latest CVS and the entire getpwnam interface has been
rewritten, so you can forget the patch.  The SIGBUS no longer happens
(although my login/connection attempts always fail now ;)

Lonnie Borntreger

> -----Original Message-----
> From: java at bitserv.vpnet.at [mailto:java at bitserv.vpnet.at]On Behalf Of
> Bertl
> Sent: Wednesday, July 07, 1999 7:15 AM
> To: lonnie at borntreger.com
> Cc: D.Bannon at latrobe.edu.au; SambaList (E-mail)
> Subject: Re: How unstable is the CVS source?: smbpasswd
>
>
> "Lonnie J. Borntreger" wrote:
> >
> > This was covered in the thread "SIGBUS Panic in smbd".  I
> submitted a fix to
> > Luke who is having some people test it before it's posted.
> (see attached
> > message and diff).
>
> Hi!
>
> I don't think the diff is right: When USE_HASHED_GETPWNAM is
> set, the pw_passwd field is allocated with strdup (line 145),
> so this looks like a (very small) memory leak...
>
> I tried the following and it seems to work with hash and without:
> (btw: why is pw-hashing not set by default?)
>
> --- username.c.orig     Thu Jul  1 16:01:52 1999
> +++ username.c  Wed Jul  7 13:38:23 1999
> @@ -228,9 +228,8 @@
>
>  struct passwd *hashed_getpwnam(const char *name)
>  {
> -#ifndef USE_HASHED_GETPWNAM
> -  return getpwnam(name);
> -#else
> +  struct passwd *ret;
> +#ifdef USE_HASHED_GETPWNAM
>    struct passwd_hash_table_s *pht=&passwd_hash_table;
>
>    DEBUG(5,("getpwnam(%s)\n", name));
> @@ -259,8 +258,11 @@
>      return NULL;
>    }
>    /* Fall back to real getpwnam() */
> -  return getpwnam(name);
>  #endif
> +  ret = getpwnam(name);
> +  if(ret != NULL)
> +    ret->pw_passwd = strdup(ret->pw_passwd);
> +  return ret;
>  }
>