Re: Compile error in current CVS, src/server.c:5500

[Derek Robert Price]

Brian Murphy wrote:

>     if (retval == PAM_SUCCESS)
>         retval = pam_authenticate(pamh, 0);
> +    else
> +       printf("E PAM Error: %s\n", pam_strerror(NULL, retval));
>
>     if (retval == PAM_SUCCESS)
>         retval = pam_acct_mgmt(pamh, 0);
> +    else
> +       printf("E PAM Error: %s\n", pam_strerror(pamh, retval));

You'll need to nest the conditionals or set some sort of "skip to end" 
variable.  Perhaps "goto error_exit;" or the like would be more 
appropriate.  The way you have things set up in this patch, an error in 
pam_start will cascade through to print two error messages.  After this, 
pam_end will be called, perhaps with a NULL or undefined pointer, though 
perhaps that is the correct usage since pam_end will spot a bad retval?

I'd also like to see a "CVS report" of what it was trying to do when the 
PAM_* error was received.  Something along the lines of "PAM returned an 
error during authentication: %s\n", pam_strerror(...).  Ditto for 
pam_start.  Ditto for pam_acct_mgmt, which I notice you are not printing 
the pam_strerror for - it's conditional is printing out the error 
message, if any, for the call to pam_authenticate.

> -    if (pam_end(pamh,retval) != PAM_SUCCESS)
> +    if ((err = pam_end(pamh, retval)) != PAM_SUCCESS)
>     {
> -       printf("E Fatal error, aborting.\n
> -               pam failed to release authenticator\n");
> +       printf("E Fatal error, aborting.\n\
> +error 0 pam failed to release authenticator\n\
> +PAM error number %d: %s\n", err, pam_strerror(NULL, err));
>         error_exit ();
>     } 
>  

What exactly is the "error 0" you cite?  I'd rather CVS not start 
outputting arbitrary error numbers without good reason.  Also, if 
pam_strerror is called on the output of pam_end, I don't think printing 
the actual error number matters much.  It shouldn't matter to an end 
user and likely not even to us with a good text message.  Then again, 
and please feel free to correct me on this one, maybe we should leave 
the error number in there through the experimental stage, until we 
discover that the text messages in error reports are satisfying?

Derek

--
               *8^)

Email: derek@ximbiot.com

Get CVS support at <http://ximbiot.com>!
--
"You use a Windows machine and the golden rule is: Save, and save often.
It's scary how people have grown used to the idea that computers are
unreliable when it is not the computer at all - it's the operating system
that just doesn't cut it."
                               -- Linus Torvalds