RE: Squid 2.5 + NTLM on Solaris 8 SPARC

Thanks Robert,
to save you a little time:
This could appear in two main places.
1) If the BASE64 encoded auth strings are wrong (they shouldn't be), the
in-squid code will get confused. It should spit errors if this was the
case, but it's worth a _quick_ eyeball.
2) The ntlm helper is at fault. We are (I thought) using the correct n2h
and h2n calls with this, but you never know.

For debugging you might try the fakeauth helper. That never talks to the
network. If that works, and shows the correct username, then the ntlmssp
helper is definately at fault and we've eliminated 1).

Cheers,
Rob

> -----Original Message-----
> From: Robert Moss [mailto:rmoss@mbox.com.au]
> Sent: Thursday, October 25, 2001 3:26 PM
> To: Robert Collins
> Subject: Squid 2.5 + NTLM on Solaris 8 SPARC
>
>
> Heya,
> Been hunting and searching for posts from people with success with
> Squid+Solaris8+NTLM .. very few with solutions, many with unresolved
> problems.
>
> Having some problems here too, and going through the debugged
> NTLM Auth
> packets has shown the problem to be a Little/Big endian problem.
>
> x86 running FreeBSD with squid-head-200109262300 works fine
> SPARC running Sol8 with squid-head-200109262300 same compile flags
> does not work, and the structure of the NTLM packets are reversed from
> left to right, which is a common sign that the little-big endianisms
> between archetecture isn't being taken into account.
>
> We're working on fixing it through changing the source code, and will
> post a patch.. This is just a heads up :-)
>
> Symptoms:
> NTLM Authentication failing without requesting username/password
> NTLM Authentication failing at stage 2: after challenge packet sent by
> squid, nothing is returned by the web browser and connection drops
>
> Attatched is the debug_options ALL,1 29,9 output of the FreeBSD and
> Solaris8 box
>
> Cheers,
> Robert Moss
>
>
Received on Wed Oct 24 2001 - 23:34:49 MDT