[16875] in Kerberos_V5_Development
Re: What is the intended calling convention on Windows between GSS
daemon@ATHENA.MIT.EDU (Greg Hudson)
Mon Jun 13 12:19:29 2011
From: Greg Hudson <ghudson@mit.edu>
To: Sam Hartman <hartmans@painless-security.com>
In-Reply-To: <tslhb7tpwzb.fsf@mit.edu>
Date: Mon, 13 Jun 2011 12:19:23 -0400
Message-ID: <1307981963.2281.197.camel@t410>
Mime-Version: 1.0
Cc: "krbdev@mit.edu" <krbdev@mit.edu>
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Errors-To: krbdev-bounces@mit.edu
On Mon, 2011-06-13 at 11:33 -0400, Sam Hartman wrote:
> I don't think MIT has shipped a working mechglue that can dynamically
> load mechanisms on Windows, although perhaps the 1.9 code has this
> functionality if built on Windows. It certainly has not been in an MIT
> KFW release.
Based on one user report, if you build 1.9 on Windows you don't get
functioning GSSAPI. I haven't looked into specifics. (When I got the
build system working again, I only tested kinit and klist.) I doubt
there's a compatibility issue here.
> The current trunk code does not decorate the function pointers in
> mglueP.h with any calling convention.
>
> I think that means that we'll use the C calling convention when calling
> through the function pointers into dynamic mechanisms.
>
> I'm guessing this is a bug.
Heimdal appears to use __stdcall on Windows in its mechglue/mech
interface. If we want it to be eventually possible to mix mechglues, we
probably need to follow suit.
I haven't been able to find any compelling reason for Windows DLLs to
decorate all of their APIS with __stdcall (NSS doesn't seem to do it,
for example), but it doesn't appear to be an uncommon practice.
_______________________________________________
krbdev mailing list krbdev@mit.edu
https://mailman.mit.edu/mailman/listinfo/krbdev
