I am still investigating the problem.  The problem is not really caused 
by the fact the CREDENTIALS definition is different in KRB5_32.DLL or 
KRB4_32.DLL.  The problem is that building Leash has a conflict with 
attempting to bring in the correct header file since they have the same 
names.

I am still not sure whether or not that is the problem.  I am still 
investigating.



Sam Hartman wrote:

>>>>>>"Jeffrey" == Jeffrey Altman [Kermit Project] <Jeffrey> writes:
> 
> 
>     Jeffrey> How would it break the ABI?  The changes are Windows
>     Jeffrey> only. The Krb5 Krb4 library is not supported (nor
>     Jeffrey> shipped) on Windows at the present time.
> 
> 
> It's a little unclear to mea how much the krb5 krb4 library is or is
> not supported on Windows.  You yourself said that you had implemented
> support for this library in Kermit because people were using it.
> 
> KFW does not ship this library, but the krb5 sources do support
> building outside the KFW framework on Windows and in that
> configuration do build the library.
> 
> 
> When we contemplated the krb524 change, you were asked to review
> whether the difference in the credential structure would be a problem.
> You said that because the KFW krb4 library had a larger credentials
> structure and that because it was allocated by the caller, no problem
> would result.
> 
> The fact that we were concerned about the difference in the
> credentials structure suggests to me that we are at least reluctant to
> change it for 1.3.