> In article <rmiobqinusk.fsf@...>,
> Greg Troxel <gdt@...> wrote:
>> fix the USE_KERBEROS=no USE_PAM=no build.
>>I think that's the wrong fix - it declares a variable unconditionally,
>>when it seems that the bug was referring to the variable in the
>>non-KERBEROS5 case. Do you think that the instance variable is
>>needed/useful in the non-KERBEROS5 case?
> Yes, have you looked at the code in HEAD?
Yes, I did look, at 1.100 of login.c. But now I see that instance is
used to assign ptr, in addition to two stores to instance that look like
they should be ifdef KERBEROS5.
Overall, it looks like residual confusion from making kerberos
conditional, or at best undercommented - the use of a variable named
"instance" for non-kerberos purposes at the same time as kerberos
purposes should be explained.