[856] in Kerberos-V5-bugs
Re: appl/bsd patches....
daemon@ATHENA.MIT.EDU (Ken Raeburn)
Fri Oct 14 09:21:35 1994
From: Ken Raeburn <raeburn@cygnus.com>
Date: Fri, 14 Oct 1994 09:20:56 -0400
To: "Theodore Ts'o" <tytso@MIT.EDU>
Cc: krb5-bugs@MIT.EDU
In-Reply-To: Theodore Ts'o's message of Fri, 14 Oct 1994 00:05:47 +0500 <9410140405.AA16539@dcl.MIT.EDU>
Date: Fri, 14 Oct 1994 00:05:47 +0500
From: Theodore Ts'o <tytso@MIT.EDU>
The problem is that you can't leave it unspecified since the comparison
must either be against 255 or -1. So you have to know whether or not
chars are signed or unsigned.
The comparison is against a value read from a char array and sign- or
zero-extended. The bytes stored in the array must be correct, and the
signedness must be consistent.
If we assume bytes are eight bits, then whether you initialize the array
elements with "255" or "-1" doesn't matter.
If we don't assume bytes are eight bits, then we have to figure out what
the support code is going to do with octets read from the network...
If you think I'm missing something, can you send me patches of how you
would modify krlogind in this case?
The "magic" array and the arrays and pointers Ezra was changing just need
to have the same signedness. I think leaving it unspecified in all cases
would work fine. Ezra's solution of making them all unsigned would work
too, but might be trivially less efficient on some systems.
