[2574] in Kerberos_V5_Development
Re: Prototype hell
daemon@ATHENA.MIT.EDU (Ken Hornstein)
Tue Oct 14 14:57:29 1997
To: krbdev@MIT.EDU
In-Reply-To: Your message of "Fri, 10 Oct 1997 20:47:49 EDT."
<199710110047.UAA27397@dcl.MIT.EDU>
Date: Tue, 14 Oct 1997 14:56:22 -0400
From: Ken Hornstein <kenh@cmf.nrl.navy.mil>
I don't believe it ... the _one_ weekend out of the entire year that
I'm out of town, I get a million replies to a message I send on krbdev!
From "Theodore Y. Ts'o" <tytso@MIT.EDU>:
>I take it you actually tried setting krb5_int32 to be an int on the
>Cray, and it failed the self-test? That's interesting. I wouldn't have
>expected that. I had thought the DES code didn't make 32-bit
>assumptions, but if it does, that would be quite unfortunate.
It dumped core way before it got to the self-tests, actually ...
My logic was that it would be simpler to solve one set of problems
(short being 8 bytes) rather than two. It's entirely possible that
my logic was flawed, and once I solved the 8-byte assumptions that
I didn't need krb5_int32 to be a short.
From: Tom Yu <tlyu@MIT.EDU>
>So where exactly is it failing? A quick perusal of the Ferguson DES
>implementation that we're using reveals a few points where bits beyond
>the least significant 32 might get clobbered, but this shouldn't
>matter because nothing tries to use them (everything that right-shifts
>an unsigned KRB5_INT32 masks the result sanely as far as I can see).
>Is it the actual DES verification test that is failing? Or is it
>something like the MD{4,5} or CRC32 or SHA tests that are failing?
I had trouble getting the DES self-test to work ... but I solved those
problems. SHA doesn't work, though :-)
>The krb5 glue to the Ferguson DES library is broken w.r.t. the size of
>a "KRB_INT32".
Sigh ... this might be the _real_ problem after all. I'll try it!
--Ken
