Error while compiling krb 1.5

Marcus Watts mdw at
Sat Jul 8 02:45:21 EDT 2006

Ken Raeburn <raeburn at MIT.EDU> writes:
> Cc: Jeffrey Hutzelman <jhutz at>, krbdev at MIT.EDU
> From: Ken Raeburn <raeburn at MIT.EDU>
> Subject: Re: Error while compiling krb 1.5 
> Date: Sat, 8 Jul 2006 01:24:14 -0400
> To: Marcus Watts <mdw at>
> On Jul 7, 2006, at 17:58, Marcus Watts wrote:
> > Jeffrey Hutzelman <jhutz at> writes:
> >> Two words: com_err
> >
> > I can reduce that to one word.
> >
> > 	"Pain".
> >
> > But I should probably expand that out for anyone who hasn't
> > run into this yet:
> > [...long list elided...]
> Yeah, it's gotten to be a sorry state.  There was someone who was  
> going to work on integrating the various branches of the MIT  
> implementation, and perhaps add in the Heimdal version of the API as  
> well, but AFAIK nothing's actually happened there yet.  If anyone  
> else is interested in picking up that little project, I'd like to  
> know....
> Ken

I tried dodging that bullet for a long time, with less and less
success.  Every successive project I'd end up with or new environment
I'd try would wind up with ad hoc bits of glue code to try to fool bits
of mit kerberos, or afs into doing the right thing, with successively
less useful results.

So, my current effort in this direction is here:
this includes:
	a compile_et (based on perl) that can produce 3 different
		kinds of initialization, also handles
		heimdal extensions, non-standard bases, etc.
	mutex code to handle list locking
	key alloc code which leaks keys (instead of hanging).
	non-standard base handling (a la heimdal, to suit afs)
	'' - using non-standard bases.
	negative error numbers
	random variations on remove_error_table,
	initialize_error_table_r, error_table_name, etc.
I don't have anything here that does the mit k5 ".init" hack yet.
I suppose it's easy enough to put in.

I had been thinking of this as more of a 'reference' implementation
than its own separate thing - so there's more work needed to make
this truely something other people would want to use.

At some point, I want to hack up some version of openafs to support
some equivalent to "with-system-et".  The current
error_message annoys me.  I also want to find all cases in openafs where
it assumes "all negative numbers are rpc errors" and constrain that to
be "-1000..-1".  This might be a small project of mine at the next
AFS hackathon, if it doesn't happen sooner.

At one point, I told a number of the various com_err developers
about the compile_et I was working on - an earlier version of
what's included here.

I wish I had more time to proselytize people on the virtues of
"--with-system-et" non-standard base handling, and general
pain non-promulgation.


More information about the krbdev mailing list