krb4 library major version bump

[Ken Raeburn]

Investigating Sam's problem with aklog and krb524init, I found one
likely possibility.

The new krb_get_admhst will fill the supplied buffer with nulls after
the hostname, to a length of MAXHOSTNAMELEN.  In the aklog sources
Athena is using, the buffer supplied has a size of MAX_HSTNM, defined
in krb.h to be 100.  So krb_get_admhst will clobber a bunch of stuff
on the stack.  This might lead to the broken pointers being passed to
strcpy as indicated in the aklog crash.

When I change the krb_get_admhst code to use strcpy instead of
strncpy, and reduce the size limit to MAX_HSTNM, Athena's aklog works
for me.

So, yes, it's an ABI change, but I don't think it's intentional or
desirable.


hartmans at MIT.EDU (Sam Hartman) writes:
> We seem to have changed the ABI on Unix for krb4 in that I seem to get
> strange failures using applications linked against 1.2.x with -current
> shared libraries.  This is not surprising, but we should make sure that:
>
> 1) We consider this acceptable
>
> 2) We know at least one change that should create ABI divergence 
>
> 3) We are willing to accept a so bump for the krb4 libs.
>
>
>
> _______________________________________________
> krbdev mailing list             krbdev at mit.edu
> http://mailman.mit.edu/mailman/listinfo/krbdev