Tests failing kdb5_util create apparently broken on the trunk

[Will Fiveash]

On Tue, Oct 03, 2006 at 05:32:57PM -0400, Sam Hartman wrote:
> 
> Hi.  I was having difficulty testing the Redhat patch and then
> discovered my problems happen even without that patch applied.
> 
> kdb5_util create seems to be failing consistently in tests/dejagnu and
> the kadm5 unit tests.
> /home/hartmans/krb5-trunk/build/../src/kadmin/testing/scripts/stop_servers_local: line 37: kill: (9239) - Operation not permitted
> create: Cannot allocate memory while adding entries to the database

I ran across what looks like a similar issue while testing krb5_util
create from a recent trunk build (Revision: 18631).  What I saw is in
krb5_dbekd_decrypt_key_data() there is:
 
     plain.length = key_data->key_data_length[0]-2;
     if ((plain.data = (krb5_octet *) malloc(plain.length)) == NULL)
         return(ENOMEM);
 
and plain.length is ending up with a huge value (4294967294) so malloc()
blows up.  Ken tells me that this is very likely a ASN.1 bug he is
working on.

> Running valgrind shows several problems in the plugin code.

I also intend to do test using libumem on Solaris to catch memory
issues.

> This is on i386-linux with a reasonably modern libc.

-- 
Will Fiveash
Sun Microsystems Inc.
Austin, TX, USA (TZ=CST6CDT)