malloc hang inside krb5_sendto_kdc

[Sam Hartman]

>>>>> "Jeremy" == Jeremy Allison <jeremy.allison at hp.com> writes:

    Jeremy> Ken Weaverling wrote:
    >> In article <4020E692.2040100 at nyc.rr.com>,
    >> Jeffrey Altman  <jaltman2 at nyc.rr.com> wrote:
    >> 
    >> 
    >>> The most useful thing would be to add code before and after
    >>> the krb5_locate_kdc() call to check for context == realm as
    >>> they are both being set to 0x1.  Write to a log file when you
    >>> find the condition.
    >> 
    >> 
    >> Will do that today.

    Jeremy> If this is on an x86 Linux box I would recommend running
    Jeremy> under valgrind to try and track this down. In fact I would
    Jeremy> hope the Kerberos developers are already using this. It's
    Jeremy> a wonderful way of finding and fixing bugs (we use it a
    Jeremy> *lot*).

We tend to use Purify on Solaris, but yes.

I'm somewhat dubious this is a bug in our code.  Or rather, there is
another bug we're seeing periodically here at MIT that has similar
backtraces but seems more like a glibc problem than a bug in our code.
I'm trying to determine if they are the same problem.