Re: Assertion triggered in liblber if LDAP server is down (ITS#837)

[Kurt@OpenLDAP.org]

At 09:54 PM 10/16/00 +0000, gombasg@inf.elte.hu wrote:
>On Mon, Oct 16, 2000 at 01:43:54PM -0700, Kurt D. Zeilenga wrote:
>
>> such as appropriate error checking?  ldap_adandon() should
>> not be called unless ldap_simple_bind() (or other request
>> generating call) was successful.  And if any routine returned
>> LDAP_SERVER_DOWN or other such error, then about the only
>> sensible call to make is ldap_unbind().
>
>I did not say it is done right in nss_ldap, I just wanted to say that
>OpenLDAP should handle it and should not abort the application.

And I'm not saying I wasn't going to fix the problem you reported.
I was providing additional information on appropriate use of API
calls so that you might provide feedback to the developer's of
the application.


>> Stack trace?
>
>It would not help, the call path was completely legal.

Whether the call path is legal or not has nothing to do with
my desire to have a stack trace.  I desire a stack trace to
ensure the bug I was able to reproduce is in fact the same
bug as you are reporting.  I am fairly sure it is from your
original description, but a stack trace eliminates guess
work.

>Just ld->ld_sb->sb_iod
>was not initialized during the unsuccessful bind operation and ldap_unbind()
>tried to send data without checking it.

ldap_unbind?  I assume you mean ldap_abandon?
A stack trace, of course, would answer these questions.