Re: bazaar: krb5_fcc_ops has get_cache_first/next/end set

["Henry B. Hotz" <hotz@jpl.nasa.gov>]

I've been able to load both Sun/MIT and Heimdal modules in a Sun LDAP  
server using a similar technique (on Solaris with the Sun compiler/ 
linker).  I suggest you look very carefully at all the options  
available for your linker.

FWIW the heimdal-using module is a .a, not a .so, and that seemed to  
be necessary.  It's still dynamically loaded though.

On Sep 13, 2006, at 9:14 AM, Michael B Allen wrote:

> On Mon, 11 Sep 2006 07:53:45 +0200
> Love Hörnquist Åstrand <lha@kth.se> wrote:
>
>> 8 sep 2006 kl. 17.21 skrev Michael B Allen:
>>
>>> Is there any way for the krb5_fcc_ops get_cache_first,  
>>> get_cache_next,
>>> and end_cache_get functions to somehow get set?
>>
>> They don't exists in a heimdal release, so I assume you are running a
>> snapshot. If your code somhow load the wrong (older) shared library,
>> that would look like this problem.
>
> We found the problem. Another PHP module (pgsql.so) was dynamically
> linked with stock Heimdal. My PHP module (plexsso.so) is dynamically
> linked with a libarary containing my custom Heimdal (mechglue) plus
> other stuff. As a result, apparently there was some symbol collision.
>
> For some reason I thought statically linking everything into one  
> shared
> library for the PHP module to dyncamically link with would give me
> some isolation. Apparently that's not the case. Can anyone recommend
> a workaround? Are symbols versioned on Linux or once loaded are they
> shared regardless of their source?
>
> Mike
>
> -- 
> Michael B Allen
> PHP Active Directory SSO
> http://www.ioplex.com/

------------------------------------------------------------------------ 
----
The opinions expressed in this message are mine,
not those of Caltech, JPL, NASA, or the US Government.
Henry.B.Hotz@jpl.nasa.gov, or hbhotz@oxy.edu