I've been trying to get Slony installed on some Postgres 8.2.5 
instances I've got running on a couple of AIX boxes, but with no luck 
at all, and I'm hoping someone's run into the same problem and has a 
solution.

Postgres itself builds, tests, and runs just fine. More than fine, 
it's a screamer and leaves the Postgres instances I have on some 
Fujitsu Solaris boxes in the dust. Slony 1.2.11 also builds and 
installs just fine. The trouble comes in as soon as I try to get any 
sort of replication going.

Doing an init cluster with the AIX box as a master, or a store node 
with the AIX box as a slave, crashes Postgres. From the core file 
left behind, the death is in slony's .so, when it tries to call 
palloc. According to gdb, palloc's function address is NULL, and so 
the task gets an illegal instruction trap and dies.

This is very consistent; I've built Postgres and Slony as both 32 and 
64 bit executables, and with both xlc6 (IBM's compiler) and gcc. The 
results are the same, palloc's address is always NULL and things die.

 From what I can tell, it looks like when slony's .so is loaded, the 
palloc call *back* into Postgres' shared objects never resolves, so 
it's left NULL and things die. Unfortunately AIX's way of handling 
shared objects is... somewhat unique and I've got no experience with 
it. I have a feeling there's a simple fix somewhere, but beats me 
what it is. Google's been no help, but that's not too much of a 
surprise since I don't quite know what I'm looking for. (The obvious 
simple searches haven't turned up much of use)

Has anyone come across this? Better yet, anyone know a fix?
-- 
				Dan

--------------------------------------it's like this-------------------
Dan Sugalski                          even samurai
dan at sidhe.org                         have teddy bears and even
                                       teddy bears get drunk