Re: 7.0.2 regressions testing on Sparc running 2.5.1

On Fri, 1 Sep 2000, Tom Lane wrote:

> My bet is that you are chasing the wrong idea.
...
> I'd be looking at issues like whether you have
> things configured for dynamic shared-object linking at all. For
> example, on HPUX it's necessary to use a special linker switch (-E,
> I think) while linking the main "postgres" executable, or else its
> global symbols won't be set up to be accessible by dynamically loaded
> objects. Dunno if there's a similar incantation needed on your
> platform, but that seems like the sort of issue to be looking at.

I think that's it. I did a make clean on the main makefile, and
ran make again. This time directing all output to a file.

There are a few places where -fPIC occurs (which is part of the
shared object business). However, doing a search about -fPIC
turned up:
Forum: comp.unix.unixware.misc
Subject: Re: building shared object with gcc/g++
Author: Casper H.S.Dik - Network Security Engineer
<Casper(dot)Dik(at)Holland(dot)Sun(dot)Com>

...

Do not use ld directly.
Do not use "gcc -G"
Use gcc -shared

...

You must always link with gcc, including when linking shared
libraries.

-----------------------------------------------------------

So, I guess all of my compiles aren't quite right. Hopefully
this will fix things. I still possibly have an issue with
Run-time link paths. I have been adding -R /path/to/lib
but I guess the proper gcc incantation is
-Xlinker -R -Xlinker /path/to/lib

Have a great weekend.
Gord

Matter Realisations http://www.materialisations.com/
Gordon Haverland, B.Sc. M.Eng. President
101 9504 182 St. NW Edmonton, AB, CA T5T 3A7
780/481-8019 ghaverla @ freenet.edmonton.ab.ca