🚀 go-pugleaf

>Has anyone out there gotten isode 7.0 to compile cleanly on a system 5
>release 4 system?

It's a mess. We have the base ISODE running now; most of the problems were in
terrible varargs ab-usage, and easy enough to kludge around. We also went with
the ksocket library interface for RFC1006, instead of TLI. (Besides the fact
that the TLI TCP/IP interface is a joke, we wanted to keep TLI clear for use
with Pyramid's CLNP and CLNS interfaces, and ISODE really only supports one
TLI interface at a time.) That of that took about a week of tinkering.

PP is another problem, and QUIPU is worse. The base ISODE code at least runs
on SVR3, which is enough to get you going nicely on SVR4. But PP and QUIPU
were never brought up on SVR3, and so rely heavily on the BSD compatability
library, which in every SVR4 I have seen is so shot full of holes as to be
almost unusable. More importantly, PP and QUIPU haven't been tested on more
than a handful of architectures, and seem to keep running into nagging little
problems with alignment and arguments that don't work the way they were
expected. 

And to think this is some of the most portable code I've ever worked on! UNIX
is just getting to big to grasp any more. :-(

My first step was to bring *everything* under a source-code debugger. I use
dbx, but sbd or whatever will do. I found bugs in secords that would have
taken hours to isolate otherwise. You will need about 300MB of disk to hold
all the binaries. 

My attack strategy for PP and QUIPU has been to much more tightly couple them
to the "compat" library and general.h header files of ISODE, and pull in the
various compatability hooks in source files when needed. The target is SVR3
first, then SVR4. This seems to be going well so far.  But this is obviously
not a weekend project.

<csg>