Cfengine for HP-UX on IA64

[Tod Oace]

Has anyone else had the displeasure of compiling cfengine on an Itanium 
HP-UX system? I finally got it going but was getting an "I don't 
understand what architecture this is!" warning when I ran it. This was 
the same error I got when I tried running my PA-RISC binary (which IA64 
HP-UX emulates). I modified the following and the error is gone but I'm 
not sure if this is the best solution:

# diff classes.c.orig classes.c
90c90
<    {"hp-ux","9000*","*"},        /* hpux */
---
>    {"hp-ux","*","*"},            /* hpux */

"uname -a" looks like:
HP-UX nwhpfs02 B.11.23 U ia64 0035364093 unlimited-user license

I may just go back and try recompiling the PA-RISC binary with the 
classes.c tweak. Compiling on HP-UX Itanium was definitely not fun. I 
had to install m4, bison and flex. Then it took me a while to figure 
out that openssl compiled 64 bit object files while other things did 32 
bit. Setting CFLAGS and LDFLAGS to "+DD64" took care of that (might as 
well use the 64 bits, right?). I think I also had to toss in an 
"-L/usr/local/lib" on the LDFLAGS for cfengine.

I set LDOPTS to "-a archive_shared" so cfengine would pick up the 
static Bison and Flex libraries rather than skipping over them and 
trying to use the incompatible system yacc and lex libraries. You can 
stick the "-a ..." on the ld's command line but not on cc's. There's a 
man page for ld but not for cc, so I don't if there's an equivalent 
CFLAGS setting I can use. Maybe "cc +help" would've told me, but the 
HTML page it references doesn't exist. (sigh)

I couldn't figure out how to get a fully static binary, which is what 
I've tried to standardize on so I don't have to worry about all the 
right shared libraries being installed correctly on all systems in 
order for cfengine to run. ld has a "-noshared" option, but cc won't 
have it. Instead, cc wants "-minshared", but turning that on caused 
something else to fail. So I'll settle for just having openssl and 
BerkeleyDB statically linked in like I settled for on Solaris.

There were a bunch of cfengine compile warnings that I didn't look 
into. The Itanium HP-UX compiler doesn't like "-hreads", and there were 
lots of warnings about declarations in /usr/include/sys/vfs.h. No idea 
whether or not that's something cfengine should be concerned about, but 
I can spend some time digging into it if it should be fixed.

Anyway... I'd appreciate any feedback or advice on this. Thanks...

--
Tod Oace, Intel Corporation <tod@intel.com>