Andy Hassall wrote:
> --- Tim Bunce <[EMAIL PROTECTED]> wrote: > On Mon, Jan 12, 2004 at
> 05:48:48PM -0500, Jeff Urlwin wrote:
>>>
>> Um, Andy, you didn't mention that one in your "Solaris 2.7 / 9.2.0.4
>> / 32-bit"
>> results. Did you see that?
>
> Hm, I'm trying to compile it again and I'm now hitting the 32/64 bit
> problem again. You'd better ignore my results until I work out what
> I've done - sorry.
>
> It _does_ link if I use '-l' when running Makefile.PL though - but
> I'm sure I didn't do that the first time. What's puzzling me is that
> the 1.15 I built yesterday and installed into a test directory is
> correctly linked against $ORACLE_HOME/lib32.
For what it's worth, I found out what was going on - was a
non-deterministic problem resulting from clocks being out of sync between
the server I was compiling on, and the fileserver the files were on.
From a freshly unpacked directory, got the following from Makefile.PL:
"
Attempting to discover Oracle OCI build rules
cc -c -o DBD_ORA_OBJ.o DBD_ORA_OBJ.c
by executing: (make -f /opt/oracle/9.2.0.C/rdbms/demo/demo_rdbms.mk build
ECHODO=echo ECHO=echo GENCLNTSH='echo genclntsh' CC=echo OPTIMIZE=
CCFLAGS= EXE
_ORA_OBJ.o)
Oracle oci build prolog:
+ make: *** Warning: File `DBD_ORA_OBJ.o' has modification time in
the future (2004-01-14 13:15:13.817217 > 2004-01-14 13:15:13.155464156)
Oracle oci build command:
+ -L/opt/SUNWcluster/lib -R/opt/SUNWcluster/lib -o build
-L/opt/oracle/9.2.0.C/rdbms/lib/ -L/opt/oracle/9.2.0.C/lib/ -o DBD_ORA_EXE
DBD_ORA_OBJ.o
host9 -lnus9 -lnldap9 -lldapclnt9 -lnsslb9 -lnoname9 -lntcp9 -lntcps9
-lnsslb9 -lntcp9 -lntns9 -lnsl -lsocket -lgen -ldl
-R/opt/oracle/9.2.0.C/lib -laio
-lthread
+ make: warning: Clock skew detected. Your build may be
incomplete.
Unable to interpret Oracle build commands from
/opt/oracle/9.2.0.C/rdbms/demo/demo_rdbms.mk. Using fallback approach.
"
The clock skew gets picked up as an error, and it then goes to another
method, which correctly picks up the lib32 directory instead of lib, and it
builds and links correctly.
But because the clock problem is intermittent, the next time I tried it the
first method 'succeeded', but picked 'lib' and not 'lib32' and so the link
failed.
--
Andy Hassall <[EMAIL PROTECTED]> / Space: disk usage analysis tool
<http://www.andyh.co.uk> / <http://www.andyhsoftware.co.uk/space>
