Operating System - Linux
1753479 Members
4921 Online
108794 Solutions
New Discussion

make test fails DBD::Oracle 1.20

 
Ben Wern
Occasional Advisor

make test fails DBD::Oracle 1.20

I'm trying to build DBD::Oracle 1.20 using Merijn's current Perl and GCC builds for HPUX 11.11 against Oracle Instant Client 10.2. Ran into two problems (solved) and an stuck with one I have not been able to crack. The two solved issues (since they represent changes I made to the system):

1. modified gcc's type.h to include:

# ifndef _CMPTID_T
# define _CMPTID_T
typedef int32_t cid_t; /* For compartment IDs */
typedef cid_t cmpt_t;
# endif /* _CMPTID_T */
(http://forums11.itrc.hp.com/service/forums/questionanswer.do;HP-FORUMS-S-WPA-IDX=HpgT1TKY5s3Fhbb8hpzTPmZ541VKx1mL2HyKf1FWq1V4Tz2g1mK6!-1200587909!1449797411?threadId=964154&admit=109447626+1205772307767+28353475)
2. Modified the Makefile produced by Makefile.PL to NOT include -Wl in the ld line.

Make completes with several warning regarding int and unsigned int in various places (I can provide this output if relevant, but I don't think it is.) Make test kicks back:

t/10general.............DBI connect('','USER/PASS',...) failed: ERROR OCIEnvsCreate. Check ORACLE_HOME env var, NLS settings, permissions, etc. at t/10genal.t line 20

I have confirmed that TNS_ADMIN, ORACLE_BASE, ORACLE_HOME, SHLIB_PATH, PATH, and LD_LIBRARY_PATH all contain the instant client directory. TWO_TASK is set to the correct SID, and that instance is defined in the tnsnames.ora file. I have also tried specifying ORACLE_SID with no change in behavior. ORACLE_USERID is set with correct USER/PASS information. I have also confirmed that I can log in via sqlplus using the user/password/SID.

Most of the posts I've been able to find (here and elsewhere) center of environment variables missing when running through apache or similar environments (mostly loosing LD_LIBRARY_PATH and SHLIB_PATH), but that doesn't apply here.

If I change paths to the full oracle client (which I'm trying to replace) and recompile, DBD::Oracle tests fine, and appears to function normally, so I think this is an Oracle Instant Client specific issue.


Has anyone seen this or have suggestions on a next step?