On 20/03/12 12:12, Peter J. Holzer wrote:
A few months ago we upgraded two of our servers from RHEL 4 to RHEL6.
Unfortunately the upgrade involved the OS, the Perl version (5.8.8 ->
5.10.1), the Oracle client version (still 10g, but now instant client)
and DBD::Oracle (1.19 -> 1.30 (now 1.38)), and we didn't note the
problem at once, so it's unclear which of the many changes is the
culprit, but I suspect it's a bug in DBD::Oracle.
Anyway, since the upgrade sometimes (not always) array inserts returned
the error
DBD::Oracle::st execute_array failed: ORA-02005: implicit (-1)
length not valid for this bind or define datatype (DBD ERROR:
OCIBindByName)
Here is a test script which semi-reliably (>= 50% of the time) produces
the error message:
<snipped script>
Some notes:
ARGV[0] is supposed to be the name of a file containing space-separated
connect data, e.g.:
dbi:Oracle:ORCL scott tiger
When the script fails, it always fails on the second iteration of the
loop. If it gets through the second iteration it completes all 1000
iterations successfully.
When the call to prepare_cached is replaced by a simple prepare, the
script fails less often.
I have run the script with DBI_TRACE=4, but I don't see any significant
difference between successful and unsuccessful runs.
hp
I've run this a couple of dozen times and it never has failed for me. I was
using Linux Ubuntu 10.10, Perl 5.10.1 (with threads), DBD::Oracle (from
subversion trunk) and Oracle Instant Client 11.2.
Martin
--
Martin J. Evans
Easysoft Limited
http://www.easysoft.com
