Fixes #5

Fixes #7

Fixes #8

repeatedly executing output bound params can cause the
bound scalars to continually increase in size

Fixes issue #10

Fixes #11

…tions

This is a rather hairy change as charsetid and ncharsetid were global
and used all over the place.
NOTE, you cannot actually build this because there is no VERSION in
Oracle.pm because the repo is missing dist.ini and is probably
using distzilla. To make it build I had to add
$DBD::Oracle::VERSION = '1.63_00';
after require 5.006 in Oracle.pm
It is also missing ABSTRACT

github-flavored markdown

fine-tune CONTRIBUTING.mkd

Fixes #2

…e connections"

This reverts commit 957f547.

Fixes #13

Fixes #17

Because this test file is a giant mess of way too many tests caught
in a clump. They badly need to be cleaned up

Some documentation was "strongly inspired" by DBD::Pg
documentation. So strongly, that we forgot to change 'Pg'
to 'Oracle' in a few places. :-)

Caught by Martin J. Evans.

Mostly Tux's "Try and set -l when the build would have failed"

Accept change request from Yanick. Change DBD::Pg to DBD::Oracle.

Fix git issues 21 and 15 - previous unrelated fix broke this

Merge in fixes for instant client path from Martin.

Ignore disabled constraints in primary/foreign key_info.

This code was trying to print an OCI ub4 value using the format string
"%lu". On systems where unsigned long is wider than ub4 (including several
64-bit Linux systems), this will cause undefined behaviour (which might in
turn look like stack-smashing).

This code was trying to print two arguments that are likely to be size_t (as
the result of strlen() calls) using the "%d" format string. On systems where
size_t is wider than int (including most 64-bit systems), this will cause
undefined behaviour (which might in turn look like stack-smashing).

enhancement from dbi-users to stop hiding connection information messages

… 32 bit.

Previously this was fixed for 64 bit only (git issue #20).

Search for Oracle Instant Client .mk file location when compiling for 32-Bit Instant Client SDK.

Fix printf format-string errors on 64-bit Linux

Set zero-length lob to undef when using ora_piece_lob option

…ent handle that was not invalidated.

and the build is for 64 bit.

Have ora_libdir check $ORACLE_HOME/lib/64 when it exists

Fix for Oracle version default value.

This error caused a SIGSEGV in t/51scroll.t with ORA_DRCP=1

This fixes #27 - among other cases. Note that with ithreads, the global imp_drh environment handle still leaks.

The previous error paths and the attempted partial fix in #28 could cause crashes because of missing curly brackets in if-statements calling OCIHandleFree_log_stat (which is a macro).

The addition of a driver handle destructor also prepares for timely destruction of DRCP session pools. This requires DBI >= 1.623, which adds logic to force destruction of children before parents during global destruction.

Don't set OCI_ATTR_DRIVER_NAME on an unallocated session handle, potentially causing corruption or crashes.

Set the default driver name to "DBD::Oracle : <version>", using the de facto standard driver naming scheme.

This fixes #32

Facilitate setting of OCI_EVENTS in a new OCI environment via the attribute ora_events or the environment variable ORA_EVENTS.

Facilitate setting of OCI_EVENTS in a new OCI environment via the attribute ora_events or the environment variable ORA_EVENTS.

Rewrite of DRCP session pooling, influenced by PHP's oci8.c

Enable sharing (reuse) of server sessions, and implement caching of the client side of the session pool(s) between connects, keeping the connection(s) to the DRCP Connection Broker. Depending on server hardware and network latency, subsequent connects are now up to several hundred times faster than before the rewrite.

Use session tagging to fetch shared server sessions with correct NLS character sets.

The attribute ora_drcp_rlb or the environment variable ORA_DRCP_RLB controls runtime connection load balancing for Oracle RAC.

Documentation update.

All tests pass with ORA_DRCP=1, with the exception of t/60reauth.t (it goes without saying that reauthentication does not work with session pooling).

(updated ocitrace.h - Dean Hamstead)