ID: 37130
User updated by: jabramson at wgen dot net
Reported By: jabramson at wgen dot net
Status: Bogus
Bug Type: OCI8 related
Operating System: Windows XP
PHP Version: 5.1.2
New Comment:
The way i ran into the first place was a unit test where I need to
assert that I *don't* have more than one actual connection in use (per
set of logon credentials) at a time. (My app does very tight oracle
logging and connection/transaction control, so being able to manage and
verify the true identity of connections is important.)
The test input is a set of Arrays, for example, each containing the
user, oracle_sid, and the oci8 resource. If i have more than one
matching pair of user/oracle_sid, then I want to compare the resources
and make sure they also point to the same connection.
Problem is it will always fail for the pconnects if those resources
always get a new handle/id#.
I can accept the behavior if it's by design, but still believe it's at
odds with the documented functionality, and find it counter-intuitive
as well.
Thx,
Jim
Previous Comments:
------------------------------------------------------------------------
[2006-04-19 16:20:10] [EMAIL PROTECTED]
>Nevertheless this is a clear difference in behavior, no?
No
>Is it by design?
Sure it is.
>Does it mean that, aside from querying
>oracle server for session info and comparing those
> results, there is no reliable identity test for
>pconnect() resources?
In common case - yes.
But you can try to do it manually in the same way OCI8 does:
looking at the parameters (login, password, db); persistent or not; and
what character set is used.
Anyway, wth do you need it?
------------------------------------------------------------------------
[2006-04-19 16:01:10] jabramson at wgen dot net
OK, yes, it's understood I am merely comparing an identifier and not
the resource itself. Nevertheless this is a clear difference in
behavior, no? Is it by design?
Note that '===' test fails on those pconnect() resources as well (but
passes on connect() ones). Does it mean that, aside from querying
oracle server for session info and comparing those results, there is no
reliable identity test for pconnect() resources?
------------------------------------------------------------------------
[2006-04-18 21:43:51] [EMAIL PROTECTED]
Comparing resources with "==" doesn't make much sense, as they
identificators may differ, but that doesn't mean underlying connections
are not the same.
------------------------------------------------------------------------
[2006-04-18 21:31:02] jabramson at wgen dot net
Description:
------------
- Apache/1.3.34 (Win32) PHP/5.1.2
- php_oci8.dll is latest build from PECL (dated 4/17/06) : $Revision:
1.293 $
- all related ini settings are default (i.e. no explicit oci8
directives)
according to php manual, non-persistent and persistent db connections
should behave identically.
also according to the manual - this is specified under docs for
oci_connect() - "The second and subsequent calls to oci_connect() with
the same parameters will return the connection handle returned from the
first call..."
i have found that oci_pconnect is creating a new connection handle for
each invocation, whereas oci_connect behaves properly according to the
docs.
Reproduce code:
---------------
<?php
assert_options(ASSERT_WARNING,1);
$cx = oci_connect("user","pass","db");
$cx2 = oci_connect("user","pass","db");
if (assert($cx==$cx2)) {
print 'oci_connect behaves properly';
}
$pcx = oci_pconnect("user","pass","db");
$pcx2 = oci_pconnect("user","pass","db");
if (assert($pcx==$pcx2)) {
print 'oci_pconnect behaves properly';
}
?>
Expected result:
----------------
EXPECTED: no warnings, the assertions pass.
Actual result:
--------------
ACTUAL: the oci_connect assertion passed, but the pconnect version does
not.
------------------------------------------------------------------------
--
Edit this bug report at http://bugs.php.net/?id=37130&edit=1
