Re: [Sqlrelay-discussion] sqlr-status and ps aux doesn't show the same number of connections

[Martin F. <mar...@po...>]

BlacKarma schreef:
> Hi everybody,
>
> I'm experiencing some problem using SqlRelay (0.39.4) over Oracle 10g.
> SqlRelay start and after a while the listener fork but the open client 
> connection is less than the open server connection and when a check 
> with a "ps aux" the number of process is less than the open server 
> connection displayed by sqlr-status.
Probably the same bug we found and reported, but David hasn't responded 
on it yet.

Here is my previous email:

Hello David,

Just an update for this reported bug: we have commented the line 1658 
(with "OCIHandleFree(def[i],OCI_HTYPE_DEFINE);") and haven't had any 
problems since, we are using the code in production enviroment for 2 
months. So I'd say: you can safely update the sourcecode of sqlrelay, to 
prevent the connection-crash-bug that was bugging us.

Regards,

Martin Fleurke

Ingmar Brouns schreef:
Hi David,

I have been trying to debug what's going on. The connection process
crashes on this statement

                OCIHandleFree(def[i],OCI_HTYPE_DEFINE);

line 1744 of oracle8connection.C with a segmentation fault. I've not
been able to figer out why this happens. I am new to C++ and GDB and I
cannot view the contents of the OCIDefine object. However, when I
looked into the Oracle Call Interface Programmer's Guide

http://www.ewp.rpi.edu/hartford/webgen//appdocs/oracle/doc/appdev.920/a96584/oci02bas.htm 


it said the following:

OCI_HTYPE_DEFINE = OCI define handle

and

Bind and define handles are freed when the statement handle is freed
or when a new statement is prepared on the statement handle.

So my question is, why is it being freed explicitly anyway? I've
compiled sqlrelay with the statement commented out, and now everything
seems to be going fine. For what situations is the
OCIHandleFree(def[i],OCI_HTYPE_DEFINE) statement neccesary?

Anxiously awaiting your reply,

Ingmar