[unixODBC-support] Core dump issue from UnixODBC on AIX platform

Hassan Shaik hshaik at arcot.com
Tue Nov 6 18:39:42 GMT 2007


Hi Nick,
 
Sure, I will try to build 2.2.13pre overall.
 
Actually, we are seeing this problem on a production environment from a customer. That's why its little tricky and we can't directly test on production server. We are trying to recreate this issue locally on our servers. If I am able to reproduce it locally, then I can perform these tests with 2.2.13pre binaries.
 
- The threads in our application ranges from 5 (min) to 15 (max) based on load.
- We are able to see core dump within 18-24hour duration.
 
We already used Purify for code validation on AIX ( to do mem leak testing ). No issues were reported from Purify tool.
 
Appreciate your help on this.
 
Thanks
Hassan

________________________________

From: unixodbc-support-bounces at easysoft.com on behalf of Nick Gorham
Sent: Mon 11/5/2007 8:34 AM
To: Support for the unixODBC project
Subject: Re: [unixODBC-support] Core dump issue from UnixODBC on AIX platform



Hassan Shaik wrote:

>Hi Nick,
>
>I need few more clarifications:
>
>1. I extracted UnixODBC2.2.13pre source and the source code in
>__handles.c looks similar. Will there be any impact on this problem from
>code changes in other files?
>
>2. Adding "Threading = 0" to odbcinst.ini file is just a suggestion, and
>may not solve this issue, right?
>
>3. If the above two can't solve this problem, we need to analyze the
>current code. I checked the Makefile under the directory where I built
>UnixODBC2.2.12. The flag HAVE_LIBTHREAD is set there. So, following code
>is active in __handles.c,
> 
>
Build the entire thing, don't just pull individual files. Part of the
reason for building 2.2.13 is that it does hav thread related fixes, and
the other part is if I look at the problem I will be starting with that
code.

>
>But mutex_entry( &mutex_lists ) and mutex_exit( &mutex_lists ) calls are
>used at start and end of these functions. I am wondering how come str
>can become NULL suddenly, as we already validating for it in while loop
>(using while( ptr )).
>
>Sorry, if I am troubling you with many questions.
>
>Please clarify on this.
> 
>
Well, I guess if we knew this we would know what the problem is.

How many threads does your app use, how long does it run before failing,
does it always fail at the same place?

I ask these question to ge a feeling for how much effort would be
required to fix the problem.

Have you tried any code validation tools? Maybe IBM's Purify?

--
Nick
_______________________________________________
unixODBC-support mailing list
unixODBC-support at easysoft.com
http://mail.easysoft.com/mailman/listinfo/unixodbc-support


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.unixodbc.org/pipermail/unixodbc-support/attachments/20071106/44f13907/attachment.html>


More information about the unixODBC-support mailing list