Re: sqlloader segmentation fault error

From: George Brown <gbrown_at_partech.com>
Date: 1995/05/22
Message-ID: <gbrown-2205951650480001_at_browng.partech.com>#1/1


In article <3pafr8$bt6_at_reeve.research.aa.wl.com>, groomev_at_aa.wl.com (Valerie Groomes) wrote:

> Hi,
>
> Having a problem with sqlloader. Getting the message
> 'segmentation fault (core dumped)'. Analyzed the 8.7 Mb core dump file with
> dbx and got the following:
>
> program terminated by signal SEGV (no mapping at the fault address)
> (dbx) where
> ulpttb() at 0x103f8
> ulpcnt() at 0xff3c
> ulmcep() at 0x3b74
> main() at 0x35f8
>
> I am running ORACLE 7.0.15, Sun/OS 4.1.3, size of sga is 8Mb.
> Sqlloader successfully loaded up to
> 25Mb of data to a table prior to a server reboot.
>
> Any ideas??? I am working with the ORACLE hotline on this but so far we
> are stumped.
>
> Thanks.
>
> Val

I doubt you'll learn much from the core dump. Are you loading free-format data (ie. comma separated data)? I had a problem where a field was defined to be char(10) and I had a string that had 12 characters. Result: segmentation fault... SQL*Loader is not very tolerant when it comes to stuff like this. It might be the difference between a fixed format where the parser can determine in advance the length of fields and give reasonable error messages and recoveries.

What I did to narrow the problem was to cut out all of the preceding data prior to the fault. If you set ROWS=1, you will get the line number-1 of the bad line.

-- 
George Brown                    Voice: 315-738-0600x250
8383 Seneca Turnpike            FAX: 315-738-8304
New Hartford, NY  13413-4991    e-mail: george_at_partech.com
Received on Mon May 22 1995 - 00:00:00 CEST

Original text of this message