Re: SQL*Loader Error 523 on Win2K 8.1.7.0.0 - cause & possible solution

FC wrote:
> Hello folks,
> today I had a couple of hours of pure entertainment trying to figure out
> what could cause the SQL*Loader Error 523.
> I have a text file with fixed length fields but the possibility of the last
> column to be truncated at a some point before its expected size.
> The control file defines a variable record-size file, with fixed column
> positions, terminated by CR LF ( X'0D0A').
> It has been running for months without a hiss, until today. Suddenly, the
> program failed with error 523.
> The error is somewhat misleading, because I spent time checking user
> authorizations, disk quotas and so on, until I finally realised that the
> reason must be somewhere else.
> I changed the ROWS from 128 to 1 and the BINDSIZE/READSIZE to the maximum
> expected length of the record (including the expected CR LF) and the problem
> was gone.
> Before that, I tried without success some alternative methods like adding
> TRAILING NULLCOLS and OPTIONALLY TERMINATED BY WHITESPACE.
> During one test I also noticed that when it arrives at the last record of
> the file, if the last column being loaded is shorter than its expected
> length and the last record terminator is absent, SQL*Loader adds an extra
> character of junk.
>
> Now, the question is, are these SQL*Loader bugs or what ?
>
> Thanks,
> Flavio
>
>
Sounds like some Oracle developer was told to patch up a commonly used hacker technique (buffer overflows). Fixed it all right. Now those extra long records don't let a hacker have an easy time of it. Looking at it from that viewpoint, it is kinda, sorta an authorization thing. :) Received on Sat Jan 11 2003 - 00:29:59 CET