Re: Buffer overflow in extract_diag_error_w()

On 11/11/15 06:32, Andrew Punch wrote:
> Hi,
>
> I am working with Netezza in Perl. The script was crashing with a 
> segfault. When I used gdb to investigate the problem was isolated to 
> the wide_strcpy() call in extract_diag_error_w() due to a larger than 
> expected error message from the Netezza driver. The error message from 
> the Netezza driver was not corrupt and was in the correct format.
>
> I solved this problem in the short term by increasing 
> SQL_MAX_MESSAGE_LENGTH from 512 to 1024. Ideally something like 
> wcscpy_s() should be used.
>
> Thanks
> -Andrew
>
> diff -ur unixODBC-2.3.1/include/sql.h 
> unixodbc-2.3.1-patched/include/sql.h
> --- unixODBC-2.3.1/include/sql.h    2011-07-04 18:54:09.000000000 +1000
> +++ unixodbc-2.3.1-patched/include/sql.h    2015-11-11 
> 16:13:26.709201111 +1100
> @@ -46,7 +46,7 @@
>  #define SQL_NTSL                  (-3L)
>
>  /* maximum message length */
> -#define SQL_MAX_MESSAGE_LENGTH   512
> +#define SQL_MAX_MESSAGE_LENGTH   1024
>
>  /* date/time length constants */
>  #if (ODBCVER >= 0x0300)
> _______________________________________________
> unixODBC-dev mailing list
> [email protected]
> http://mailman.unixodbc.org/mailman/listinfo/unixodbc-dev

Hi, SQL_MAX_MESSAGE_LENGTH is defined by the MS headers so we don't want 
to alter that, but have you tried a current build of unixODBC, as 2.3.1 
was from 26th November 2011. I think the problem is fixed in 2.3.3, the 
changes contain "Fix buffer overrun returning long diagnostic from driver".

Try the current, 2.3.5-pre and if the problem is still there, provide 
details and we will sort it out. How long is the message that comes from 
the driver?

-- 
Nick




_______________________________________________
unixODBC-dev mailing list
[email protected]
http://mailman.unixodbc.org/mailman/listinfo/unixodbc-dev