Overflow in GetDataDecimal()

[jbq]

Using pyodbc version 3.0.7 and 3.0.8-beta06 (latest from github) I get a bus error when trying to parse a decimal value returned from a bogus expression on iSeries (AS400).

For example this simple test segfaults every time:

print cur.execute("SELECT decimal('', 10, 2) FROM SYSIBM.SYSDUMMY1").fetchone()

It seems like SQLGetData() returns a bogus overflowed value in the cbFetched variable:

Program received signal SIGBUS, Bus error.
0x00007ffff560aca7 in GetDataDecimal (iCol=<optimized out>, cur=<optimized out>) at /home/jbq/pyodbc/src/getdata.cpp:486
486             if (buffer[i] == chDecimal)
(gdb) list
481
482         int cch = (int)(cbFetched / sizeof(SQLWCHAR));
483
484         for (int i = (cch - 1); i >= 0; i--)
485         {
486             if (buffer[i] == chDecimal)
487             {
488                 // Must force it to use '.' since the Decimal class doesn't pay attention to the locale.
489                 buffer[i] = '.';
490             }
(gdb) print cbFetched
$14 = 4294967295

Any help would be appreciated, as I haven't been able to figure out what's wrong. I'd be glad to provide more information if needed. I need this code to be working and willing to sponsor development if this could speedup resolution.

Thanks!

[mkleehammer]

That looks like a 32-bit -1 coerced to an unsigned integer. IIRC, -1 is SQL_NULL_DATA.

A couple of questions:

• Is this a 32 or 64 bit system?
• What is the value SQL_NULL_DATA?

There is an explicit check for SQL_NULL_DATA, but if the data types are somehow wrong it could fail.

[jbq]

This is a 64bit system, and SQL_NULL_DATA is -1

[mkleehammer]

Thanks for your patch!