FreeTDS Development Group

[Frediano Ziglio <freddyz77 AT tin.it>]

Il gio, 2002-11-07 alle 23:12, Lowden, James K ha scritto:
> > From: Frediano Ziglio [mailto:freddyz77 AT tin.it]
> > Sent: November 7, 2002 4:35 PM
> > > 
> > > But 5.0 sends ee, ae, d1, 1, 0?
> > > Received packet @ 2002-11-07 15:43:34.383225
> > > 0000  ee 0d 00 01 00 00 30 0a  00 00 00 6c 21 4d 12 00
> > 
> > Row, numeric of 33 (yes, 33!) bytes. So we must change 
> > numeric limit...
> > (in tds.h is defined as 17 bytes...)
> 
> I'm trying to interpret that packet:
> 
> ee            TDS_ROWFMT
> 0d 00                 13 byte packet
> 01 00         1 column
> 00            Name length = 0
> 30            status nullable, updatable (!)
> 0a 00 00 00   user type numeric
> 6c            nullable numeric
> 21            33 bytes wide
> 4d            precision 64+13 = 77
> 12            scale 18
> 00            no locale
> 
> Is that right?  
> 
> Wow.  That's one big number, 10 ^ (77 - 18), like a gazillion or something. 
> 

Yes, there are solution and a workaround:
Solution: handle huge (256 bit) numeric types. (note that Sybase
proprietary ctlib use this type) (check all code that use numeric, check
all buffers for hold numeric).
Workaround: sybase/ms db allow numeric only of 128 bit precision, so
reading it discard high info (modify only token.c).

The solution is best (we are not on freeze). I'll try this way...

freddy77