FreeTDS Development Group

["Mark J. Lilback" <mark AT lilback.com>]

At 4:14 PM -0400 7/11/01, Brian Bruns wrote:
> Just to be plain about this, the server does the conversion.  Most
> protocols in this situation rely on the client to make the switch, or just
> agree on network order (which means two conversion for little endian to
> little endian comm).

For user data, or for all network traffic?

I've been lazy about submitting patches lately, but I had to add the 
ifdef below to token.c fix a value that wasn't being properly 
converted when talking to SQL Server 7:

static int tds_process_msg(TDSSOCKET *tds,int marker)
{
int rc;
int len;
int len_msg;
int len_svr;
int len_sqlstate;

        /* packet length */
        len = tds_get_smallint(tds);

        /* message number */
        rc = tds_get_int(tds);
#ifdef WORDS_BIGENDIAN
        rc = ByteSwap32(rc);
#endif
        tds->msg_info->msg_number = rc;


I also added the following to my copy of tdsutil.h to help with this 
and a couple of other byte swaps:

#define ByteSwap16(value)                 \
        (((((unsigned short)value)<<8) & 0xFF00)   | \
         ((((unsigned short)value)>>8) & 0x00FF))

#define ByteSwap32(value)                     \
        (((((unsigned long)value)<<24) & 0xFF000000)  | \
         ((((unsigned long)value)<< 8) & 0x00FF0000)  | \
         ((((unsigned long)value)>> 8) & 0x0000FF00)  | \
         ((((unsigned long)value)>>24) & 0x000000FF))

--

__________________________________________________________________________
                                         "The best assumption to have
Mark J. Lilback                           is that any commonly held
<mark AT lilback.com>                        belief is wrong." -- Ken Olsen,
http://www.lilback.com/                   founder, Digital Equip. Corp.