FreeTDS Development Group

["Thompson, Bill D (London)" <ThompBil AT exchange.uk.ml.com>]

OK,

I've tracked down these errors. As I suspected they are both symptoms of the
same bug.

I remember now that my fix for this last time was to change dbconvert(), but
that change doesn't seem to have made it into CVS ( for good reason I'm sure
- it was to do with my use of strlen, I think).

As I have no wish to re-visit the convert code :-) , I've fixed the issue at
source in bcp.c
here it is, a one line change.

in function _bcp_exec_out() :

           for (i = 0; i < dbproc->bcp_colcount; i++) {

               bcpcol = dbproc->bcp_columns[i];
               curcol = resinfo->columns[bcpcol->tab_colnum - 1];

               if (is_blob_type(curcol->column_type)) {
                   /* FIX ME -- no converts supported */
                   src = curcol->column_textvalue;
                   len = curcol->column_cur_size;
               }
               else
               {

                   src = &resinfo->current_row[curcol->column_offset];

                   srctype = tds_get_conversion_type(curcol->column_type,
curcol->column_size);

                   if ( srctype != bcpcol->db_type )
                   {
                      bcpcol->db_type = srctype;
                   }

                   if (is_numeric_type(curcol->column_type))
                      memcpy(bcpcol->data, src, sizeof(TDS_NUMERIC));
                   else
                      memcpy(bcpcol->data, src, curcol->column_size);

  --                   bcpcol->data_size = curcol->column_size;
++                   bcpcol->data_size = curcol->column_cur_size;
               }
           } 

A quick test shows that this resolves both issues. 
Perhaps someone could put this patch in ? 
Ta.

Bill

> -----Original Message-----
> From: Thompson, Bill D (London) [SMTP:ThompBil AT exchange.uk.ml.com]
> Sent: Wednesday, October 23, 2002 5:12 PM
> To:   TDS Development Group
> Subject:      [freetds] Re: need some help with freebcp
> 
> All, 
> 
> The "padded out with NULLs" bug was raised a while ago. 
> I know that I included a fix for it in my last-batch-of-changes-bar-one,
> so
> it may have gone away by now. 
> It was to do with the length returned by dbconvert(). 
> 
> If it's still there, I can take a look.
> 
> Garbage appended to shorter values. 
> Sounds to me like it could be a different symptom of the same problem.
> 
> I've just looked , and I can reproduce BOTH problems. 
> Unless anyone else wants to take it up, leave it with me. I'll sort them
> tomorrow
> 
> 
> Looks like I could spend some time on freebcp - sorting out the text
> fields
> issue, getting error handling better and so on. I could also de-dblib the
> code so that it can be called from ctlib and odbc...
> 
> Bill 
> 
> > -----Original Message-----
> > From:       James K. Lowden [SMTP:jklowden AT speakeasy.org]
> > Sent:       Wednesday, October 23, 2002 1:30 PM
> > To: TDS Development Group
> > Subject:    [freetds] Re: need some help with freebcp
> > 
> > On Tue, 22 Oct 2002 22:05:11 -0400 (EDT), "Bob Kline"
> > <bkline AT rksystems.com> wrote:
> > > I can see
> > > some evidence of possible problems that might be related to the
> > > difficulties John is experiencing.  For example:
> > > 
> > >  1. In addition to the tab separator, values are padded out with
> nulls.
> > 
> > That's a bug.
> > 
> > >  2. The tail ends of longer values from earlier rows are carried
> > >     forward as garbage appended to shorter values in subsequent rows.
> > 
> > That's a bad bug.
> > 
> > > As I say, I'm not intimately familiar with the way Sybase's version is
> > > supposed to work, but this doesn't look right to me.
> > 
> > Ahem.  Methinks thou doest protest too much.  
> > 
> > --jkl
> > 
> > ---
> > You are currently subscribed to freetds as:
> [thompbil AT exchange.uk.ml.com]
> > To unsubscribe, forward this message to
> > $subst('Email.Unsub')
> 
> 
> ---
> You are currently subscribed to freetds as: [thompbil AT exchange.uk.ml.com]
> To unsubscribe, forward this message to
> $subst('Email.Unsub')