FreeTDS Development Group

["James K. Lowden" <jklowden AT freetds.org>]

ZIGLIO, Frediano, VF-IT wrote:
> 
> BUF_FULL is returned if buffering is turned on and reading the next row
> would cause the buffer to be exceeded. In this case, no row will have
> been read. To read any more rows, at least one row must first be pruned
> from the top of the row buffer by calling dbclrbuf.
> 
> so BUF_FULL is not returned if buffer is already full but only if buffer
> will be full

How does FreeTDS differ from Microsoft's behavior?  

As I read the specification, "would ... be exceeded" means "is full".  If
the buffer holds 3 rows, dbnextrow() returns BUF_FULL when trying to read
the 4th row.  

To me, dbnextrow() is correct: it checks to see if there is room in the
buffer.  If not, it returns BUF_FULL without reading a row.  

(I'm hoping to fix dbnextrow() and dbresults() today to remove
dbproc->dbresults_state.)

Regards, 

--jkl