On the FreeTDS mailing list, James K. Lowden (one of the FreeTDS maintainers)
took a look at the pymssql result-processing code and suggested some changes to
make it use the DB-API of FreeTDS in a more canonical way. I don't know of any
bugs in our current code, but James K. Lowden knows FreeTDS well, so it's
probably worthy of reviewing his comments --we may have some yet unknown bugs:
http://permalink.gmane.org/gmane.comp.db.tds.freetds/15122
Original issue reported on code.google.com by msabr...@gmail.com on 17 Sep 2013 at 2:57
Original issue reported on code.google.com by
msabr...@gmail.com
on 17 Sep 2013 at 2:57