Re: Commit to odbc.c from 5/19/2010 breaks prepared statements with pyodbc

Another regression, fix unknown so far.  Between the same two revisions, the following has broken.

Before:

nortonp@sys4 17:04 ~/work/odbc_tests $ ./test_mssql_select.py
Testing sybase
[('SYBQA1', )]
Testing mssql
[('DBDEVNYC2', )]

After upgrading to 0.83.dev.20100902 or later:

Testing sybase
[('SYBQA1', )]
Testing mssql
[(u'DDVY', )]

I haven't narrowed this down to a particular changeset yet.

-Peter


On Fri, Oct 15, 2010 at 01:58:45PM -0400, Peter C. Norton wrote:
> Hi, I've got another bug report.  At my site we've run into a
> regression when going from the following:
> 
> pyodbc-2.1.7
>  depends on
> unixODBC-2.2.14
>  which depends on
> freetds-0.82.1.dev.20091223
> 
> to the same pyodbc and unixODBC, but upgrading to
> freetds-0.83.dev.20100902 + various patches from upstream.  When users
> tested it I got the following succinct bug report:
> 
> import pyodbc
> def the_test(db_name):
>     db = pyodbc.connect("DSN=%s;UID=<USERNAME>;PWD=<PASSWORD>;DATABASE=<any database>;APP=python_test" % db_name)
>     q = db.execute("select * from <any table> where type = ?", (1,))
>     print q.fetchall()
> 
> the_test("SYBASE")
> 
> results in the following error:
> 
> Error: ('HY000', 'The driver did not supply an error!') 
> 
> Confirmed on solaris x86 and linux/64-bit.  I've bisected the changes
> made back to 5/19/2010 via the freetds git repository and I found that
> the following broke prepared statements from pyodbc:
> 
> http://github.com/brianb/FreeTDS/commit/2ed78ea4dd9018927b56822e318e0b5ac9401b98#diff-2
> 
> Specifically the chunk between line 1301 and 1313. The relevent chunk of code is this:
> 
>  		if (desc->type == DESC_IRD && ((TDS_STMT*)desc->parent)->need_reprepare && \
>  		    odbc_update_ird((TDS_STMT*)desc->parent, errs) != SQL_SUCCESS) \
>  			exit; \
> 
> Reverting the patch to the pre-2010-05-19 state has the preprocessor
> replace the above with:
> 
> #define IRD_UPDATE(desc, errs, exit) do { } while(0)
> 
> which may just be masking an error, but I do not know.  Backing this
> out causes prepared statements to work again.  Are there any gotchas
> to reverting this change?  And is there a better/more correct fix for
> this?
> 
> By the way, this works fine when I point the same test an an MSSQL 2005 or 2008 server.
> 
> Thanks,
> 
> -Peter
> 
> 
> 
> _______________________________________________
> FreeTDS mailing list
> [email protected]
> http://lists.ibiblio.org/mailman/listinfo/freetds