<div dir="ltr">This is mostly a big "thank you" for Adam Vandenberg. (Anyone who knows him, please buy him a beer!)<br>Adam submitted a bug report on sourceforge indicating that one should call CoInitialize() BEFORE Dispatch(), not AFTER. I made the change on my development machine, and could see no changes. Everything worked before and still worked after.<br>
Meanwhile, I had a long running python script on another machine which kept erroring out and crashing several hours into its operation. More investigation and several re-runs later, I found that it was an InterfaceError exception at adodbapi.connect() time, right where Adam said the error was.<br>
I put the patched adodbapi on the production machine and *magic* -- the error went away!<br>Score one for open source software. I would NEVER have found that bug.<br>--<br>The patched software is on CVS.<br>Also included is a bug fix from Bob Kline so that nextset() will work even when one of the result sets is empty, and some cleanup of the unit tests also suggested by Bob Kline. Thank you, Bob. I also removed all attempts to test against an AS400 database. The SQL was too weird. I would have had to modify every CREATE TABLE, SELECT, UPDATE, and INSERT statement in the test suite.<br>
--<br>Of course, the fixed version of adodbapi will be included with the next release of pywin32. I note that the download rate on <a href="http://adodbapi.sourceforge.net">adodbapi.sourceforge.net</a> has gone to zero for the last two months (with the release of pywin32 v 211). Thank you Mark.<br>
--<br>Vernon Cole<br><br></div>