Hi Kristian, I've at last figured out the join_nested issue. To sum up, it was a false alarm, the fix is to adjust the test suite and I've pushed it to the lp:~maria-captains/maria/5.1-merge tree. On Sun, Dec 27, 2009 at 07:48:33PM +0300, Sergey Petrunya wrote:
At file:///home/psergey/dev/5.1-merge-look/
------------------------------------------------------------ revno: 2777 revision-id: psergey@askmonty.org-20091227164827-038uergwe82hyd1m parent: knielsen@knielsen-hq.org-20091214083138-axta6hzovkyt65wg committer: Sergey Petrunya <psergey@askmonty.org> branch nick: 5.1-merge-look timestamp: Sun 2009-12-27 19:48:27 +0300 message: Make pbxt.join_nested test pass - The reason the test failed was competition between 3+ QEPs with identical costs. Before, two plans were competing, and that was addressed by using --sorted_result on the EXPLAIN output because they were different only in join order. Now we've got a 3rd plan which differs with "Using where" and that doesn't work anymore. - This patch fixes it by removing 'Using where' from EXPLAIN output. Test coverage is somewhat reduced but probably still ok as PBXT and nested outer join processing have no interaction and we don't expect any bugs here.
BR Sergey -- Sergey Petrunia, Software Developer Monty Program AB, http://askmonty.org Blog: http://s.petrunia.net/blog