UPDATE 2: Found it! 1) I reproduced the problem in revision that you referred to 2) It passed on Google MariaDB as...ta ta... our (yet unpublished) sparse gtid index (that maps between gtid and offset) made it skip over the out of order transaction. this was pure "luck", if I disable it...it will scan the entire binlog and get the duplicate key... puh...that took a while to track down. --- I'll now revise my patch so that it works... /Jonas On Thu, Mar 5, 2015 at 2:55 PM, Jonas Oreland <jonaso@google.com> wrote:
strange indeed! i'll test the specific git revision...
/Jonas
On Thu, Mar 5, 2015 at 2:42 PM, Kristian Nielsen <knielsen@knielsen-hq.org
wrote:
Jonas Oreland <jonaso@google.com> writes:
update: just tested your new testcase with my patch...and it passes... i.e the test does not expose any problem in my patch
Strange. It failed for me with duplicate key error, when I tested it with the patch you attached applied to f4f37533a09b9776e8d5ac3f3a27957f553c9043.
- Kristian.