Saturday, April 20, 2013

Fun with Bugs #6 - MySQL 5.6.11

Last time I had written about bugs here it was a beginning of February, and hardly somebody could imagine that we were going to have 2 more months of winter, 2 more months to wait for next MySQL 5.6 GA release or that in 2 months and two weeks I'll be speaking about MySQL in public for 50 minutes based on my own presentation, and speaking about bugs (surely)...

In any case, MySQL 5.6.11 looks promising and had got a lot of bugs fixes (check at least this nice post by Todd Farmer). Time for me to get back to bugs I've mentioned in the previous "Fun with Bugs" issue and then check for some new ones already reported in 5.6.11.

So, from that list (that I've actually sent to customers first weeks after MySQL 5.6 GA more than once), we still have the following bugs NOT fixed in 5.6.11 it seems:

Bug #68299 (EXPLAIN output for UPDATE/DELETE can still be wrong)
- Bug #68251 (semisync replication) - that's sad, actually...
- Bug #68250 (semisync replication) - this is now in "No feedback" status, and it's strange as, even if reporter does not reply to further comments or questions, it was repeated by Oracle engineers. Maybe patch is not ready to use, but the bug is real.
- Bug #68220 (minor, when replication info is stored in tables)
- Bug #68192 (data types) - it was reported at 5.5.29 and 5.6.9 times and got no visible attention since verification
- Bug #68171 (missing manual) - still no comments, since 5.6 RC times
- Bug #68144 (custom character sets, regression) - no visible progress since 5.6 RC times
- Bug #68097 (incomplete manual) - here we've got a blog post on the topic at least, and even a definitive guide from Domas, but not sure if manual itself had improved. Still "Verified".
- Bug #68079 (scalability for joins) - this my report had not got any visible attention since February 1. Not a big deal, 5.6 still a bit better than 5.5 in that case, but it seems I'll have to ask DimitriK about it in person during the Conference
- Bug #68041 (zero date, regression) - still no comments
- Bug #67982 (partitioning) - still no comments, since last year

So, 11 out of 20 bugs from my original list are still there, in 5.6.11, even though several of them were reported by one of top Community bug reporters before GA announcement. 

Now, let's check some recent serious bugs affecting 5.6.11:

Bug #68350 - optimizer bug with a simple test case from Shane (since February 16) that is still repeatable in 5.6.11
Bug #68525  - replication bug verified since February 28
Bug #69005 - regression bug in optimizer

As a result, I'd say that 5.6.11 is surely better than 5.6.10 (nobody should continue using that version now it seems), but 5.6 GA still has a long way ahead before all claims about the best version ever become definitely true. It seems that it is still a lot of work to do on replication, optimizer and even InnoDB bug fixing (to say nothing about the manual).