[MDEV-6240] Wrong "Impossible where" with LIST partitioning Created: 2014-05-15 Updated: 2020-12-08 Resolved: 2017-10-03 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Optimizer |
| Affects Version/s: | 5.5.37, 10.0.11 |
| Fix Version/s: | 5.5.39, 10.0.13 |
| Type: | Bug | Priority: | Major |
| Reporter: | Federico Razzoli | Assignee: | Sergei Petrunia |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | partitioning, upstream | ||
| Issue Links: |
|
||||||||
| Description |
|
I think this is the simplest way to repeat the bug:
Notice that the inserted row DOES match the queries. Without the WHERE, the SELECT finds it. If you re-define the table without the first value for p0, the bug disappears. |
| Comments |
| Comment by Federico Razzoli [ 2014-05-15 ] | ||||||||||||||||||||||||||||||||||||||||
|
In this case, an impossible where is NOT detected. I suspect it is the same bug, but if you want me to report it separately, please let me know.
| ||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2014-05-30 ] | ||||||||||||||||||||||||||||||||||||||||
|
Hi Federico, Thanks for the report and the great test case. It seems to be an upstream bug, and as you know we report those to Oracle as well. Would you like to do it yourself, or should I create a bug report on your behalf? | ||||||||||||||||||||||||||||||||||||||||
| Comment by Federico Razzoli [ 2014-05-30 ] | ||||||||||||||||||||||||||||||||||||||||
|
Sorry, I didn't test it on MySQL before. Reported: | ||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2014-05-30 ] | ||||||||||||||||||||||||||||||||||||||||
|
Thanks! Let's see if it gets fixed in 5.5 any time soon; otherwise we might need to fix it ourselves. | ||||||||||||||||||||||||||||||||||||||||
| Comment by Olivier Bertrand [ 2014-06-10 ] | ||||||||||||||||||||||||||||||||||||||||
|
Apparently it is a MySQL regression bug that is or will be Fixed in MySQL 5.5.39, 5.6.20, 5.7.5. | ||||||||||||||||||||||||||||||||||||||||
| Comment by Sergei Petrunia [ 2014-07-23 ] | ||||||||||||||||||||||||||||||||||||||||
|
Thanks for the pointer Olivier. Looked at their fix, seems to be correct. I've backported it and added all the testcases. | ||||||||||||||||||||||||||||||||||||||||
| Comment by Olivier Bertrand [ 2017-09-25 ] | ||||||||||||||||||||||||||||||||||||||||
|
Testing recently this example with MariaDB 10.2.6 this bug is not fixed! | ||||||||||||||||||||||||||||||||||||||||
| Comment by Olivier Bertrand [ 2017-09-25 ] | ||||||||||||||||||||||||||||||||||||||||
|
Not fixed in MariaDB 10.2.6 | ||||||||||||||||||||||||||||||||||||||||
| Comment by Sergei Petrunia [ 2017-10-02 ] | ||||||||||||||||||||||||||||||||||||||||
|
I am using the following MTR .test file:
and I get this result:
both in 10.2.9 and in 10.2.6. | ||||||||||||||||||||||||||||||||||||||||
| Comment by Sergei Petrunia [ 2017-10-02 ] | ||||||||||||||||||||||||||||||||||||||||
|
Also checked Frederico's testcase from the first comment (The issue reported there is different: it is about NOT detecting "Impossible WHERE" when it's actually possible to do so): I get Impossible WHERE on both 10.2.6. and 10.2.9 :
| ||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2017-10-02 ] | ||||||||||||||||||||||||||||||||||||||||
|
bertrandop, could you please clarify what's not fixed – what exactly you tried, and what you got? I don't see any problem on 10.2.6 with any of the above test cases. | ||||||||||||||||||||||||||||||||||||||||
| Comment by Olivier Bertrand [ 2017-10-03 ] | ||||||||||||||||||||||||||||||||||||||||
|
Neither do I! |