[Maria-developers] Please review MDEV-9604 crash in Item::save_in_field with empty enum value
Hi Sergei, It fixes the reported crash, and additionally inconsistencies when comparing temporal columns to empty strings: In some cases it returned rows, in other cases it returned no rows, depending on: - the engine - the type of comparison arguments (columns vs literals). - presence of an index Now everything works in the same way. Note, this additional problem is not 10.1 specific, it existed in 10.0. But 10.0 and 10.1 worked differently in certain cases. Thanks!
Hi, Alexander! On Mar 22, Alexander Barkov wrote:
Hi Sergei,
It fixes the reported crash, and additionally inconsistencies when comparing temporal columns to empty strings:
In some cases it returned rows, in other cases it returned no rows, depending on: - the engine - the type of comparison arguments (columns vs literals). - presence of an index
Now everything works in the same way.
Note, this additional problem is not 10.1 specific, it existed in 10.0. But 10.0 and 10.1 worked differently in certain cases.
Ok to push. In 10.0, right? And, by the way, please send a full commit next time, not just a diff. It was quite a challenge to figure out what was the bug and what were you fixing without a commit comment :( Regards, Sergei Chief Architect MariaDB and security@mariadb.org
Hi Sergei, On 03/22/2016 09:33 PM, Sergei Golubchik wrote:
Hi, Alexander!
On Mar 22, Alexander Barkov wrote:
Hi Sergei,
It fixes the reported crash, and additionally inconsistencies when comparing temporal columns to empty strings:
In some cases it returned rows, in other cases it returned no rows, depending on: - the engine - the type of comparison arguments (columns vs literals). - presence of an index
Now everything works in the same way.
Note, this additional problem is not 10.1 specific, it existed in 10.0. But 10.0 and 10.1 worked differently in certain cases.
Ok to push. In 10.0, right?
The patch depends on my earlier changes in equal field propagation in 10.1. Okey to push to 10.1 only?
And, by the way, please send a full commit next time, not just a diff. It was quite a challenge to figure out what was the bug and what were you fixing without a commit comment :(
Thanks. Agreed.
Regards, Sergei Chief Architect MariaDB and security@mariadb.org
Hi, Alexander! On Mar 22, Alexander Barkov wrote:
On 03/22/2016 09:33 PM, Sergei Golubchik wrote:
Hi, Alexander!
On Mar 22, Alexander Barkov wrote:
Hi Sergei,
It fixes the reported crash, and additionally inconsistencies when comparing temporal columns to empty strings:
In some cases it returned rows, in other cases it returned no rows, depending on: - the engine - the type of comparison arguments (columns vs literals). - presence of an index
Now everything works in the same way.
Note, this additional problem is not 10.1 specific, it existed in 10.0. But 10.0 and 10.1 worked differently in certain cases.
Ok to push. In 10.0, right?
The patch depends on my earlier changes in equal field propagation in 10.1. Okey to push to 10.1 only?
It means that 10.0 patch won't merge into 10.1 anyway. Yes, push into 10.1 then and please create a bug report for 10.0. Don't spend time on fixing it now, though, 10.0 release is still few weeks ahead. Regards, Sergei Chief Architect MariaDB and security@mariadb.org
Hi Sergei, On 03/22/2016 09:48 PM, Sergei Golubchik wrote:
Hi, Alexander!
On Mar 22, Alexander Barkov wrote:
On 03/22/2016 09:33 PM, Sergei Golubchik wrote:
Hi, Alexander!
On Mar 22, Alexander Barkov wrote:
Hi Sergei,
It fixes the reported crash, and additionally inconsistencies when comparing temporal columns to empty strings:
In some cases it returned rows, in other cases it returned no rows, depending on: - the engine - the type of comparison arguments (columns vs literals). - presence of an index
Now everything works in the same way.
Note, this additional problem is not 10.1 specific, it existed in 10.0. But 10.0 and 10.1 worked differently in certain cases.
Ok to push. In 10.0, right?
The patch depends on my earlier changes in equal field propagation in 10.1. Okey to push to 10.1 only?
It means that 10.0 patch won't merge into 10.1 anyway. Yes, push into 10.1 then and please create a bug report for 10.0.
Sure. I'll file a separate report for 10.0. Just to clarify: in 10.0 it does not crash. It only returns non-consistent results when comparing temporal columns to empty strings. So the problem is less serious.
Don't spend time on fixing it now, though, 10.0 release is still few weeks ahead.
Okey.
Regards, Sergei Chief Architect MariaDB and security@mariadb.org
participants (2)
-
Alexander Barkov -
Sergei Golubchik