[MDEV-28567] Assertion `0' in open_tables upon function-related operations Created: 2022-05-14 Updated: 2022-07-13 Resolved: 2022-06-29 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Locking, Stored routines, Views |
| Affects Version/s: | 10.9 |
| Fix Version/s: | 10.9.2 |
| Type: | Bug | Priority: | Major |
| Reporter: | Elena Stepanova | Assignee: | Aleksey Midenkov |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Attachments: |
|
||||||||||||
| Issue Links: |
|
||||||||||||
| Description |
|
The test case fails for me every time on a 10.9 debug build, but it is still a concurrent test, so maybe it can fail to fail in some circumstances. Try to run with --repeat then.
Reproducible with at least MyISAM and InnoDB. The failure started happening on 10.9 after this commit:
However, I can't claim it is a regression, because the commit added the assert itself:
|
| Comments |
| Comment by Aleksey Midenkov [ 2022-06-18 ] | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Good and Bad script behaviour:
Good: SELECT seized v1; ALTER waiting v1; CREATE waiting f; SELECT seized t1; COMMIT released t1, v1, f; etc... Bad: ALTER seized v1, t1; SELECT waiting v1; CREATE waiting f; ALTER waiting f, deadlock f. In Good ALTER is waiting on v1 until COMMIT releases f and that avoids deadlock situation. For details see diff between attached good.log and bad.log. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Aleksey Midenkov [ 2022-06-20 ] | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Please review bb-10.9-midenok | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Oleksandr Byelkin [ 2022-06-21 ] | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
OK to push |