[MCOL-5002] ExeMgr crash with certain subselects Created: 2022-02-25 Updated: 2022-03-14 Resolved: 2022-03-14 |
|
| Status: | Closed |
| Project: | MariaDB ColumnStore |
| Component/s: | ExeMgr |
| Affects Version/s: | 6.3.1 |
| Fix Version/s: | 6.3.1 |
| Type: | Bug | Priority: | Blocker |
| Reporter: | David Hall (Inactive) | Assignee: | Daniel Lee (Inactive) |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Sprint: | 10.0.20 |
| Description |
|
Consider this query:
With version 6.2.3-1, this query completes correctly. With version 6.3.1 (unreleased) it raises sig6 SIGABT. Some analysis shows that the underlying problem has always existed. Starting with 6.3.1, we have enabled more extensive array checking, and this exposes the problem. In tuplejoiner.cpp around line 604, we find:
There is no guarantee that the index in largeKeyColumns is a valid index in smallRG. largeKeyColumns[n] should only be used to index into largeRG, not smallRG. So the task here is to determine whether the author intended to comapare smallRG to smallKeyColumns[0] or should it be largeRG to largeKeyColumns[0]. Frankly, in this particular context, it may not matter, but it is important that we compare apples to apples. |
| Comments |
| Comment by Daniel Lee (Inactive) [ 2022-03-14 ] | ||||||||||||||||
|
Build verified: 6.3.1-1 (#4049) I could not reproduce the issue using drone build 3571. I don't know at what point the 6.3.1 was broken. I verified build 4049 and it is working fine. I verified it using both 1gb and 10gb databases.
|