Bug #7405 | group_concat with distinct and rollup => ignores distinct in some rows | ||
---|---|---|---|
Submitted: | 18 Dec 2004 16:12 | Modified: | 1 Jun 2005 3:08 |
Reporter: | Martin Friebe (Gold Quality Contributor) (OCA) | Email Updates: | |
Status: | Closed | Impact on me: | |
Category: | MySQL Server | Severity: | S3 (Non-critical) |
Version: | 4.1.7 | OS: | FreeBSD (freebsd (probably others)) |
Assigned to: | Ramil Kalimullin | CPU Architecture: | Any |
[18 Dec 2004 16:12]
Martin Friebe
[18 Dec 2004 16:16]
Martin Friebe
# the order by, does also break, when used with rollup (I report this to the same bug, as it thems to be quiet related) select a, group_concat(distinct b order by b) from t1 group by a with rollup; +------+-------------------------------------+ | a | group_concat(distinct b order by b) | +------+-------------------------------------+ | 1 | 0,0,0,0,0,0 | | NULL | 1,2,3 | +------+-------------------------------------+ # all values in the concat are 0
[18 Dec 2004 17:21]
Martin Friebe
It seems that the values with order by are not always 0, but sometimes random values (either from other tables, or maybe pointers?) This could explain that I had the above crashing the servers (on a different set of data (but cant reproduce)
[17 May 2005 7:12]
Ramil Kalimullin
I got server crashes for that and similar queries.
[31 May 2005 6:12]
Ramil Kalimullin
fixed in 4.1.13
[31 May 2005 16:27]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/internals/25395
[1 Jun 2005 3:08]
Paul DuBois
Noted in 4.1.13 changelog.