Kristian Nielsen wrote:
"Philip Stoev" <pstoev@askmonty.org> writes:
This code was added as a fix for MySQL bug #38005 and then removed as a fix for bug #46639
Yes.
Turns out that the removal part of Bug#46639 was lost in MariaDB in this merge:
revno: 2732 [merge] revision-id: igor@askmonty.org-20091112043128-yd6odg8zr5ribjal parent: psergey@askmonty.org-20091104224158-nk2s2luvlqwa02bl parent: igor@askmonty.org-20091110023239-vgttyweq2qmh0y25 committer: Igor Babaev <igor@askmonty.org> branch nick: maria-5.2-vcol timestamp: Wed 2009-11-11 20:31:28 -0800 message: Merge of the patch introducing virtual columns into maria-5.2
This merge gives the following conflict (and two more like it) in sql/ha_partition.cc:
<<<<<<< TREE /* MyISAM engine can fail if we call index_first() when indexes disabled that happens if the table is empty. Here we use file->stats.records instead of file->records() because file->records() is supposed to return an EXACT count, and it can be possibly slow. We don't need an exact number, an approximate one- from the last ::info() call - is sufficient. */ if (file->stats.records == 0) { error= HA_ERR_END_OF_FILE; break; } error= file->ha_index_first(buf); ||||||| BASE-REVISION /* MyISAM engine can fail if we call index_first() when indexes disabled */ /* that happens if the table is empty. */ /* Here we use file->stats.records instead of file->records() because */ /* file->records() is supposed to return an EXACT count, and it can be */ /* possibly slow. We don't need an exact number, an approximate one- from*/ /* the last ::info() call - is sufficient. */ if (file->stats.records == 0) { error= HA_ERR_END_OF_FILE; break; } error= file->index_first(buf); ======= error= file->index_first(buf);
>> MERGE-SOURCE
In MariaDB 5.2, these conflicts were resolved by keeping the lines deleted in the patch, this reverting this part of the patch for Bug#46639.
Igor, did you deliberately revert the of the patch like this, or is it just a mistake during merge?
If a mistake, I can push a correction for this to 5.2.
If deliberate, should the reverting also be done in MariaDB 5.1, or is it only necessary for 5.2+?
Probably it was a mistake. Regards, Igor.
- Kristian.
[Note that for the tokudb problem, I believe they still need to fix their engine to not return zero estimate as per comments in handler.h. This appears to be just a symptom.]