This bug was introduced by the patch for WL#3220.
If the memory allocated for the tree to store unique elements
to be counted is not big enough to include all of them then
an external file is used to store the elements.
The unique elements are guaranteed not to be nulls. So, when
reading them from the file we don't have to care about the null
flags of the read values. However, we should remove the flag
at the very beginning of the process. If we don't do it and
if the last value written into the record buffer for the field
whose distinct values needs to be counted happens to be null,
then all values read from the file are considered to be nulls
and are not counted in.
The fix does not remove a possible null flag for the read values.
Rather it just counts the values in the same way it was done
before WL #3220.
This bug could result in returning 0 for the expressions of the form
<aggregate_function>(distinct field) when the system variable
max_heap_table_size was set to a small enough number.
It happened because the method Unique::walk() did not support
the case when more than one pass was needed to merge the trees
of distinct values saved in an external file.
Backported a fix in grant_lowercase.test from mariadb 5.5.
Fixed warnings by valgrind for sum_distinct.test
Enable buffered-record-reads after filesort for InnoDB tables with short primary key
Enabled sort-with-data for MyISAM temporary files
BitKeeper/etc/ignore:
added tools/mysqltestmanager
client/mysqltest.c:
Ensure that BIG_TEST is always set to 0 or 1
Fix the 'eval' also honors 'require'
mysql-test/mysql-test-run.sh:
Enlarge InnoDB table space for --big tests
mysql-test/r/heap.result:
Fix after adding more optimzation for filsort
mysql-test/r/sum_distinct.result:
Move 'slow' part of test to sum_distinct-big.test
mysql-test/t/heap.test:
Ensure that results are indpendent of optimizer
mysql-test/t/sum_distinct.test:
Move 'slow' part of test to sum_distinct-big.test
sql/filesort.cc:
Use 'sort with data' also on temporary files and with INSERT ... SELECT
sql/ha_innodb.h:
Remove HA_FAST_KEY_READ to enable buffered-record-reads after filesort
sql/handler.h:
More comments
sql/mysql_priv.h:
A bit smaller limit for cache for buffered-records-read (after testing)
sql/records.cc:
Don't use buffered-record-reads if ref_length > MAX_REFLENGTH
Fixed warning from valgrind in 'sum_distinct'
sql/sql_select.cc:
Ensure that tempory tables has query_id set for all fields
(Required for sort-with-data to work on temp files)
