Hi!
>>>>> "Peter" == Peter Zaitsev <pz@stripped> writes:
Peter> Today I was truing to run test suite for both reiserfs and ext2 to figure
Peter> out much the first one is faster or slower (I'll post results later) And
Peter> I've found some questions should be asked.
Peter> 1) Do I understand right - I have to use --opt option to have results which
Peter> can be compared to other in database or to results on MYSQL website
No you don't have to use --opt; This option doesn't do anything for
run-all-tests.
Note also that most of the tests run on our website is comparison to
other SQL servers. We haven't had time to do a full run on all
the machines we have available; We shall however try to collect such
a test relatively soon.
Peter> 2) Some of the test return tesult marked with "?" As I understand this is
Peter> not quete ok but I can't understand if this meens the result produced by
Peter> script is wrong or just one of the queries returned wrong result ?
Peter> Here is the example:
Peter> order_by 55 27.91 16.76 0.00 10
Peter> order_by_key 39 28.01 11.00 0.00 10
Peter> outer_join 74 0.01 0.01 0.00 10
Peter> outer_join_found 72 0.01 0.00 0.00 10
Peter> outer_join_not_found 604 0.14 0.01 0.00 500 +
Peter> outer_join_on_key 58 0.00 0.00 0.00 10
Peter> select_1_row 3 0.42 0.67 0.00 10000
Peter> select_2_rows 3 0.57 0.73 0.00 10000
Peter> select_big 74 41.51 18.43 0.00 10080
Peter> select_diff_key 219 0.40 0.07 0.00 500
Peter> select_distinct 13 2.23 0.79 0.00 800
Peter> select_group 74 2.32 0.50 0.00 2911 ?
Peter> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Peter> select_group_when_MANY_tables 28 0.98 1.92 0.00 10000
The ? means that there exists a warning for some query in some of the
result files in the output directory. Try searching after the word
'warning' in output/* for more information. (It may be harmless).
Peter> 3) Is "+" marked lines appears to be OK ? I assume Yes.
Yes; Numbers marked with '+' are estimated according to previous runs
because the query took longer than a given time-limit to finish. The
estimation shouldn\'t be far from the real result thought.
Peter> Best Regards
Peter> Peter Zaitsev
Regards,
Monty
| Thread |
|---|
| • Test Suite | Peter Zaitsev | 13 Feb |
| • Test Suite | Michael Widenius | 15 Feb |