On 02/21/11 10:18 AM, Øystein Grøvlen wrote:
> On 02/18/2011 01:22 PM, Olav Sandstaa wrote:
>> Thanks for fixing this problem with the best_ref array not keeping its
>> sort order.
>> The changes look fine and I have only some minor comments (see inline).
>> I understand why you have introduced the large "perl generator" for
>> testing this. I think it is a good idea but I do not quite like that the
>> "failure criterion" will be a test timeout. Unfortunately I do not have
>> any other or better suggestions for how to make such a test. This has
>> also increased the time for running the greedy_optimizer test from less
>> than 1 second to almost 1 minute. Some things to consider: will
>> "pruning" some of the generated test cases help? or maybe move this to a
>> new separate test? I leave you to decide and if nobody else have
>> objections it is fine to keep it as it is.
> I think a 1 minute test is to long for main suite. I suggest either to mark it as big
> (then it is run when supplying the
> --big option to mtr, and weekly? in pushbuild.)
It turns out that it was the filling of the ~60 InnoDB tables used in the test which
consumes most of the time. The size of these tables has been reduced to 25% of their
original size, and the test now completes in ~8s.
>> OK to push when you have looked at the minor comments (mostly on coding