Re: bzr commit into mysql-5.1-bugteam branch (martin.hansson:3201) Bug#48459

[Tor Didriksen]

On Tue, 17 Nov 2009 16:49:13 +0100, Martin Hansson  
<Martin.Hansson@stripped> wrote:

>
>>
>> Shouldn't you do the same swap a couple of more places in this file?
>>
>> -- didrik
> Maybe. Who knows which code paths actually end up being taken? I don't  
> want to make changes in the code that aren't covered by a test case. In  
> fact, I think we have code coverage tools that catch this. At least we  
> used to with PB1. Also, it makes it far too easy to become the target of  
> blame for regressions later on.
>

OK, makes sense.

-- didrik

> The good news is that Phillip is tweaking the RQG so that we can get  
> more code coverage out of range opt. It will be more bugs and more  
> csets, but everything covered with test cases.
>
> The bottom line is, SQL is too blunt a tool to test such an intricate  
> data structure as the range optimizer. We should really have a unit test  
> calling the functions inside one by one and examining the output. Maybe  
> if we get a lot of bugs on the range optimizer I can motivate our  
> managers to get the time needed to set something up.
>
> /Martin