[Chandler-dev] (no subject)

[D John Anderson]

On Jul 3, 2007, at 12:45 PM, Heikki Toivonen wrote:


>
>
>> Having the test framework keep track of which test failed and  
>> report it
>> make more sense to me.
>>
>
> This seems to contradict what you have been saying earlier. assert is
> not test framework as I understand it.
>

The recorded tests are run by code that catches asserts and logs  
them. It also reports which test fail.


>
>
>> Whether or not we use assert to raise errors as I propose, or  
>> explicitly
>> raise AssertionErrors as you propose, I think it makes more sense  
>> to run
>> with asserts for all the tests except for performance tests. I don't
>> know of any case where running asserts and __debug__ code  
>> introduced a
>> bug and I know of many cases where they detected a bug.
>>
>
> And I am saying that this misses important cases of failure, and  
> that we
> should do the following:
>
> * Tbox should run release with -O and debug without -O
> * Developers, if they run just release or debug, should run without -O
> (but if they run both then that becomes the Tbox case)
>

I think probably makes sense to also be running release without -O --  
which I suspect is likely to find more bugs than running release with  
-O.