Re: [Alephmodular-devel] A note on films

[Woody Z. I. <woo...@sb...>]

On Sunday, January 19, 2003, at 01:35  AM, Br'fin wrote:

> On Saturday, January 18, 2003, at 03:38  PM, Woody Zenfell, III wrote:
>>
>> (I assume it's obvious that the "checkfile" would be generated from 
>> the filmfile as early as possible... then occasionally in the game's 
>> development lifetime, that original filmfile and checkfile would be 
>> used to make sure everything's still working the same way internally.)
>
> Output a game tick and all the attitude changes of objects within that 
> tick.
>
> I agree that this could be useful, unfortunately I'm unsure when such a 
> test harness could be implemented. (It seems to require we have our 
> interfaces defined already :/ )

Right, this could be a stronger form of checkfile for later use.

In the meantime, one could dump the value of get_random_seed() at the 
beginning of each tick (when generating) or compare the value against 
the checkfile's value (when playing back a film that has a checkfile).  
Should be reasonably straightforward, no?  And you can check for 
divergence throughout the restructuring against the *current* logic, not 
merely the logic you might end up with post-restructuring 
(post-test-harnessing).

Actually, if the (exact) same checkfile mechanism were used in A1, we 
could make sure that A1 is still film-file compatible with M2 (well, or 
at least the current AM) (which A1 may not be; even very subtle changes 
can break films).

> I rather like the idea of the diffable checkfiles. You don't just know 
> that things diverged, but you know the tick and calculations where they 
> did.

Diffable is good - that's effectively how I used this technique to debug 
A1 - but having the machine check for you (while the film is playing 
back) would be even more convenient I think.

Woody