Re: [glob2-devel] Checking for nondeterminism

[Erik Søe Sørensen]

Leo Wandersleb skrev:

> actually there is a checksum computed over the game state. so as i 
> guess you must have seen that i must have misunderstood. maybe you can 
> explain what exactly you want to compare.

Well, I can't claim that I've read all, or even much, of the 
documentation - and I haven't touched the networking parts yet (but I 
found the relevant wikipage just now and am going to).

But I notice that the game state in this case does not include the 
gradients. Not does it include guardArea, clearArea or canRessourcesGrow.
Not that this is not fine and reasonable for normal playing; for serious 
testing-before-releasing, however, it'd probably be nice to know that 
also the values which ended up not being used (such as probably most of 
the gradient value) actually match.
After all, each test run should test on behalf of a thousand real 
network games ;-)

> besides that, automated testing with multiple clients on one machine 
> is not yet implemented.

About that: If this is implemented, it would probably be best if it was 
possible to test two different glob2 binaries against each other... i.e. 
what is really needed is some way to start a multiplayer game from the 
command line, and some separate script to set up several versions 
playing the game  on the same machine, using that mechanism.

/Erik