[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] NNGS games in regression/games/nngs

From: Heikki Levanto
Subject: Re: [gnugo-devel] NNGS games in regression/games/nngs
Date: Tue, 8 Jan 2002 10:01:17 +0100
User-agent: Mutt/1.2.5i

On Mon, Jan 07, 2002 at 04:36:27PM -0500, Trevor Morris wrote:
> Accessible, yes, but not so easy to make regression tests on.
> For example, here are the steps I have to take:
>  - download the game.
>  - figure out a reasonable name for it, and hope no one else has
>      named it something different.
>  - when submitting the test cases, make sure to submit this new file.
> These are just enough steps to make it annoying to create and submit 
> tests on these games.
> >I think automatically entering every game into regression/games is
> >a bad idea.
> I disagree.  As a compromise, perhaps enter just the games that
> are bold on the recent games page?

Anopther compromise:Keep the games where they are, but at least give them
unique names that can be used in the regressions directly. This solves the
second step in Trevors complaints, without inflating the CVS at all.
> Also, perhaps it would be useful to store the annotated version of
> the game, after re-running it through the most recent GNU Go (as
> of being added), so that GNU Go's current best moves are indicated
> in the SGF file, rather than the (possibly time-constrained) actual
> moves?

What if the problem only occurs with a time-constraint. Those are 
worth fixing as well, are they not?

- Heikki

Heikki Levanto  LSD - Levanto Software Development   <address@hidden>

reply via email to

[Prev in Thread] Current Thread [Next in Thread]