Winboard Forum

Hi all,

Alessandro Scotti found a small bug in Glaurung 1.0.1: When a search was aborted (happens when the user stops the search, or when the maximum allocated time for a move is consumed), a few incorrect draw scores would be written to the hash table. This probably doesn't hurt the program noticably when playing games, but it is very annoying when running the program in analysis mode (when searches are aborted prematurely every time the user makes a move).

I have now released Glaurung 1.0.2, which fixes this bug. There are no other changes compared to 1.0.1. For those of you who create rating lists: Don't bother to restart your tests, but please use 1.0.2 in future tournaments. Let it keep the rating of 1.0.1.

Thanks to Alessandro for pointing out the bug! Yet another example of the benefits of open source.

Download from my Glaurung and Scatha page.

Tord

Hi Tord ,

thanks for the new version. Many small improvements over the time can also sum up to a bigger advance I think.

So I encourage all CEGT testers to let the new version replace Glaurung 1.0.1. For your info: In our rating list all games for Glaurung 060105, Glaurung 1.0, 1.0.1 and 1.0.2 are combined now to Glaurung 1.0.2. But you will surely tell us when there is a version that you want to see with separate stats.

In the league the start version (1.0.1) will have to finish the tournament, as this is a rule.

Looking forward to more Glaurung versions and games.

Best Regards
Heinz

Hi Heinz!

Heinz van Kempen wrote:Hi Tord ,

thanks for the new version. Many small improvements over the time can also sum up to a bigger advance I think.

So I encourage all CEGT testers to let the new version replace Glaurung 1.0.1. For your info: In our rating list all games for Glaurung 060105, Glaurung 1.0, 1.0.1 and 1.0.2 are combined now to Glaurung 1.0.2.

Yes, that's what I wanted. Restarting the tests with 1.0.2 would be an awful waste of time.

But you will surely tell us when there is a version that you want to see with separate stats.

I can tell you already now: All versions of the form 1.0.x should be combined. I hope 1.0.2 will be the last such version, but perhaps I'll discover some more minor bugs. When I release a version which is not of the form 1.0.x, it should be given separate stats.

In the league the start version (1.0.1) will have to finish the tournament, as this is a rule.

No problem. I expect the rating difference between the two versions to be less than one Elo point anyway.

Tord