testing stockfish 17 from test repo- it works

[entropyfoe]

A brief report on this testing package.

Thanks to the team (Stevo?) for packaging this up for us at MX that play chess.
stockfish is the strongest chess engine in the world right now, and we have free access to the latest version with a few clicks ! (nearly, I think 17.1 is out).

The new version is a rating bump of 50 ELO over the main repo's stockfish An unbelievable, inhuman rating of 3850, Magnus would have no chance.

So, I did two installs, first on a fully updated MX-21 system. In both cases I used the mx package installer, since it is easier than enabling repos , and safer, than in synaptic.
I enable the test repo tab, and check the stockfish to get 17.0.1.

Installs fine. stockfish is a chess engine, but you also need a chess interface, many programs can connect to stockfish. I use from the MX repos, scid, a powerful chess database.
I configured scid to use the new stockfish, saved, and it worked, perfectly.

The second install is on my main machine, MX-23.5. This time it seemed like the mxpi installer got stalled out at the princeton math repo (?).
So I gave up. A week later I tried again, and it worked.
When I tried to run stockfish, by configuring scid, when I saved the configuration, I got an error "the analysis engine terminated without exit code"

I redid the scid configuration with a " new" engine called stockfish_17. It used the same stockfish command and directory, and when I saved , it worked. It is working great, it seems fast, I will do more testing. The error may have been on the scid side, but in any case, it seems to work great on both MXs.

[Stevo]

Thanks for the tests. I've managed to get the source, nnue (neural network files), and refresh the patches for 17.1, so that will be in test pretty soon. We'll put 17 in main.

[entropyfoe]

Thanks Steveo. I thought that was you maintaining and packaging this.
As usual this puts MX Linux as living up to the "BEST" as the benchmark.

Here is the announcement for 17.1 on lichess

https://lichess.org/@/stockfishnews/blo ... y/POlhf9l0

[Stevo]

I have to get the stockfish source code, then run "make net" in the /src folder for it to download the neural network .nnue files, then compress that into the .orig source tarball. Stockfish itself compiles very quickly. Debian's is a lowest common denominator build, so the package could be rebuilt by a user to take advantage of more advanced CPU features, such as AVX2, most of which are not enabled.