Comments/Ratings for a Single Item
A convinient way to share code would be to fork jocly on github in the name of cvp to obtain contributions. The project seems to compile on windows but not on linux/apple. Do you want to give it a try that way? It would also be interesting to upgrade the models on some games that have upgraded versions...
H. G. Muller wrote on Sat, Oct 28, 2023 08:25 AM EDT:The Jocly version on CVP is a bit problematic, in that the variants added by me were added by hacking them directly into the library. So there doesn't exist a source code that would produce the current library on compiling. I have started to backport some of the added variants to source, and for some variants this has been done, but this project is not finished yet. It stalled when I ran into a problem w.r.t. opening books, which I did not know how to solve.
If you can provide the source of a model and view file for the variant, I can try to hack it into the library as well. I think I still remember how to do this.
If we want to propose a jocly implementation of the game :
Should we add an external link or can we send you a github pull request for the repository of www.chessvariants.com/play/jocly/ ?
5 comments displayed
Permalink to the exact comments currently displayed.
Well, it doesn't seem to make any difference whether I fork it on GitHub or on my own on-line git repository (which I already did; branch 'hgm' there). The real problem is that the latest version in that fork is still not able to reproduce the library that we have at CVP, not which website it is on.
I was never able to compile the Jocly source on Windows, but I can compile it on a Linux 16.04 system I have running in VirtualBox. (Which I installed solely for that purpose.)
The problem that halted further backporting of the games I added was that the hashing scheme in the common chessbase model file was flawed, and turned out to be woefully inadequate for detecting repetitions (and thus perpetual checks) in variants with piece drops. But is is also used to access the opening book, so fixing the position hashing would break the book. Unfortunately Michel Guiterrez did not have the code anymore through which he created the books, so we could not redo the books using the fixed key.
Now only Chess and Xiangqi appear to have opening books, so we could fork the chessbase model file, and let the build process keep using the old ones for these two games. This seemed a very ugly solution, though.