cutechess-cli 0.6.0 released
Posted: Fri Jul 12, 2013 6:52 pm
For those of you who are not on our mailing list:
It's been a really long time since the last release, so we decided to delight our users with a new release of cutechess-cli 0.6.0. As usual, both source code and binary packages for supported platforms can be downloaded at https://github.com/cutechess/cutechess
Anyone using cutechess-cli 0.5.1 or earlier is encouraged to update. If you're using the version that was made for the Stockfish testing framework, there's probably no reason to update - unless you use the SPRT feature which was improved lately.
New features or changes:
* CLOP glue script is compatible with Python 3
* Opening positions can be picked from EPD files
* ECO codes and opening names as included in PGN output
* Custom XBoard engine options are supported. This should make Gaviota users happy.
* Added "trust" engine option that makes cutechess-cli trust any result claim from the engine.
* The correct PGN move number format is used in games where black moves first
* The precision of the relative score for engine A is increased
* Initialization timeout period is longer so that some slowly starting engines are detected properly.
* A Sequential Probability Ratio Test (SPRT) can be used as a termination criterion for the match.
* The format for setting some options has changed to a less error-prone format
* The "-draw" option now has a "movecount" parameter for setting the number of consecutive scores that have to be within the chosen treshold.
* The "-pgnin" and "-pgndepth" options have been replaced be the "-openings" option that supports both EPD and PGN formats, and two ways to pick openings - sequentially or randomly. In "random" mode the openings file is first indexed, so it can take a few seconds with large files (> 100 MB), but it's only done once. It's also possible to set the number of the first opening to be picked (only in sequential mode).
* Gaviota tablebases can be used to adjudicate games
Bugs fixed:
* Missing command or invalid protocol in engines.json results in a crash
* UCI combo options couldn't be set properly on some systems
* Memory leak when the "-pgnout" option is not used
* On some systems the random number generator produced the same sequence of numbers in each thread, which would result in the same opening used all the time in tournaments with more than two players.
* Threading problems. The multi-threading code is now simpler and more effective and some possible race conditions have been eliminated. If you previously saw warning messages about QThread objects being destroyed while the thread is running, cross-thread killTimer() warnings or random crashes at the start of a match, then this fix should make you very happy.
* Cutechess-cli used to stall if it received an error message but didn't receive a move after sending the "move now" command to an engine. Some Deuterium versions made this bug show up.
It's been a really long time since the last release, so we decided to delight our users with a new release of cutechess-cli 0.6.0. As usual, both source code and binary packages for supported platforms can be downloaded at https://github.com/cutechess/cutechess
Anyone using cutechess-cli 0.5.1 or earlier is encouraged to update. If you're using the version that was made for the Stockfish testing framework, there's probably no reason to update - unless you use the SPRT feature which was improved lately.
New features or changes:
* CLOP glue script is compatible with Python 3
* Opening positions can be picked from EPD files
* ECO codes and opening names as included in PGN output
* Custom XBoard engine options are supported. This should make Gaviota users happy.
* Added "trust" engine option that makes cutechess-cli trust any result claim from the engine.
* The correct PGN move number format is used in games where black moves first
* The precision of the relative score for engine A is increased
* Initialization timeout period is longer so that some slowly starting engines are detected properly.
* A Sequential Probability Ratio Test (SPRT) can be used as a termination criterion for the match.
* The format for setting some options has changed to a less error-prone format
* The "-draw" option now has a "movecount" parameter for setting the number of consecutive scores that have to be within the chosen treshold.
* The "-pgnin" and "-pgndepth" options have been replaced be the "-openings" option that supports both EPD and PGN formats, and two ways to pick openings - sequentially or randomly. In "random" mode the openings file is first indexed, so it can take a few seconds with large files (> 100 MB), but it's only done once. It's also possible to set the number of the first opening to be picked (only in sequential mode).
* Gaviota tablebases can be used to adjudicate games
Bugs fixed:
* Missing command or invalid protocol in engines.json results in a crash
* UCI combo options couldn't be set properly on some systems
* Memory leak when the "-pgnout" option is not used
* On some systems the random number generator produced the same sequence of numbers in each thread, which would result in the same opening used all the time in tournaments with more than two players.
* Threading problems. The multi-threading code is now simpler and more effective and some possible race conditions have been eliminated. If you previously saw warning messages about QThread objects being destroyed while the thread is running, cross-thread killTimer() warnings or random crashes at the start of a match, then this fix should make you very happy.
* Cutechess-cli used to stall if it received an error message but didn't receive a move after sending the "move now" command to an engine. Some Deuterium versions made this bug show up.