Some buffering problem...

[awagner]

My new engine, written in ruby, consistently blows up after the second move when running under xboard (Mac OSX). It's fine if i run the basic ui without xboard. The relevant section of the debug file is here:

15496 >first : usermove 15496 >first : c2c3
15497 <first : /Users/andrewwagner/workspace/ruby/tdchess/lib/tdchess/runner.rb:13:in `gets': Interrupt
15497 <first : from /Users/andrewwagner/workspace/ruby/tdchess/lib/tdchess/runner.rb:13:in `tick'
15497 <first : from bin/winboard.rb:15:in `<main>'
GameEnds(26, Error: first chess program (bin/winboard.rb) exited unexpectedly, 2)
Interrupting first
18069 >first : result 0-1 {Error: first chess program (bin/winboard.rb) exited unexpectedly}
18069 >first : force
18069 >first : quit

You can find the code at https://github.com/arwagner/tdchess if you're interested.

Any suggestions?

[awagner]

Here's a simple test case. This works if I just run it by itself, but has that interrupt if I run it under xboard:

Code: Select all

i = 0
moves = ["a7a6","a6a5","a5a4","a4a3"]
$stdin.each do |command|
  if command.start_with?("protover")
    $stdout.puts "feature ping=0 setboard=1 ics=1 usermove=1"
  elsif command.start_with?("usermove")
    $stdout.puts "move #{moves[i]}"
    i+=1
  end
  $stdout.flush
end


[awagner]

Aha, it's fixed. It was the SIGINT described in http://www.gnu.org/software/xboard/engine-intf.html#7 . I'm not entirely sure I understand why I would it would default to throwing those signals at me all the time, but I just turned them off in the initial feature command, and it's fixed.

[H.G.Muller]

Ah yes, I also keep getting fooled by that again and again, each time I port one of my engines to Linux. (On Windows signals do not work.) It is indeed a very silly feature of WB protocol, only good for switching off.

But it seems that the original GNU Chess, for which XBoard was designed as front end, needed the user to type ctrl-C to wake it up out of ponder search.