Eschecs FreeBASIC (UCI chess GUI)

[Roland Chastain]

Version 1.2.3. Chessboard coordinates. Creation of a INI file for settings.



Download (source code only)

[fxm]

Fingers crossed that this does not crash anymore !

Yes, let's hope. :)

Is there any disk file opening with reading or writing during the main loop (main thread) but executed from the other thread ?
'LoadFileAsString()' ?

[Roland Chastain]

Is there any disk file opening with reading or writing during the main loop (main thread) but executed from the other thread ?
'LoadFileAsString()' ?

That function is called at application start, to load JSON files. It is called from the main thread. In the latest version (1.2.3), there is an INI file, also read from the main thread.

[fxm]

I am looking for any input or output operation that would be executed from the other thread during the main loop of the main thread.

[Roland Chastain]

The procedure OnOutput writes to eschecs.log. And also the dependent procedures: OnUciOk, OnBestMove...

[fxm]

Thus, it would appear that the mutex locking block in 'ProcedureThread()' could be reduced to:

Code: Select all

sub ProcedureThread(byval param as any ptr)
  DebugLn("->")
  dim engineOutput as string
  with *cast(TListener ptr, param)
    do
      if .doReadOutput then
        engineOutput = ReadEngineOutput
      else
        engineOutput = ""
      end if
      if engineOutput <> "" then
        mutexlock(.sync)
        .procedure(engineOutput)
        mutexunlock(.sync)
      end if
      sleep(50, 1)
    loop until .quit
  end with
  DebugLn("<-")
end sub

as previously.

[Roland Chastain]

OK, thank you fxm.

By the way, maybe we could find better names. The names ProcedureThread and listener.procedure create a confusion.

[fxm]

Why not 'listener.callback' ?

[Roland Chastain]

Why not 'listener.callback' ?

Yes. Here is the final (?) version, with other minor modifications.

Code: Select all

sub ThreadProcedure(byval aParam as any ptr)
  DebugLn("->")
  dim s as string
  with *cast(TListener ptr, aParam)
    do
      if .doRead then
        s = ReadEngineOutput
        if s <> "" then
          mutexlock(.sync)
          .callback(s)
          mutexunlock(.sync)
          s = ""
        end if
      end if
      sleep(50, 1)
    loop until .quit
  end with
  DebugLn("<-")
end sub

[fxm]

Why here is a conflict (inducing crash without mutex) between 'INKEY' executed from a thread and 'PRINT #' executed from another thread ?

• The conflict can not be between 'INKEY' and 'PRINT #' (to disk) because this impacts two different devices.
• Therefore I wondered for a long time why so there was conflict between 'INKEY' and 'PRINT #' (to console), because they are using two different streams.
• But, by re-reading the 'Open Cons' documentation page, I think maybe I have a clue, thanks to the following warning paragraph:

  Warning: Presently, Open Cons does not work as described above. Without any file mode specifier, a runtime error 1 (illegal function call) occurs. With the Input file mode specifier, the only input mode is well supported. But with the Output file mode specifier, input and output modes are simultaneously supported.

[edit]
But I can not seem to highlight the crash with simple code (Windows 10) !

[Roland Chastain]

Thank you fxm for your investigations. Yes, it would be good to understand exactly the problem. Unfortunately I can't help you help much, because all that is still not very clear for me. I intend to retouch soon a project which has more or less the same configuration. I hope it will be the occasion to see clear in that affair.

I uploaded the latest version of my code, with a minor correction in the Windows version (about pointer size when compiling to 64-bit). Includes executables for Linux and Windows. Tests and comments still welcome.

[link deleted]

P.-S. I still would like to understand why the application uses much processor, even when it does apparently nothing. I tried to increase the sleep value in sGUI xSleep: it makes the reaction to mouse move slower, but the processor usage stays the same.

[Roland Chastain]

Hello everybody!

I applied the modifications suggested by neil here (thanks to him), and uploaded the project to a git repository:

https://codeberg.org/rchastain/eschecs-freebasic

Compilation successfully tested with fbc 1.10.1 under Linux.