I would like to show you what I created with this compiler, but unfortunately this is not possible.

Thanks again for your great support!!!

A solution that might seem ugly now is to change your calling function to be:

      subroutine GRAF()
.
      use appgraphics
      implicit none
.
.
      integer:: ipixxx, ipixyy, winx, winy, iiwin
      character::(20) mnubuf

      interface
            function noclosewin()
            integer::noclosewin
            end function noclosewin
      end interface

.
.
.
      iiwin = initwindow(ipixxx,ipixyy,mnubuf,winx,winy,dbflag=.false.,closeflag=.false.)
.
.
      call setwindowsystemclosecallback(noclosewin)
.
.
      end subroutine GRAF

The above should now make it clear to the compiler that noclosewin, wherever it is in your code, is a function returning an integer.

But I don't think the function is the issue.
The main source holding the call to the new subroutine gives the error!

The following lines are amongst others in my mainsource:

      subroutine GRAF()
.
      use appgraphics
      implicit none
.
.
      integer:: noclosewin, ipixxx, ipixyy, winx, winy, iiwin
      character::(20) mnubuf
.
.
.
      iiwin = initwindow(ipixxx,ipixyy,mnubuf,winx,winy,dbflag=.false.,closeflag=.false.)
.
.
      call setwindowsystemclosecallback(noclosewin)
(here the editor presents the error-message!!!)
.
.
      end subroutine GRAF

    function noclosewin ()
    implicit none
    integer::noclosewin 
    
        noclosewin = 0
        Print *, "I blocked it"
        
    end function noclosewin 

I tried the above on my system with the latest build, and it seemed to work.

              call setwindowsystemclosecallback( noclosewin )

By removing the parentheses, you're passing the function itself, not the result of calling the function, which is what your original code was doing.

Does that make sense?

I defined noclosewin as "integer"
Am I doing something wrong?

I see what you're asking now.  You want to disable the Windows "close" button?  If that ever did work in AppGraphics, it was a bug.  The closeflag only means that the program won't terminate, not that the window won't close.  If you click the close button, the window and its resources will be destroyed, but the program will keep running.

Overriding Windows' WM_CLOSE is going to be messy.  I would have to add a new API to block it.  The better way to handle this would be to assign a callback using setwindowclosecallback that will trigger safe shutdown in the main thread.  I'll see about cancelling the WM_CLOSE message, though.  It will take a bit of time to add.

It probably shouldn't outright crash, though it might be hard to modify the library to avoid that.  However, the fix would simply cause kbhit to return .FALSE., leaving this example in an infinite loop.

Basically, if all the AppGraphics windows are closed, it is extremely dangerous to continue making calls to the library.

      Program test
!
      use appgraphics
!
      implicit none
!
      INTEGER ipixxx, ipixyy, maxpx, maxpy, winx, winy, iiwin, ic
      INTEGER*2 port, imode, ipixelx, ipixely
      CHARACTER*6 mnubuf
      logical keypressed
!
      maxpy = getmaxheight()
      maxpx = getmaxwidth()
      port = 38
      imode = 180
      if (maxpx .gt. 1500) then
         ipixely = ((int2(maxpy)-int2(150)) / port) * port
         ipixelx = ((int2(maxpx)-int2(50)) / imode) * imode
      else
         ipixely = ((int2(maxpy)-int2(30)) / port) * port
         ipixelx = ((int2(maxpx)-int2(5)) / imode) * imode
      end if
!
      winx = (maxpx - ipixelx) / 2
      winy = (maxpy - ipixely) / 3
      ipixxx = ipixelx
      ipixyy = ipixely
!     
      mnubuf = 'Dynrel'
      iiwin = initwindow(ipixxx,ipixyy,mnubuf,winx,winy,dbflag=.false.,
     +        closeflag=.FALSE.)
!
450   Keypressed = .FALSE.
      Keypressed = kbhit()
      IF (Keypressed) then
         Ic = getch()
      end if
      IF (Ic .LE. 0) GOTO 450
!
      stop
!
      END program test

I can't replicate this behavior.  Do you have a simple code sample that causes this crash?