Clean build of kaffe-snap on Solaris?

Godmar Back gback at cs.utah.edu
Mon Jul 27 09:43:08 PDT 1998


 John,

could you elaborate on your problems some?

> 
> Has anyone managed to build the current kaffe-snap on Solaris? I am using
> Solaris 2.4 and have patched the FileInputStream.c codeas suggested in a 
> previous post but with no luck.
> 

 That was a build problem?  I thought that was run-time problem where
IOExceptions were thrown.  A coherent patch for that one still hasn't
been posted, to my knowledge.

> Initially I got a FIOASYNC problem as Solaris reports FIOASYNC on an
> inappropriate device as ENOTTY not ENXIO. When I patched this some of

Where?  Are you referring to these lines in jthread.c?

        /*
         * On some systems, this will not work if a terminal fd is redirected.
         * (Solaris sets errno to ENXIO in this case.)
         */
        r = ioctl(fd, FIOASYNC, &on);   
        if (r < 0 && errno != ENXIO) {  
                /* Defines ENXIO to be an acceptable error */
                perror("FIOASYNC");
                return (r);
        }

Could you be more specific as to what fd is passed to jthreadedFileDescriptor:

    a) a stdio fd (0, 1, 2) during initialization
    b) a stdio fd that is redirected to a file during initialization
    c) a file fd that is created later, say during class loading
    d) a socket fd.

I could swear that I saw an ENXIO, but I don't have the time to check that
right now, and it takes half an hour to build Kaffe on the slowlaris
machines we have here.

> the tests work but the awt/WidgetsDemo and a number of others die with
> an Illegal instruction error. A backtrace under GDB lists callMethodV as
> the last named routine with a number of unnamed routines (presumably JIT)
> following.
> 

 You might want to try using the interpreter and --with-staticlib, that
should help in debugging.

	- Godmar



More information about the kaffe mailing list