Compiling Kaffe 1.0.5 on Alpha running OSF

[Godmar Back]

The F_SETOWN is needed on BSD and I think also Linux systems in order to
get asynchronous (SIGIO) notifications on sockets.  On some (most?) systems, 
it's only needed and legal for sockets.  I think that's why at some
point I decided to ignore a failure of the fcntl.

However, if it fails, it shouldn't affect the state of the filedescriptor 
or make it unusable, should it?

You could also verify that the jthread_write() method is being called.
Maybe the problem is there.
I think if you run with -vmdebug, it will leave fd 2 in standard mode,
so you should always see debugging output generated with write(2, ...).

	- Godmar

> 
> I have taken a closer look at the problem that we have observed
> in the multithread  java programs in test/regression (ttest.java).
> 
> The threading appears to be working on the Alpha.  When I "-vmdebug
> JTHREAD", I can see the processor switching between the threads.  I
> also wrote some simple test cases code that showed that the child
> threads were actually doing something.
> 
> The problem apprears to the the input/output code and no output is
> being generated.  When I turned on the "-vmdebug JTHREAD", I got a
> number of error messages:
> 
> F_SETOWN: Not a typewriter
> 
> The DEC unix does not like the way the some of the file handling is
> being performed for multithreaded programs.  This error message is
> coming from line 2020 of
> kaffe/kaffevm/systems/unix-jthreads/jthread.c:
> 
> #elif defined(F_SETOWN)
>         /* On some systems, this will flag an error if fd is not a socket */
>         r = fcntl(fd, F_SETOWN, pid);
>         if (r < 0) {
>                 DBG(JTHREAD, perror("F_SETOWN"); )
>         }
> #endif
> 
> I am not sure this is the root cause of why multithreaded programs
> such as ttest do not produce any output.  However, it is a suspect.
> Any suggestions?
>