IO problems?

[Stephen Weeks]

> This is the ancient pre-glibc version of libc.  Not anything we compile with
> or that you even CAN compile with under Red Hat.  It is only there for
> compatibility (so you can run old binaries).

Ahh.  So I am running

% rpm -q glibc
glibc-2.1.2-17

> What I was talking about was that at some point (I think long past) there was
> the issue of how mlton would know, for instance, that SIGKILL was 9, or what
> SIG_IGN was.  At the time you toyed with the idea of running a C program
> which would produce a .h file which the runtime system would include which
> would tell it these things.  I seem to recall that this turned out not to be
> needed because you could just have the runtime system use SIGKILL and include
> signal.h, but I remember that there was some problem with this.  I think that
> the problem was solved, but just wanted to be sure.

Now, mlton produces C code with the symbolic constant "SIGKILL".  This 
is something we'll have to figure out out to solve with the new backend.

> I must admit to pretty close to zero sensitivity to people who are running
> MLton under Windows or NT.

I am sure my sensitivity is higher than yours, but I mostly want to
make sure it isn't a bug in the basis library code.