[MLton] MinGW hosted MLton

[Wesley W. Terpstra]

On Mon, May 02, 2005 at 12:21:30PM -0700, Stephen Weeks wrote:
> > 0. 'wait' is reported unused by MLton, but if removed it fails to
> > build
> 
> That it is unused makes sense, because, you removed the only use when
> you changed the definition of OS.Process.system.  It fails to build
> because the signature constraint OS_PROCESS_EXTRA requires wait.  This
> is clearly no longer needed, so I took wait out of the signature, and
> things compile fine.

I thought OS_PROCESS_EXTRA was exported somewhere, that's why I left it.
Removing it sounds good to me!

> * In the definition of OS.Process.system, why do you think it's OK to
>   remove the signal handler stuff that causes SIGINT and SIGQUIT to be
>   ignored?  In all honesty, I don't know why it was there, except that
>   it was probably inherited from SML/NJ.  But not knowing makes me
>   hesitant to remove it.

I switched to using system() b/c there is no /bin/sh on windows. :-)

If you read the manpage for system() it says that SIGCHLD will be blocked
and SIGINT and SIGQUIT will be ignored. I presume that SML/NJ was trying to
fake a proper system call. However, since we are using a real system call
now, I don't think that it is necessary to do this ourselves...?

Under windows, of course, this is all irrelevant as there are no signals.
(That's why I don't/can't use protect in the waitpid->cwait method...)

> * In main.fun, you changed the default MinGW temp directory from
>   "C:/WINNT/TEMP" to "C:/WINDOWS/TEMP".  This seems like a better
>   choice because it will work on more machines, although I wonder if
>   there isn't a more portable (across versions of Windows) solution.

The environment variable TEMP is used and the above is a fallback.

So, yes, there is a more portable thing to do: use TEMP. :-)
I just thought WINNT was a bad fallback, so changed it.

-- 
Wesley W. Terpstra <wesley@terpstra.ca>