[MLton] Re: [MLton-devel] MLton HOL

[Joe Hurd]

On Sun, 19 Oct 2003, Stephen Weeks wrote:

> I'm trying to understand how to compare this with the 10 minute
> compile that you saw earlier.  Were these both with 20030716?  What
> were you doing to fix the Interrupt problem that let you get the 10
> minute compile?

The 10 minute compile was prior to realizing the Interrupt problem
existed, so I wasn't doing anything. When I ran the resulting program,
it immediately threw an exception and terminated. I had been assuming
that the mlton optimizer "knew" this would happen, and so got rid of
large chunks of the program, resulting in a quick compile. Is my
anthropomorphizing reasonable in this case?

> There is an expert option that allows you to turn off any of the SSA
> optimizations.  In this case you could do
>
> 	-drop-pass knownCase

OK. At the moment moving to the 1.5Gb machine seems to have solved
that problem, but I'll bear this in mind for the future.

> Yes, this is a bug.  We changed how equality is handled, and only
> propagated the change to the 2002 basis.  I've checked in a fix to our
> CVS.  If you want to get it, you can use the 20031018 binary package
> to process the latest CVS basis library sources and rebuild
> world.mlton by using the following command
>
> /usr/lib/mlton/mlton-compile <path>/<to>/basis-library/ /usr/lib/mlton/world
>
> Of course, you must have write permission on
> /usr/lib/mlton/world.mlton in order to do this.  Let us know if you
> need any other help to get this working.  I'll remake binary packages
> soon, but I'd like to try to solve some of your other problems before
> doing so.

OK, I'll give this a go. Would 20031018 suffer the same problems with
writing to /tmp as 20030716?

Joe


_______________________________________________
MLton mailing list
MLton@mlton.org
http://www.mlton.org/mailman/listinfo/mlton