[MLton] Finished (?) MLton.Child

Stephen Weeks MLton@mlton.org
Sun, 28 Nov 2004 10:30:05 -0800


> > Sorry about that.  Another reversed condition.  I had forkIsEnabled
> > set to true when *not* using mmap (of course it should be enabled when
> > using mmap). 
> 
> Is this also responsible for why mlton fails to use system()?

Yes.  Have a look at system/process.sml, which shows that system is
implemented using MLtonProcess.spawn, whose behavior mlton/process.sml
shows depends on forkIsEnabled.

> > If you change the basis library, all you need to do is 'make world',
> > which takes seconds.  If you change the runtime, all you need to do is
> > 'make world', which might take a minute.
> 
> That doesn't seem to work for me. When I install the newer basis library and
> try to use the 20041109 compiler, I get a big whack of errors. I didn't
> investigate deeper because I assumed that this was not supported.

It is often possible.  But in the CVS we often make language or
primitive changes that make a previous mlton-compile binary obsolete.
Unfortunately, that happened almost immediately with 20041109.  If you
had kept your working version from a couple of days ago, you should
have been able to use that.  I certainly debugged this problem without
any self compiles.  The only advice I have is that since rebuilding is
so slow for you, it might be worth keeping around a couple of recent
working versions while testing a new one.  A bit late, I know :-).