[MLton-devel] Re: size increase

[Stephen Weeks]

> I would STRONGLY argue that any implementation of something called system
> MUST ignore SIGINT and SIGQUIT.

OK.  I must have misinterpreted some earlier mail of yours.

> 2.) It is trivial in things like the MLton compiler to not use system() and to
> 	instead use special code which understands that the compiler passes
> 	are not interactive.
> I would argue that 2 should be done any way since in the case of MLton, if I
> hit an interrupt I expect things to die, not stay alive.

It is not trivial to get rid of the *reference* (not use) to system.
The problem is that the MLton sources refer to OS.Process.system,
which refers to MLton.Signal.setHandler -- just the reference of this
in a program is enough to bring in the thread stuff.  It's too late to
get it anytime after CoreML dead code elimination.  The SSA dead code
elimination is not smart enough to undo the introduction of threads.

I don't like being on the edge of this cliff, and will try to think
some more about avoiding bringing in all the signal stuff unless it is
actually used.

_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel