getrusage bug?

[Stephen Weeks]

> The child times only include children that have exited, so your test is bad.

Agreed.

> This has nothing to do with the MLton compiler reporting 0 times for the
> assemble and link times.  That could be caused by the real work being done
> by grandchildren, not children.

I didn't think child vs grandchild mattered.

>  I don't see any solution to that.
> Where in the source is the place where the fork/exec of gcc gets done?

src/lib/mlton/basic/process.sml

> We MUST be waiting for the children that do the assembler stuff, otherwise we
> would be starting the link before the .o is created.

Yep.

The problem was in src/mlton/control/control.sml.  The time function was not
even looking at the childrens' times.   A simple fix.  Sorry for the confusion.