bug NOT dead

[Stephen Weeks]

> Damn, damn, damn.  I checked out the latest MLton with the `fix' and although
> it solves the problem for the reduced version of the  failure,  the  original
> code, compiled with
> 
>     -g -native false -keep g -keep ssa -keep dot
>             -DGC_EVERY_CHECK -DINSTRUMENTED
> seg-faults.  No assertion failure, just a simple seg-fault.

There were three versions of the code that we looked at

1) rico2.sml, which contained hash sets in addition to readFile
2) die.sml, which just contained readFile
3) a two line file that just allocated a large array.

I don't know if by reduced you mean (2) or (3).
I don't know if by original you mean (1) or (2).

Is the seg-fault the same as the error we were seeing before?  Is it
happening in about the same place, just after creating the 0.25G
array?