> Is it worth (or perhaps already done) putting in a call to MLton.GC.pack > just before the fork that runs gcc? Already done. But this would have no effect with fixed-heap, which was part of the problem.