[ts-gen] segmentation faults on Debian

[Bill Pippin]

Russ, I don't find any problem with the shim on 32bit debian,
and given a successful compile as we've had for a week or so
now, I can't see why there would be architecture-specific problems
there; I didn't see any.

The machine is sufficiently underpowered that the Ruby script runs
way ahead of the shim at startup, so the stderr trace info from the
two processes is confusing or even misleading, but otherwise both
test.rb and risk.rb run fine.

I'm going to wait for you to reproduce this before I dig any
farther.

Thanks,

Bill