To exercise the parallel code, I changed sflag to default to 0, and
recompiled as a new file pmk.dis, but this also dispatches and
completes all jobs in the same order, and waits till one finishes
before starting the next one.

To test this I gave mk 10 shell scripts (0 to 9) to run at once, each
of which sleep for 10-n seconds (so script 0 sleeps for 10 seconds).
If they were executed in parallel, I would expect to all 10 started
before 0 finishes, and finish in roughly reverse order.

Is this a known issue with mk in acme-sac, or an unavoid(able|ed)
problem with the Windows port?  I don't see that the latest code
(which presumably runs in OSX and Linux too) of mk.b is any different
as regards sflag, and there's no issue apparently about this in the
tracker.

This maybe because the VM doesn't support multiple cores so setting
NPROC > 1 and getting true concurrency out of inferno mk isn't really
an option.

This question might also be asked on the inferno-os list since there is really
no difference between mk in acme-sac and inferno-os.

Caerwyn