Computing Scienceipnx Edition 8 Release 1.0
ipnx
Menu

Manual · Section 8 of 8

Troubleshooting

When the machine will not stop, and what a clean halt is for.

The rule that matters

Never force-quit ipnx. The emulated machine has a filesystem with data in memory that has not reached the disk, exactly as a real Unix does, and killing the app is pulling the plug on a VAX.

Quitting normally is safe and is not a shutdown — it takes a snapshot of the running machine and resumes it next time. That is the right behaviour for closing a lid, and it is what the app does when you quit or when the system suspends it.

Shutting down properly

If you want the machine genuinely stopped — before backing up its disk, say — do it from inside, as an operator would have:

cd /
sync; sync
/etc/halt
syncing disks... done
halting (in tight loop); hit
	^P
	HALT
The kernel flushes, waits for the I/O to drain, and then spins. That is a clean stop.

Two syncs is not superstition, it is the habit of a generation: the first schedules the writes, the second gives them time to happen. /etc/halt then does its own flush before stopping. Once you see the halt message, quitting the app is safe.

If it was force-quit anyway

The next launch will notice and run a filesystem check automatically. Usually it repairs itself and boots normally.

Usually is carrying weight in that sentence. fsck restoresconsistency, not data: it makes the filesystem mountable again, and several of its repairs work by throwing something away — truncating a file to a length it can prove, clearing an inode it cannot reconcile, moving an orphaned directory intolost+found under a number. A clean pass is not evidence that your work survived.

If the boot stops at a single-user prompt instead of a login, the automatic check found something it could not decide alone. The most reliable way forward is Reset (below); the disk can also be exported first and examined on a desktop SIMH if what is on it matters.

A terminal that has gone quiet

Press Return. This is nearly always it. A resumed session has nothing left to say — getty printed its prompt once, before the snapshot, and a logged-in shell prints nothing unasked — so a perfectly healthy restored terminal can look frozen. The app sends a nudge for you; Return is the manual version.

If Return does nothing on the 5620 specifically, use Restart terminal. That power-cycles the terminal only, leaving the VAX and your session untouched. It is the fix when a mux session was snapshotted and resumed — the window system lived in the terminal's memory, and the terminal always power-cycles.

Everything is slow

Check whether a large build or a find is running inside the machine — a VAX-11/780 was about one MIPS and it is being simulated honestly. If the machine is idle and the app is still busy, the 5620 window is the likely cost: it does not idle, because the real terminal did not. Close it.

Starting again

Settings → Disk → Reset replaces the emulated disk with the one the app shipped and re-creates your account on first boot. Everything inside the emulated machine is lost; everything on your Mac, including shared folders, is untouched. Export the disk first if there is anything on it you want.

Still stuck

Support has what to include in a report. The one thing worth adding here: if the machine printed anything unusual, the exact text is far more useful than a description of it — this is a system whose error messages are precise and occasionally beautiful.