Adventures in OpenBSD (and NetBSD) parts 1 through 4, MacBSD

My first sorties into OpenBSD were almost twenty years ago, trying to install OpenBSD and NetBSD on old Macintosh hardware. Ran into SCSI and floating point issues. And bad CPU mask set problems.

Really wish I had known Apple was offering to replace the bad mask set 68LC040 in the LC/Performa 630 series Macintoshes. Found out about it too late.

The bad mask set CPU in the Performa 630 had faulty MMU operations during exceptions in certain address ranges. It almost functioned reliably for Mac 7.0 and 7.1, but not so much for Mac OS 7.5 or 8.x. Loading OpenBSD or NetBSD, you'd get tantalizing close, and then it would panic. (Actually logged in once to a completely installed NetBSD, looked around a bit, then tried to conpile something, at which point it died. Later versions would not even get that far.)

Was somewhat tempted by offers of a full 68040 on ebay, for $100 plus or minus, but I needed the money for rent and food for the kids.

The problem with the Performa 550 was lack of hardware floating point. (Always regretted not splurging 6,000 yen for the 68882 when I had a chance to get it on a closeout at, I think it was Ninomiya in Sannomiya.) My understanding was that the SCSI problems there were partly due to lack of floating point. Mac OS 7 ran pretty well, though.

Still have both machines, although my wife wishes I'd get rid of them. The 550 seems to have developed whiskers in the floppy circuitry that cause phantom disk insertion interrupts. The 630 has something in the monitor circuitry that ruins focus and color. Maybe that's whiskers, too. (Early lead-free PC boards tended to whisker around the solder joints.)

I keep them for sentiment, I guess. And because I someday hope to have the time to implement figFORTH and my bif version of FORTH in 68K assembler.

I extracted the ROM image from the 630 recently, to run the BasiliskII emulator under OpenBSD and on Android. It looks nice, but a major part of the magic of the old Macintosh was the careful timing that made the mouse work smoothly. That timing does not work under emulation. It's bad enough to prevent me using either emulated environment to start working on my re-implementation of FORTH during commute time on the train.

No current OS, including Mac OS X, really gets the mouse timing right, by the way.

To really get the timing right, the GUI really should be controlled by a separate CPU.

Consider what could have been in 1992, had Apple simply started putting both the Mac OS and A/UX in a single box, each with its own CPU, instead of running off into fantasies with the first abortive re-design which was theoreticallly intended to become Mac OS 8 but ended up as a pipe dream.