charles@hp-sdd.hp.com (Charles Keith) (05/31/89)
Friday night I managed to solve the problems I was experiencing installing Bruce Evans' Protected Mode version of MINIX. The reward was the banner line: Memory size = 3584K MINIX = 127K RAM disk = 0K Available = 3457K Hard disk root is GREAT. A little slower, but no more mounting the root disk so I can change /etc/ttys or some such, and no more forgetting to put something on the 'permanent' root. And NO more waiting for MAKE to BREAK me. Good job and THANKS BRUCE! In response to my last posting, Bruce sent a note suggesting that I might be using the old make: >> The makefile has problems like a leading space in front of the defines >> up to line 75, and a missing ) on line 123 (target clean:). If I clean >> these up and make the makefile like I think I want it, make hangs without > >This is probably caused by using the old (1.2 single source file) version >of make. It doesn't expand nested macros properly. This can be done by >hand as a last resort. It may also have some space problems, causing the >hang. I used the 1.3 make, with chmem=20000 and LZ (line size) increased >from 1024 to 8192. > I was using 1.3 make, but make did have some problems. For starters, I needed to compile it with -DMINIXPC to get the .c.s rule turned on. I knew that problem was there, but this time I chased it down. My make/makefile had matched the v1.4a crcs, lulling me into a false sense of security. The second thing I did was compile make with -i (separate I&D option). This improved make's ability to withstand the onslaught of macro definitions in axmakefile. >> workaround. I have been having problems with kernel/axmakefile however. >> >> From crc-dist: >> 29962 8507 kerneldif/axmakefile.cdi >> and the file I unshared: >> 29962 8507 axmakefile.cdi >> patch produces: >> 57119 7773 axmakefile >> while crc-after says I should have: >> 42574 7662 kernel/axmakefile > >Crc-after is wrong. It must be out of date. The values produced by patch >are correct. > I think axmakefile did settle a bit in shipment, but the problems were just the cosmetic ones mentioned earlier. Another problem I had was that build kept telling me that kernel and mm click_shift didn't match. This was because although I had dutifully fixed libc.a, mm and fs grab head.s directly from /usr/lib, and were therefore grabbing the old one. Bruce, checking for this in build probably saved my skin. As for the future of official MINIX, I am somewhat ambivalent. It seems like the 386/680x0/RISC future belongs to GNU or some other cheaply licensed version of UNIX. Andy holds the power over official MINIX, and major enhancements like Bruce's protected mode are left with no legal means of distribution except the diffs sent over the net (not to be installed by the casual user or the faint of heart). This is sad. Because many and diverse people are working on it, the natural tendency of MINIX is to diversify. It would be nice if, after ordering MINIX 5 years from now, one would receive 57 diskettes containing 10 versions of the operating system for '86 '286 '386 '486 and maybe a few other processors, network versions, server versions, instrumented versions, and so on. From a teaching standpoint, this would be a feast, and far better than a standard shrink wrapped product. >Please post this. I don't think posting from here will work. > >Bruce Sure thing, Bruce. --- Charles Keith TELE: (619) 592-4472 Hewlett-Packard Co. UUCP: {hplabs|ucsd|hpfcla}!hp-sdd!charles 16399 W. Bernardo Dr. San Diego, CA 92127 Disclaimer: I am absolutely irresponsible. ---