bobm@convexs.UUCP (05/03/87)
Is anybody out there running Minix on a machine with a non-Intel
processor yet? I like what I've read of Minix, but I refuse to
own a machine with segment registers. Call me prejudiced.
Along the same line, what's the cheapest 68k or NS32k box you
know of that could run Minix? It seems like an MMU is mandatory.
The Mac II will have an MMU (optional); anybody know of anything
cheaper? I'm not really very keen on building my own, even though
I would consider porting Minix.
"Computers are my forte." K<bob>
deep_thought() Bob Miller
{ Convex Computer Corporation
sleep(years_to_seconds(7500000)); Richardson, Texas
return 42;
} {ihnp4,cbosgd,allegra,sun}!convex!bobm
Disclaimer: The above is not just the opinion of the author; it is the opinion
of all sentient beings in this universe and all other known universes. The
author's employer may, however, not be sentient.
hwe@beta.UUCP (05/07/87)
In article <132200005@convexs>, bobm@convexs.UUCP writes: > > Is anybody out there running Minix on a machine with a non-Intel > processor yet? I like what I've read of Minix, but I refuse to > own a machine with segment registers. Call me prejudiced. > ... I would love to work on porting minix to both 68K and PDP-11. The problem is that when I look at my UniPress price list, and see $10,000 for the Amsterdam Compiler Kit... I can't afford $200 for a new disk drive! And No, I am not, myself, entitled to an educational discount. I believe that a port will require compiler source. No, I DO NOT want to modify the compiler. However, the compiler will be needed in at least three different environments. First, on a minix pc, as a cross compiler to the target.(e.g. a PDP-11). Second, on the target under the existing host system (U**X, or whatever) to test pieces. Finally, as a native compiler under minix on the target architecture. This last would be the distributed in binary form with the finished port. I do feel that minix ports should be done with one compiler base. While I have a version of PCC that generates 68K code, and I have both the Ritchie compiler and PCC for 11s, one of the largest lingering problems with U**X portability (other than forgetting that short and int are the same size :-) is the minor (but crippling) differences between all the Cs in use out there. I understand that the distribution of the Amsterdam Compiler Kit through UniPress was a really good thing at the time. It allowed some really nice stuff to get distributed. However, when that agreement was reached, I would guess that the situation with minix was not forseen. Is there a chance that some scheme could be worked out so that those of us wanting to port minix could have access to the compiler technology?? I certainly don't mind licensing that would protect UniPress's ability to sell the product to those who would buy it today. I don't mind licensing restrictions that would prevent me from letting anyone else have it. Certainly, distribution of any port would have to route back through Prentice Hall, and the compiler distributed with a port would be binary (as in the current system). Anyone out there (UniPress, Prentice Hall, Distinguished Author??) care to come up with a scheme that lets me build a compiler and cross compiler to support a minix port while, at the same time, protecting the interests of UniPress, and Prentice Hall?? Skip hwe@lanl.gov hwe@lanl.UUCP
merlin@hqda-ai.UUCP (David S. Hayes) (05/07/87)
In article <5129@beta.UUCP>, hwe@beta.UUCP (Skip Egdorf) writes: > I would love to work on porting minix to both 68K and PDP-11. > The problem is that when I look at my UniPress price list, and > see $10,000 for the Amsterdam Compiler Kit... > I believe that a port will require compiler source. There is no need to use the Amsterdam Compiler Kit. Richard Stallman and Len Tower, of Project GNU (recursive: Gnu's not Unix), have created the GNU C compiler. Briefly, project GNU aims to write a Unix replacement, much like Minix, with the source code freely available. GNU will be much more robust than Minix, since it is intended to be used, not studied. Some pieces of GNU are already available. These include BISON (unix yacc), GNU debugger, and GNU Emacs. This last is the best editor I have ever seen on any computer. The C compiler was announced about a month ago. It is currently considered to be in beta-test, but you can still get it. The Free Software Foundation (the administrative entity that does the distribution work of project GNU) has an archive of GNU sources maintained on the Internet at "prep.ai.mit.edu", address 128.52.14.14. This can be accessed via anonymous ftp. For those who don't have ftp access, you may write to Free Software Foundation 1000 Massachusetts Avenue Cambridge, MA 02138 Phone: 617-876-3296 GNU is not public domain. The license allows anyone to redistribute as many copies of the software as they want, subject to two basic restrictions: 1) You can't sell it for a profit. 2) You must supply it as you got it, i.e., source code. The quality of the code from GNU is outstanding, and everything is written to be ported to as many different machines as possible. GNU is also subject to frequent updates, enhancements, and bug fixes. Most of these are very minor, and can be safely ignored until the next major release, but it's good to know they're there. The updates are available in the same places the source is. For those interested in following the doings of FSF and Project GNU, have a look in "comp.emacs". They occassionally stuff a note or two in there. -- David S. Hayes, The Merlin of Avalon PhoneNet: (202) 694-6900 UUCP: *!seismo!sundc!hqda-ai!merlin ARPA: merlin%hqda-ai.uucp@brl.arpa
rpw3@amdcad.UUCP (05/08/87)
In article <5129@beta.UUCP> hwe@beta.UUCP (Skip Egdorf) writes: >I would love to work on porting minix to both 68K and PDP-11. >The problem is that when I look at my UniPress price list, and >see $10,000 for the Amsterdam Compiler Kit... You don't need that. At USENIX, they announced a new package specifically for MINIX hackers -- $99 for the source of the MINIX C compiler only. Note that this does NOT include other compilers than C, nor the compiler-compiler tools, but should let you fix bugs in the compiler. Whether it's enough to let you bootstrap a non-Intel compiler is questionable, though. Talk to somebody at UniPress directly. Rob Warnock Systems Architecture Consultant UUCP: {amdcad,fortune,sun,attmail}!redwood!rpw3 ATTmail: !rpw3 DDD: (415)572-2607 USPS: 627 26th Ave, San Mateo, CA 94403
ast@cs.vu.nl (Andy Tanenbaum) (05/09/87)
In article <5129@beta.UUCP> hwe@beta.UUCP (Skip Egdorf) writes:
[Discussion about MINIX on non Intel CPUs]
Early on in this group I mentioned this, but I will mention it again for
people who have joined recently. There is a port underway to the Atari
ST. It is making good progress, but as you all know, the first 99% is the
easy part. I will post more when things are getting closer to completion.
The problem of the compiler is well known to us and will have to be dealt
with. The person doing the 68000 port has access to all the compiler
sources and technology. Precisely how will deal with this remains to be
seen, but we are keenly aware that a C compiler is very important and will
do our best to find a reasonable solution.
As far as porting to CPUs other than the 8088 family and 68000 family,
the problem is harder. Licensing of the ACK source is one problem, but a
more serious one is its size. The current ACK distribution includes about
half a dozen front ends for various languages and a dozen back ends,
massive libraries, test suites and tons of stuff. You need a 32-bit CPU
and probably at least 30-40 megabytes of empty disk space to even begin thinking
about bringing it up. It works fine on VAXes and SUNs, but not on ATs.
This means that it isn't really aimed at individuals. Universities can get
the whole source for $995, so people at a university who want the full sources
can get it for a price most universities can afford. For people at companies
who are using the company's VAX or SUN to hack MINIX, we don't really have an
adequate solution unless the company needs ACK for its business and is willing
to pay the commercial price.
Andy Tanenbaum
hwe@beta.UUCP (Skip Egdorf) (05/09/87)
In article <353@hqda-ai.UUCP>, merlin@hqda-ai.UUCP (David S. Hayes) writes: > In article <5129@beta.UUCP>, hwe@beta.UUCP (Skip Egdorf) writes: > > I would love to work on porting minix to both 68K and PDP-11. > > The problem is that when I look at my UniPress price list, and > > see $10,000 for the Amsterdam Compiler Kit... > > I believe that a port will require compiler source. > > There is no need to use the Amsterdam Compiler Kit. Richard > Stallman and Len Tower, of Project GNU (recursive: Gnu's not > Unix), have created the GNU C compiler. > I think that the Gnu C would be a good base for minix in a year or so when it is not quite so beta testy. Currently, minix hacking is a young enough sport that I think it should be kept as simple as possible for the time being. This means not complicating things with a new compiler as well as new minix ports. Remember back about 1979-1980 when the first few U**X V7 ports were done?? The general comment when the dust settled on the new port was something like "The kernal was simple once we got around byte-swapping, but getting PCC working was a real job...". Having worked with PCC, I am sure that the Gnu compiler is a much better engineered product; we have learned a lot about building such compilers in the last 5 or 10 years. (Nothing WRONG with PCC, it was quite remarkable for 1978. However the first ANYTHING, let alone a Portable C Compiler, is seldom the best ever.) Gnu C is a good suggestion. I intend to watch its progress, play with it, and hopefully contribute to it. However, given a choice between playing with Gnu C or playing with a minix port, my personal choice will be to spend my scarce hobby hacking time on the OS. Thanks for a good suggestion! Skip hwe@lanl.gov hwe@lanl.uucp
stuart@bms-at.UUCP (05/12/87)
It works great for 32-bit cpus (like 68000). It needs work for 16-bit cpu's. I am working on a 16-bit version. -- Stuart D. Gathman <..!seismo!dgis!bms-at!stuart>
henry@utzoo.UUCP (Henry Spencer) (05/13/87)
> ... GNU will be > much more robust than Minix, since it is intended to be used, not > studied... Change that to "GNU is *intended* to be much more robust..." and I'll go along with it. Intention to use is no guarantee of robustness, as witness a whole lot of current software. GNU may well score higher on robustness, but asserting that this is automatically and inherently so is hubris, to put it mildly. Try moving any of the GNU stuff to a non-32-bit machine if you want to learn about robustness; my understanding is that it's likely to be, uh, a Really Interesting Experience. > ... GNU Emacs. This last is the > best editor I have ever seen on any computer. Also the biggest, and the slowest to start up, although it's okay if you have enough memory for it and just leave it running all the time. Full GNU will need truly enormous main memories if this pattern holds... -- "The average nutritional value Henry Spencer @ U of Toronto Zoology of promises is roughly zero." {allegra,ihnp4,decvax,pyramid}!utzoo!henry