Version 1.6.24b upgrade problems

Version 1.6.24b upgrade problems

Post by Robert R. Ha » Wed, 30 Dec 1992 07:21:33



> Please report bugs on the net.  After the dust settles, I will fix any remaining
> bugs and then make this the official version for a while.

> Happy Holidays.



Ok here is what I have encountered so far. I match all the CRC's
given in the file MANIFEST, thus I believe I have a good copy
of the minix/1.6.24b directory but I find the files:
ansi.sh and stdio.sh  to be short at lot of cdiff files as
after unpacking them my CRC's for the directories:
  src/lib/ansi and src/lib/stdio
still have my ver 1.5 CRC's.

Also I can't find the share file for the directories:
 src/lib/em  src/lib/float  src/lib/math

Beware the files syscall.ansi.sh and syscall.kr.sh did
not get compressed so don't try to uncompress them.
The file name syscall.ansi.sh exceeds the 14 character
limit and gets truncated to syscall.ansi.s


 
 
 

Version 1.6.24b upgrade problems

Post by Andy Tanenba » Thu, 31 Dec 1992 07:02:57



Quote:>I find the files:
>ansi.sh and stdio.sh  to be short at lot of cdiff files as

This is correct.  I wasn't quite sure what to do with the library.  I made
a version that worked with the K&R compiler, as I promised, but it is a bit
of a hybrid with the 1.5 library.  Basically, you use the original library
(libc.a) but just add the few files I supplied, since these are essential.

I didn't supply the math and float libraries since anyone who bought the ACK
ANSI C compiler already has the source code, and they don't work with the
K&R compiler, so in theory nobody should need them.

Quote:>Beware the files syscall.ansi.sh and syscall.kr.sh did
>not get compressed so don't try to uncompress them.

This is correct.  The 14-character limit hit me here.



 
 
 

Version 1.6.24b upgrade problems

Post by Gerald He » Thu, 07 Jan 1993 01:34:49


Hi Netters, Hi Mr. Tanenbaum,

considering the size of the diffs, it seems that 1.6.24b is again
a "nearly-complete" minix-source package, as 1.6.23.... am I right?
(I can hardly imagine that there were ~100k diffs to nroff :-)

Some days ago, I had a dream that once we'll be able to download
a minimal set of diffs to 1.5.10 plus a complete library-source-tree
of the newest minix version. The library source as it is in
1.5.10 is a mess, patching it makes it even worse.
Will my dream ever become reality?

gerald heim

 
 
 

Version 1.6.24b upgrade problems

Post by Andy Tanenba » Thu, 07 Jan 1993 22:04:52



Quote:>Some days ago, I had a dream that once we'll be able to download
>a minimal set of diffs to 1.5.10 plus a complete library-source-tree

The 1.6.24b distribution IS a minimal set of diffs to 1.5.10.  It is so
large because there have been so many changes.  It is true that the library
source is not provided in complete form, but the diffs are provided.


 
 
 

Version 1.6.24b upgrade problems

Post by Dave Burges » Fri, 08 Jan 1993 04:59:39



> This is also my dream, like for monthes I spent nights trying to patches to
> 1.6.?? without any success. I have found a patch kit to 1.5.10 but this is
> for 1.3-from update. Is there a kit from 1.5 to 1.5.10 somewhere ?

  If you bought 1.5 from Prentice Hall, you already have 1.5.10.  If you try
to patch from 1.5.10 to 1.5.10 using the 1.3 to 1.5.10 patch kit, you can
be pretty well assured that the resulting files will diverge from the current
CRC list.

Quote:> the ideea of a minimal set of diffs is good, better should be a way to get
> a full library source tree than patchs to it.

  Be sure to check the list in the README to be sure you are patching from/to
the correct versions.  The 1.6.23 patchkit should have been from 1.5, I think.
The 1.6.24b should have been from 1.6.23.

TSgt Dave Burgess

--
Mail System (MMDF)

 
 
 

Version 1.6.24b upgrade problems

Post by Gerald He » Fri, 08 Jan 1993 20:51:22


Yes, I may have been wrong. Sorry.
But I have found some Programms in there where the patches are larger than
the sources (e.g. commands/ic).

Nevertheless I have tried to get 1.6.24b up and running. And I failed.
The first problem was building the library. Tsort is still buggy.
(Well not exactly: I could compile a working tsort using gcc and linux :-> )
Then I managed to build a new kernel, mm, fs and init.
I was not able to get the new tools (forgot the names, substitutes for
build and menu) running. They compiled w/o a warning, but didn't work.

It seems to me, that the filesystem format has changed a bit, but I could
find no documentation on it.

All in all, this is such a story of frustration that i have decided to:

 - Stay with 1.5.10 since this patching needs too much time even using
   a 12 MHz-286. (Where are these *s still using XTs? :->> )

 - Use the available patches for 1.5.10: shared text, virtual consoles,
   shutdown/reboot, nice ....
   This will result in a nice, stable and usable system.

 - 1.5.10 is POSIX enough for me. For special cases I will use Linux.
   Linux has a almost complete C-library, almost everything compiles fine.
   1.5.10 uses the same filesystem format as Linux, so they are pretty
   compatible this way (except symlinks, but that won't matter).

 - I won't ever sell my copy of Minix, since I will never find a
   fool that will even pay half the price I paid. :->

gjh

BTW: Has anybody numbers of ppl actually using Minix? It seems the numbers
     are decreasing, considering all the "mama, please unsubscribe meeeee!"
     news in the last weeks...

 
 
 

Version 1.6.24b upgrade problems

Post by Andy Tanenba » Sat, 09 Jan 1993 07:42:11



Quote:>I was not able to get the new tools (forgot the names, substitutes for
>build and menu) running. They compiled w/o a warning, but didn't work.

Some additional information here might help us debug it.

Quote:>It seems to me, that the filesystem format has changed a bit, but I could
>find no documentation on it.

It changed radically, to support POSIX, but that was something like two years
ago.  Most programs that care were changed too.  I think readfs may be the
only one left that is not fixed.


 
 
 

Version 1.6.24b upgrade problems

Post by Will Ro » Sat, 09 Jan 1993 05:29:41


:
:  - Stay with 1.5.10 since this patching needs too much time even using
:    a 12 MHz-286. (Where are these *s still using XTs? :->> )

Right here - I bought an XT to a) run Lotus b) run Minix, and Minix
on an XT is an awful lot better than DOS.  It's actually a nice OS
now things like communication programs have fallen into place.  Still
no decent spreadsheet, but 64KB+64KB is too small an address space
anyway, same problem as CP/M.  Furthermore, my XT works, and has done
for four years now.  The new, shiny 50MHz 486 they bought me for work
roaches the DOS filesystem once a day or so, corrupts every file it
edits, and won't accept a tape back-up unit.  This after 3 trips back
to the shop and about a weeks scuffling and cursing by yours truly.
There's a m*to this story...

Will

 
 
 

Version 1.6.24b upgrade problems

Post by Gerald He » Sat, 09 Jan 1993 22:05:43


   >I was not able to get the new tools (forgot the names, substitutes for
   >build and menu) running. They compiled w/o a warning, but didn't work.
   Some additional information here might help us debug it.

OK. I'll try. I think the name of the program was "construct". It's
appearently used to build a bootimage consisting of kernel,fs, mm, init,
monitor and the bootblock. Somewhere in the makefile it's called as
 "construct -i image <parts>", but it doesn't build a image. Since
I had no manual available, I gave up and used shoelace, with which I was able
to boot the new kernel. Init, the kernel, fs and mm ran (I could see by
pressing F1), but init didn't fork any processes, nor was it wasting time,
nor complaining about anything. The idle-task accumulated cpu-ticks as usual.
A running system, but completely unusable :->
I don't know if this was due to my filesystem (1.5.10ish, whatever that means).

   It changed radically, to support POSIX, but that was something like two years
   ago.  Most programs that care were changed too.  I think readfs may be the
   only one left that is not fixed.

Hmm. Is something like 2 years before or after the release of 1.5.10?
If it changed since then, bootstrapping a new system would be some
kind of difficult. Would it then be possible to distribute a diskimage with
the new kernel and some basic programms in /bin, maybe even a
binary of the new KnRish library? But this doesn't matter to me, since
I will definitly stay with 1.5.10 :-)

My contribution to the C-Compiler-Flame-War:
Building a new library, I tried to use the 1.6.16 tsort. It failed utterly
sorting an input of ~20k output of lorder. It gave me 38 Bytes of output!!!
I downloaded tsort.c to my linux box and compiled it with gcc (I had to
comment out the _PROTOTYPES). Now it ran ok and gave me a order of the
complete library. Strange. But this might be a library bug.

gjh

aehm: if my address in the header is wrong:

 
 
 

Version 1.6.24b upgrade problems

Post by Bruce Eva » Mon, 11 Jan 1993 01:15:50



Quote:

>Yes, I may have been wrong. Sorry.
>But I have found some Programms in there where the patches are larger than
>the sources (e.g. commands/ic).

Many of the patches are large only because the source has been reformatted.

Quote:>Nevertheless I have tried to get 1.6.24b up and running. And I failed.
>The first problem was building the library. Tsort is still buggy.

Library building is easier using bcc.  It has tsort built into the linker.

Quote:>It seems to me, that the filesystem format has changed a bit, but I could
>find no documentation on it.

The old (V1) file system has not changed.  The new (V2) file system has
32-bit block numbers and inode access and change timestamps.  The number
of inodes is still limited to 32K (or maybe 64K) and file name lengths
are still limited to 14.

Quote:> - Use the available patches for 1.5.10: shared text, virtual consoles,
>   shutdown/reboot, nice ....
>   This will result in a nice, stable and usable system.

You would need to reinstall (port?) shared text and virtual consoles to
make 1.6 as nice.
--

 
 
 

Version 1.6.24b upgrade problems

Post by Andy Tanenba » Mon, 11 Jan 1993 21:32:48



>OK. I'll try. I think the name of the program was "construct". It's
>appearently used to build a bootimage consisting of kernel,fs, mm, init,
>monitor and the bootblock. Somewhere in the makefile it's called as
> "construct -i image <parts>", but it doesn't build a image. Since
>I had no manual available, I gave up and used shoelace, with which I was able
>to boot the new kernel. Init, the kernel, fs and mm ran (I could see by
>pressing F1), but init didn't fork any processes, nor was it wasting time,
>nor complaining about anything. The idle-task accumulated cpu-ticks as usual.
>A running system, but completely unusable :->

Here is a manual (written by the author of the new boot mechanism, Kees Bot).

Andy Tanenbaum

.SH
Doing more with the Minix Boot Monitor.
.PP
This text describes the menu interface of the Minix Boot Monitor, and
the commands that may be used to customize it.
.LP
First of all, the monitor mode as distributed normally hides some of the
functionality, but shows you an explanation of the environment
variables instead.  If you add -DEXTENDED_LIST to CFLAGS in the Makefile
and recompile, then you will no longer see the long explanation (you
should know it by now), but you will see all the commands the monitor
knows about.
.SH
The commands.
.PP
The boot command has two functions, one is to load and start Minix, the
other is to boot a different operating system.  If the first partition
on your hard disk contains MS-DOS, then
.DS
.B
boot hd1
.R
.DE
will boot MS-DOS.  (Not all operating systems like to be called this
way, some insist on being on the active partition.)
.LP
The delay, ls, and other simple commands are not too difficult to
understand, just try them out.  The trap command may be used to execute
a function after a delay.  You can show a menu first and boot Minix
after 5 seconds of inactivity like this:
.DS
.B
trap 5000 boot; menu
.R
.DE
(This must be typed on one line, traps are cancelled when the prompt is
printed.)
.SH
Functions.
.PP
Functions are used to bundle commands, or to build menu items.  The best
example of a simple function is 'main', the function executed by the
monitor on startup.  Main is by default defined as:
.DS
.B
main() { menu }
.R
.DE
So that's why you see a menu at the start.  The example with 'trap'
above could be executed by main if you type:
.DS
.B
main() { trap 5000 boot; menu }
save
.R
.DE
The save command will save the changed environment of the monitor to the
second half of the boot block, the "boot parameters sector".
.LP
Functions may have one or two arguments, the first is a key to be
pressed from the menu to execute the function, the optional second
argument is the text that is to be displayed on the menu.  The single
argument functions should only be produced by construct, like this
one:
.DS
.B
AT(a) {label=AT;image=42:626;echo AT kernel selected;menu}
.R
.DE
It invites you to choose one of many kernels on a special boot floppy.
.LP
The two argument functions are used to customize the menu, once you
define one the default option disappears, so your first function will
probably be one to start Minix.  Example:
.DS
.B
minix(=,Start Minix) { boot }
dos(d,Boot MS-DOS) { boot hd1 }
save
menu
.R
.DE
Now you can type '=' or 'd' to choose between Minix and DOS.

 
 
 

Version 1.6.24b upgrade problems

Post by Laurent Chav » Tue, 12 Jan 1993 06:01:11




>:
>:  - Stay with 1.5.10 since this patching needs too much time even using
>:    a 12 MHz-286. (Where are these *s still using XTs? :->> )

>Right here - I bought an XT to a) run Lotus b) run Minix, and Minix
>on an XT is an awful lot better than DOS.  It's actually a nice OS
>now things like communication programs have fallen into place.  Still
>no decent spreadsheet, but 64KB+64KB is too small an address space
>anyway, same problem as CP/M.  Furthermore, my XT works, and has done
>for four years now.  The new, shiny 50MHz 486 they bought me for work
>roaches the DOS filesystem once a day or so, corrupts every file it
>edits, and won't accept a tape back-up unit.  This after 3 trips back
>to the shop and about a weeks scuffling and cursing by yours truly.
>There's a m*to this story...

>Will


I also have on old XT, and up to now I am very satisfied with what minix
has provided me with.
I tried to install bsd on my 486, and it is anything but trouble. I do not
have time for installation, being a student my main concerned is in learning,
minix is such a tool. It is not intended for profit, or real world application.
It is a platform for learning.
 
 
 

Version 1.6.24b upgrade problems

Post by Gerald He » Tue, 12 Jan 1993 18:42:28


forgive me, forgive me, i didn't want to start another flame war,
it's completely ok, if you still use your XT for Minix (what else
could you run on an Xt, if you wished something stable, unix-alike)
since this was intended as a joke, i had put an evilly grinning smilie
in the bracket... ":->"

sorry

gjh

 
 
 

1. FATAL 1.6.24b (PC) Upgrade error, upgrade aborted!

Hello Netters,

I own PC-Minix version 1.5 from Prentice Hall and wanted to upgrade to
1.6.24 (beta).

Before I bought 1.5, I asked the net how to get the latest version of Minix.
The advice I got was buy 1.5 and upgrade from the net. To my dissapointment
this seems to be impossible!

I've got all the upgrade files for 1.6.24b but that's not enough. First of
all the upgrade kits on star.cs.vu.nl and plains.nodak.edu are both missing
syscall.ansi.sh and syscall.kr.sh. I reported this on the net with a
question where I can get them. No response. Although I have to use a clumsy
way to ftp I found them on the ftp-site in koeln (with the correct crc's).

The commands.sh file on on star.cs.vu.nl and plains.nodak.edu doesn't
contain the file whatis.sh, although it should. From what I read I understand
that it has been updated to fix some other problems (the date of the file
differs from the rest of the kit on star.cs.vu.nl), and the whatis.sh file
was lost. I got whatis.sh from the 1.6.23 upgrade on koeln, the crc matched.

Now that I've got everything I started to apply the patch. Using only a 10Mb
partition I did it part by part, starting with the commands, then the kernel,
the include files and finaly the library. I only whished I started with the
library, because that whould have saved me a lot of time.

The library is very incomplete. The whole ansi and stdio directory is
praktically empty, and what can you do with a C library without printf,
fread, fwrite etc?? The 1.5 distribution contains the files for ansi but these
files differ from the crcs in 1.6.24b. I don't want to use them, I wanted to
have a real 1.6.24b system! I started looking again on the FTP sites on
earlyer distributions but this time no luck. So now I'm stuck. I can only
build an unstable hybrid version of minix for which there is no future path,
which is not supported by Andy, and which may or may not work, Great.

There are several people beta-testing this version. How come they didn't miss
those files. How did they get them or did they have to buy the compiler to get
these?

It's fine with me, I know the patches are free and I shouldn't complain.
I only complain because I was told to upgrade from the net, and now that I
want to it appears it can't be done. Please don't tell the next one who
wants to start with the latest version of minix to upgrade from the net.

Kind Reagards,

Henk.

--
  /   /            Henk de Groot      | Dep.: IISS-SE (System Management)

/   / (-_ / / /(   Tel: +31 55 432104 | Digital Equipment Corporation

2. Elite distribution

3. Version 1.6.24b libc.a problems

4. LADYBUG

5. PH1.5 -> 1.6.24b upgrade probs

6. Desperately seeking game...

7. One more problem on 1.6.24b with K&R-C compiler (posix/_dup.c)

8. Abnormail Behavior

9. 1.6.24b with K&R-C (2) - commands

10. Request for help with 1.6.24b

11. 1.6.24b notes

12. lib/other/mtab.c (1.6.24b)

13. 1.6.24b update 2 [long]