Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!samsung!caesar.cs.montana.edu!ogicse!plains!overby
From: overby@plains.UUCP (Glen Overby)
Newsgroups: comp.os.minix
Subject: Re: 1.5.0
Summary: It Compiles!
Message-ID: <3044@plains.UUCP>
Date: 27 Dec 89 00:52:48 GMT
References: <4945@ast.cs.vu.nl>
Reply-To: overby@plains.UUCP (Glen Overby)
Organization: North Dakota State UNIXversity, Fargo, ND USA
Lines: 73

I ran into a few snags while applying the diffs for 1.5.0 to 1.3c.  Here's a
list.  There were relatively few problems, considering that the shar files
were about 3 megabytes.  In fact, the shar files were only a few hundred KB
smaller than the resulting source files (but without the origionals, the
patches are useless!). I think there were four files in the entire system
that were not touched (tests 15, 17, 18 and 19).

Problem: fs/proto.h is > 0 and fs/putc.c doesn't exist

Symptom: the empty file, proto.h, doesn't unshar right, so fs/putc.c is in
	proto.h.

Solution: move proto.h to putc.c and delete the first few lines that have
	shell commands in them.  Then 'touch proto.h' (it should be an empty file).

Problem: test/t11a.cdif and t11b.cdif don't exist

Symptom: there are binaries in the test_00 shar file, causing it to not
	unshar correctly.

Solution: Hand edit the shar file.  The cdifs ARE in there!  Jove cut the
	file off prematurely, as did vi.  I used Gnu Emacs since it will edit
	a binary without second thoughts.

I did have a few other files that did not patch up correctly (22 files
required some attention overall), but I suspect it was caused by not exactly
matching 1.3 CRCs (I keep the sources in RCS files, which could have caused
some of the problems).

I now have a complete set of sources which match Andy's CRCs, except in
tools/bootblok.s (it's REALLY close -- the character count is only off by a
couple).

The library, kernel, fs, mm and everything in tools compiled rather
efortlessly.  Don't forget to set the makefile variables to where you hide
your cpp, lib, etc., and add a -T/usr/tmp (or whatever) to CFLAGS if your
ram disk doesn't have much free space.

Building a library will be a trick, since the 1.3 lorder and tsort don't
really work right.  Maybe I'll post the ordering I'm using later, after I've
sucessfully compiled all the commands with it.

The test cases looked like a blood bath:

tests  0, 1, 2, 4, 6, 7, 11, 12, 13, 14 : exited with "ok"
test 3: no errors, but no ok, either.
test 5: did nothing (it is supposed to fork. I could not see where that was
	ever done)

test 9: printed "Hangup" and exited.

test 10: says "ok", but I'm not sure that it was... (don't remember now what
	I thought was wrong).

test 15: strncpy fails tests 6, 9, 13

test 17: too many errors spit out to list

test 18: gives 17 errors

test 19: goes into an infinite loop, and does not appear to fork like it
	should.

test 20: POSIX calls fail miserably, leaving you with some funny files that
	are linked together and such.  Even when test16 passes, test20 fails.

Also, unlink will NOT unlink a file unless you can write the file -- being
able to write the directory is not good enough.

Now... to fix these problems.
-- 
		Glen Overby	<overby@plains.nodak.edu>
	uunet!plains!overby (UUCP)  ncoverby@ndsuvax, overby@plains (Bitnet)