Path: utzoo!attcan!uunet!munnari!otc!metro!ditsyda!evans
From: evans@ditsyda.oz (Bruce Evans)
Newsgroups: comp.os.minix
Subject: Re: V1.4 #0 (introduction)
Summary: Report on 1.4a
Message-ID: <1633@ditsyda.oz>
Date: 26 Jan 89 12:10:30 GMT
References: <1945@ast.cs.vu.nl>
Organization: CSIRO DIT Sydney, Australia
Lines: 167

I have been running the modified sh for several months and strongly
recommend it. The only new bug I know of is that sh -x does not print
assignment statements properly (try sh -x $HOME/.profile with both shells).

I found the following problems.

Packaging
---------

The diffs for commands/make/* were reversed.

No crc was posted for helpfile. Please use shar even for single files,
since even with perfect communications it's hard to tell how many newlines
are added by news.

Same for floppy.c.cdif. This is actually the full floppy.c with the diff
appended.

Packaging - string functions
----------------------------

Makefile.strin and StringTest.c are in the crc list but were not posted.
I recovered them from the previous posting.

Running make is supposed to produce string.h and memory.h for /usr/include.
This was not done. Instead there is string.h.proto in the include directory.
This needs to be run through sed to produce the other files. Sed doesn't
work with multiple commands separated by ';', so the makefile doesn't work
directly.

The makefile needs to be edited for cc to produce the tester program. It
still has .o files.

This is always a pain when porting Unix programs. Why not just modify cc
and (preferably not) friends so that

	.o file == packed assembler file
	.s file == human readable assembler file

I won't be doing it since I only use cc for testing.

Lib
---

fdopen.c appears to have 2 fresh bugs. It no longer seeks to the end
for the append case, and sets the IOMYBUF flag even for a null buffer.

termcap.c appears to have some of the bugs fixed by EFTH. I think it
is based on an old ST version.

utime.c doesn't compile. It has an undeclared argument.

The library order is still hard to get right. Now index calls strchr and
rindex calls strrchr. I had to move strchr and strrchr later in the
library to match. Same for time which is now called by utime.

I forgot the -LIB flag when recompiling this library. Why does asld
distinguish extern symbols in the library? I understand why .define is
needed for the library, but why is .globl good enough elsewhere?

The CFLAGS for the library are getting out of hand:

	-Di8088 -DCONST='' -DSIZET=int -DVOIDSTAR='char *' -LIB

The strings package really belongs in a directory by itself.

The helpfile for more is called more.help in more.c, but help.more in the
distribution.

I think lib should be reserved for source files and not cluttered up with
help files. It is already too large. On a hard disk I use /usr/sys/lib for
source so there's plenty of room in /usr/lib. It is not clear how to
organize it for floppies. Does anyone still care?

Commands
--------

pr.c clears the L_BUF pointers before it allocates them. The bug shows
up in commands like

	ls | pr -t -l1 -5

(same before this supposed fix). It looks like a botched diff - the
indentantion is wrong too. I'm posting my 1 line fix again rather than
mess around with this 5 line fix.

Old cc bugs
-----------

The string tester program showed 2 problems. One involving strerror() is
system-dependent and doesn't matter. The other is that memchr() is broken.
This is because cc is broken. Cc incorrectly thinks

	(unsigned char) *"\203" == 0xFF83.

(unsigned char) 0203 is OK. The program ccbug.c shows the bug and its
effect on memchr. I've been bitten by this before when "porting" to cc.

	cc -D/user/include -c somefile.c

gives an amusing error message. The -D is a mistyped -I.

#! /bin/sh
# Contents:  ccbug.c pr.c.cdif
# Wrapped by sys@besplex on Thu Jan 26 23:00:25 1989
PATH=/bin:/usr/bin:/usr/ucb ; export PATH
if test -f 'ccbug.c' -a "${1}" != "-c" ; then 
  echo shar: Will not clobber existing file \"'ccbug.c'\"
else
echo shar: Extracting \"'ccbug.c'\" \(619 characters\)
sed "s/^X//" >'ccbug.c' <<'END_OF_FILE'
X/* show bug (unsigned char) *"\203" == 0xFF83 */
X
Xchar one[10];
Xchar *scan = "\203";
Xint uc;
Xint ucharwanted;
X
Xmain()
X{
X	strcpy(one, "a\203b");
X	if (memchr(one, 0203, 3) != one+1) {
X		printf("something wrong with memchr:\n");
X		printf("memchr(one, 0203, 3) = %04x\n", memchr(one, 0203, 3));
X		printf("one + 1 = %04x\n", one + 1);
X	}
X	ucharwanted = 0203;
X	uc = (unsigned char) ucharwanted;
X	if ((unsigned char) *scan != uc) {
X		printf("no, something wrong with cc:\n");
X		printf("(unsigned char) *\"\\203\" = %04x\n",
X			(unsigned char) *"\203");
X		printf("(unsigned char) 0203 = %04x\n",
X			(unsigned char) 0203);
X	}
X}
END_OF_FILE
if test 619 -ne `wc -c <'ccbug.c'`; then
    echo shar: \"'ccbug.c'\" unpacked with wrong size!
fi
# end of 'ccbug.c'
fi
if test -f 'pr.c.cdif' -a "${1}" != "-c" ; then 
  echo shar: Will not clobber existing file \"'pr.c.cdif'\"
else
echo shar: Extracting \"'pr.c.cdif'\" \(269 characters\)
sed "s/^X//" >'pr.c.cdif' <<'END_OF_FILE'
X*** /user/sys/commands/pr.c	Thu Oct  6 02:08:47 1988
X--- pr.c	Thu Oct  6 02:15:49 1988
X***************
X*** 468,473 ****
X--- 468,474 ----
X  		fprintf(stderr,"malloc returned %d\n",ptr);
X  		exit(1);
X  	}
X+ 	memset(ptr, 0, (unsigned)size);
X  	return (char *) ptr;
X  }
X  
END_OF_FILE
if test 269 -ne `wc -c <'pr.c.cdif'`; then
    echo shar: \"'pr.c.cdif'\" unpacked with wrong size!
fi
# end of 'pr.c.cdif'
fi
echo shar: End of shell archive.
exit 0
D