Fix handling of files without newlines in the head and tail builtins (#365)

The head and tail builtins don't correctly handle files that lack
newlines[*]:
    $ print -n foo > /tmp/bar
    $ /opt/ast/bin/head -1 /tmp/bar  # No output
    $ print -n 'foo\nbar' > /tmp/bar
    $ /opt/ast/bin/tail -1 /tmp/bar
    foo
    bar$
This commit backports the required changes from ksh93v- to handle files
without a newline in the head and tail builtins. (Also note that the
required fix to sfmove was already backported in commit 1bd06207.)

src/lib/libcmd/{head,tail}.c:
- Backport the relevant ksh93v- code for handling files
  without newlines.

src/cmd/ksh93/tests/builtins.sh:
- Add a few regression tests for using 'head -1' and 'tail -1' on a file
  missing and ending newline.

[*]: https://www.illumos.org/issues/4149

NEWS

@@ -15,6 +15,9 @@ Any uppercase BUG_* names are modernish shell bug IDs.
   return value that fits in a signed integer, typically a 32-bit value.
   Note that $? is truncated to 8 bits when the current (sub)shell exits.
 
+- The head and tail builtins now correctly handle files that do not have an
+  ending newline. (Note that the tail builtin is not compiled in by default.)
+
 2021-12-05:
 
 - Fixed an issue on illumos that caused some parameters in the getconf

src/cmd/ksh93/tests/builtins.sh

@@ -1396,5 +1396,22 @@ got=$?; exp=1
 got=$?; exp=2
 (( got == exp )) || err_exit "cd -eP to empty string has wrong exit status (expected $exp, got $got)"
 
+# ======
+# The head and tail builtins should work on files without newlines
+if builtin head 2> /dev/null; then
+	print -n nonewline > "$tmp/nonewline"
+	exp=nonewline
+	got=$(head -1 "$tmp/nonewline")
+	[[ $got == $exp ]] || err_exit "head builtin fails to correctly handle files without an ending newline" \
+		"(expected $(printf %q "$exp"), got $(printf %q "$got"))"
+fi
+if builtin tail 2> /dev/null; then
+	print -n 'newline\nnonewline' > "$tmp/nonewline"
+	exp=nonewline
+	got=$(tail -1 "$tmp/nonewline")
+	[[ $got == $exp ]] || err_exit "tail builtin fails to correctly handle files without an ending newline" \
+		"(expected $(printf %q "$exp"), got $(printf %q "$got"))"
+fi
+
 # ======
 exit $((Errors<125?Errors:125))

src/lib/libcmd/head.c

@@ -1,7 +1,7 @@
 /***********************************************************************
 *                                                                      *
 *               This software is part of the ast package               *
-*          Copyright (c) 1992-2012 AT&T Intellectual Property          *
+*          Copyright (c) 1992-2013 AT&T Intellectual Property          *
 *          Copyright (c) 2020-2021 Contributors to ksh 93u+m           *
 *                      and is licensed under the                       *
 *                 Eclipse Public License, Version 1.0                  *
@@ -15,8 +15,8 @@
 *                            AT&T Research                             *
 *                           Florham Park NJ                            *
 *                                                                      *
-*                 Glenn Fowler <gsf@research.att.com>                  *
+*               Glenn Fowler <glenn.s.fowler@gmail.com>                *
-*                  David Korn <dgk@research.att.com>                   *
+*                    David Korn <dgkorn@gmail.com>                     *
 *                                                                      *
 ***********************************************************************/
 #pragma prototyped
@@ -28,7 +28,7 @@
  */
 
 static const char usage[] =
-"[-n?\n@(#)$Id: head (AT&T Research) 2012-05-31 $\n]"
+"[-n?\n@(#)$Id: head (AT&T Research) 2013-09-19 $\n]"
 "[--catalog?" ERROR_CATALOG "]"
 "[+NAME?head - output beginning portion of one or more files ]"
 "[+DESCRIPTION?\bhead\b copies one or more input files to standard "
@@ -81,6 +81,7 @@ b_head(int argc, register char** argv, Shbltin_t* context)
 	register off_t		keep = 10;
 	register off_t		skip = 0;
 	register int		delim = '\n';
+	off_t			moved;
 	int			header = 1;
 	char*			format = (char*)header_fmt+1;
 
@@ -145,9 +146,16 @@ b_head(int argc, register char** argv, Shbltin_t* context)
 			sfprintf(sfstdout, format, cp);
 		format = (char*)header_fmt;
 		if (skip > 0)
-			sfmove(fp, NiL, skip, delim);
+		{
-		if (sfmove(fp, sfstdout, keep, delim) < 0 && !ERROR_PIPE(errno) && errno != EINTR)
+			if ((moved = sfmove(fp, NiL, skip, delim)) < 0 && !ERROR_PIPE(errno) && errno != EINTR)
+				error(ERROR_system(0), "%s: skip error", cp);
+			if (delim >= 0 && moved < skip)
+				goto next;
+		}
+		if ((moved = sfmove(fp, sfstdout, keep, delim)) < 0 && !ERROR_PIPE(errno) && errno != EINTR ||
+			delim >= 0 && moved < keep && sfmove(fp, sfstdout, SF_UNBOUND, -1) < 0 && !ERROR_PIPE(errno) && errno != EINTR)
 			error(ERROR_system(0), "%s: read error", cp);
+	next:
 		if (fp != sfstdin)
 			sfclose(fp);
 	} while (cp = *argv++);

src/lib/libcmd/tail.c

@@ -1,7 +1,7 @@
 /***********************************************************************
 *                                                                      *
 *               This software is part of the ast package               *
-*          Copyright (c) 1992-2012 AT&T Intellectual Property          *
+*          Copyright (c) 1992-2013 AT&T Intellectual Property          *
 *          Copyright (c) 2020-2021 Contributors to ksh 93u+m           *
 *                      and is licensed under the                       *
 *                 Eclipse Public License, Version 1.0                  *
@@ -15,8 +15,8 @@
 *                            AT&T Research                             *
 *                           Florham Park NJ                            *
 *                                                                      *
-*                 Glenn Fowler <gsf@research.att.com>                  *
+*               Glenn Fowler <glenn.s.fowler@gmail.com>                *
-*                  David Korn <dgk@research.att.com>                   *
+*                    David Korn <dgkorn@gmail.com>                     *
 *                                                                      *
 ***********************************************************************/
 #pragma prototyped
@@ -29,7 +29,7 @@
  */
 
 static const char usage[] =
-"+[-?\n@(#)$Id: tail (AT&T Research) 2012-10-10 $\n]"
+"+[-?\n@(#)$Id: tail (AT&T Research) 2013-09-19 $\n]"
 "[--catalog?" ERROR_CATALOG "]"
 "[+NAME?tail - output trailing portion of one or more files ]"
 "[+DESCRIPTION?\btail\b copies one or more input files to standard output "
@@ -164,6 +164,7 @@ tailpos(register Sfio_t* fp, register Sfoff_t number, int delim)
 	register Sfoff_t	last;
 	register char*		s;
 	register char*		t;
+	unsigned char		incomplete;
 	struct stat		st;
 
 	last = sfsize(fp);
@@ -175,6 +176,7 @@ tailpos(register Sfio_t* fp, register Sfoff_t number, int delim)
 			return first;
 		return offset;
 	}
+	incomplete = 1;
 	for (;;)
 	{
 		if ((offset = last - SF_BUFSIZE) < first)
@@ -184,6 +186,15 @@ tailpos(register Sfio_t* fp, register Sfoff_t number, int delim)
 		if (!(s = sfreserve(fp, n, SF_LOCKR)))
 			return -1;
 		t = s + n;
+		if (incomplete)
+		{
+			if (t > s && *(t - 1) != delim && number-- <= 0)
+			{
+				sfread(fp, s, 0);
+				return offset + (t - s);
+			}
+			incomplete = 0;
+		}
 		while (t > s)
 			if (*--t == delim && number-- <= 0)
 			{
@@ -415,6 +426,7 @@ b_tail(int argc, char** argv, Shbltin_t* context)
 	char*			t;
 	char*			r;
 	char*			file;
+	Sfoff_t			moved;
 	Sfoff_t			offset;
 	Sfoff_t			number = DEFAULT;
 	unsigned long		timeout = 0;
@@ -765,8 +777,8 @@ b_tail(int argc, char** argv, Shbltin_t* context)
 			if (number < 0 || !number && (flags & POSITIVE))
 			{
 				sfset(ip, SF_SHARE, 1);
-				if (number < -1)
+				if (number < -1 && (moved = sfmove(ip, NiL, -(number + 1), delim)) >= 0 && delim >= 0 && moved < -(number + 1))
-					sfmove(ip, NiL, -number - 1, delim);
+					(void)sfgetr(ip, delim, SF_LASTR);
 				if (flags & REVERSE)
 					rev_line(ip, sfstdout, sfseek(ip, (Sfoff_t)0, SEEK_CUR));
 				else