• use BAFH for hash ipv NZAAT

• prep for release
• fix minor nits in manpage and tests

check.t

@@ -1,4 +1,4 @@
-# $MirOS: src/bin/mksh/check.t,v 1.639 2014/01/05 21:57:22 tg Exp $
+# $MirOS: src/bin/mksh/check.t,v 1.640 2014/01/11 18:09:37 tg Exp $
 # OpenBSD src/regress/bin/ksh updated: 2013/12/02 20:39:44
 #-
 # Copyright © 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010,
@@ -27,7 +27,7 @@
 # http://www.freebsd.org/cgi/cvsweb.cgi/src/tools/regression/bin/test/regress.sh?rev=HEAD
 
 expected-stdout:
-	@(#)MIRBSD KSH R49 2014/01/05
+	@(#)MIRBSD KSH R49 2014/01/11
 description:
 	Check version of shell.
 stdin:
@@ -36,7 +36,7 @@ name: KSH_VERSION
 category: shell:legacy-no
 ---
 expected-stdout:
-	@(#)LEGACY KSH R49 2014/01/05
+	@(#)LEGACY KSH R49 2014/01/11
 description:
 	Check version of legacy shell.
 stdin:
@@ -2075,12 +2075,18 @@ stdin:
 	echo [!-ab]*
 	echo [!ab]*
 	echo []ab]*
+	:>'./!bc'
+	:>'./^bc'
+	echo [^ab]*
+	echo [!ab]*
 expected-stdout:
 	-bc abc bbc
 	-bc abc bbc
 	cbc
 	-bc cbc
 	abc bbc
+	^bc abc bbc
+	!bc -bc ^bc cbc
 ---
 name: glob-range-2
 description:
@@ -7685,14 +7691,9 @@ stdin:
 	typeset -i8 foo=10
 	bar=baz
 	unset baz
-	bla=foo
 	print ${foo@#} ${bar@#} ${baz@#} .
-	print ${foo@#123} ${bar@#456} ${baz@#789} .
-	print ${foo@#bla} ${bar@#bar} ${baz@#OPTIND} .
 expected-stdout:
-	D50219A0 20E5DB5B 00000000 .
+	9B15FBFB CFBDD32B 00000000 .
-	554A1C76 004A212E CB209562 .
-	6B21CF91 20E5DB5B 124EA49D .
 ---
 name: varexpand-special-quote
 description:

dot.mkshrc

@@ -1,5 +1,5 @@
 # $Id$
-# $MirOS: src/bin/mksh/dot.mkshrc,v 1.87 2014/01/05 22:17:57 tg Exp $
+# $MirOS: src/bin/mksh/dot.mkshrc,v 1.88 2014/01/11 18:09:39 tg Exp $
 #-
 # Copyright (c) 2002, 2003, 2004, 2006, 2007, 2008, 2009, 2010,
 #		2011, 2012, 2013, 2014
@@ -333,9 +333,12 @@ function Lb64encode {
 	(( u )) || set -U
 }
 
-# mksh NUL counting, never zero
+# Better Avalanche for the Jenkins Hash
-typeset -Z11 -Uui16 Lnzathash_v
+typeset -Z11 -Uui16 Lbafh_v
-function Lnzathash_add {
+function Lbafh_init {
+	Lbafh_v=0
+}
+function Lbafh_add {
 	[[ -o utf8-mode ]]; local u=$?
 	set +U
 	local s
@@ -348,45 +351,19 @@ function Lnzathash_add {
 	local -i i=0 n=${#s[*]}
 
 	while (( i < n )); do
-		((# Lnzathash_v = (Lnzathash_v + s[i++] + 1) * 1025 ))
+		((# Lbafh_v = (Lbafh_v + s[i++] + 1) * 1025 ))
-		((# Lnzathash_v ^= Lnzathash_v >> 6 ))
+		((# Lbafh_v ^= Lbafh_v >> 6 ))
 	done
 
 	(( u )) || set -U
 }
-function Lnzaathash_end {
+function Lbafh_finish {
-	((# Lnzathash_v *= 1025 ))
+	local -Ui t
-	((# Lnzathash_v ^= Lnzathash_v >> 6 ))
-	((# Lnzathash_v += Lnzathash_v << 3 ))
-	((# Lnzathash_v = (Lnzathash_v ^
-	    (Lnzathash_v >> 11)) * 32769 ))
-	print ${Lnzathash_v#16#}
-}
-function Lnzaathash {
-	Lnzathash_v=0
-	Lnzathash_add "$@"
-	Lnzaathash_end
-}
-function Lnzathash {
-	Lnzathash_v=0
-	Lnzathash_add "$@"
-	Lnzathash_end
-}
-function Lnzathash_end {
-	if (( Lnzathash_v )); then
-		Lnzaathash_end
-	else
-		Lnzathash_v=1
-		print ${Lnzathash_v#16#}
-	fi
-}
-function Lnzathash_mix {
-	local -Uui16 t
 
-	((# t = ((Lnzathash_v >> 7) & 0x01010101) * 0x1B ))
+	((# t = (((Lbafh_v >> 7) & 0x01010101) * 0x1B) ^ \
-	((# t ^= (Lnzathash_v << 1) & 0xFEFEFEFE ))
+	    ((Lbafh_v << 1) & 0xFEFEFEFE) ))
-	((# Lnzathash_v = t ^ (t <<< 24) ^ (Lnzathash_v <<< 8) ^ \
+	((# Lbafh_v = t ^ (t >>> 8) ^ (Lbafh_v >>> 8) ^ \
-	    (Lnzathash_v <<< 16) ^ (Lnzathash_v <<< 24) ))
+	    (Lbafh_v >>> 16) ^ (Lbafh_v >>> 24) ))
 	:
 }
 

eval.c

@@ -23,7 +23,7 @@
 
 #include "sh.h"
 
-__RCSID("$MirOS: src/bin/mksh/eval.c,v 1.146 2014/01/05 21:57:25 tg Exp $");
+__RCSID("$MirOS: src/bin/mksh/eval.c,v 1.147 2014/01/11 18:09:39 tg Exp $");
 
 /*
  * string expansion
@@ -412,27 +412,10 @@ expand(
 					if (stype)
 						sp += slen;
 					switch (stype & 0x17F) {
-					case 0x100 | '#': {
+					case 0x100 | '#':
-						char *beg, *end;
-						mksh_ari_t seed;
-						register uint32_t h;
-
-						beg = wdcopy(sp, ATEMP);
-						end = beg + (wdscan(sp, CSUBST) - sp);
-						end[-2] = EOS;
-						end = wdstrip(beg, 0);
-						afree(beg, ATEMP);
-						evaluate(substitute(end, 0),
-						    &seed, KSH_UNWIND_ERROR, true);
-						/* hash with seed, for now */
-						h = seed;
-						NZATUpdateString(h,
-						    str_val(st->var));
-						NZAATFinish(h);
 						x.str = shf_smprintf("%08X",
-						    (unsigned int)h);
+						    (unsigned int)hash(str_val(st->var)));
 						break;
-					}
 					case 0x100 | 'Q': {
 						struct shf shf;
 

lex.c

@@ -2,7 +2,7 @@
 
 /*-
  * Copyright (c) 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010,
- *		 2011, 2012, 2013
+ *		 2011, 2012, 2013, 2014
  *	Thorsten Glaser <tg@mirbsd.org>
  *
  * Provided that these terms and disclaimer and all copyright notices
@@ -23,7 +23,7 @@
 
 #include "sh.h"
 
-__RCSID("$MirOS: src/bin/mksh/lex.c,v 1.191 2014/01/05 19:11:45 tg Exp $");
+__RCSID("$MirOS: src/bin/mksh/lex.c,v 1.192 2014/01/11 18:09:40 tg Exp $");
 
 /*
  * states while lexing word
@@ -1440,6 +1440,7 @@ getsc_line(Source *s)
 		alarm(0);
 	}
 	cp = Xstring(s->xs, xp);
+	rndpush(cp);
 	s->start = s->str = cp;
 	strip_nuls(Xstring(s->xs, xp), Xlength(s->xs, xp));
 	/* Note: if input is all nulls, this is not eof */

main.c

@@ -34,7 +34,7 @@
 #include <locale.h>
 #endif
 
-__RCSID("$MirOS: src/bin/mksh/main.c,v 1.276 2014/01/11 16:26:28 tg Exp $");
+__RCSID("$MirOS: src/bin/mksh/main.c,v 1.277 2014/01/11 18:09:40 tg Exp $");
 
 extern char **environ;
 
@@ -142,21 +142,6 @@ rndsetup(void)
 	return ((mksh_uari_t)h);
 }
 
-uint32_t
-chvt_rndsetup(const void *bp, size_t sz)
-{
-	register uint32_t h;
-
-	NZATInit(h);
-	/* variation through pid, ppid, and the works */
-	NZATUpdateMem(h, &rndsetupstate, sizeof(rndsetupstate));
-	/* some variation, some possibly entropy, depending on OE */
-	NZATUpdateMem(h, bp, sz);
-	NZAATFinish(h);
-
-	return (h);
-}
-
 void
 chvt_reinit(void)
 {
@@ -359,9 +344,14 @@ main_init(int argc, const char *argv[], Source **sp, struct block **lp)
 #endif
 
 	/* import environment */
-	if (environ != NULL)
+	if (environ != NULL) {
-		for (wp = (const char **)environ; *wp != NULL; wp++)
+		wp = (const char **)environ;
+		while (*wp != NULL) {
+			rndpush(*wp);
 			typeset(*wp, IMPORT | EXPORT, 0, 0, 0);
+			++wp;
+		}
+	}
 
 	/* for security */
 	typeset(initifs, 0, 0, 0, 0);

mksh.1

@@ -1,4 +1,4 @@
-.\" $MirOS: src/bin/mksh/mksh.1,v 1.328 2014/01/05 21:57:27 tg Exp $
+.\" $MirOS: src/bin/mksh/mksh.1,v 1.329 2014/01/11 18:09:41 tg Exp $
 .\" $OpenBSD: ksh.1,v 1.149 2013/12/18 13:53:11 millert Exp $
 .\"-
 .\" Copyright © 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009,
@@ -74,7 +74,7 @@
 .\" with -mandoc, it might implement .Mx itself, but we want to
 .\" use our own definition. And .Dd must come *first*, always.
 .\"
-.Dd $Mdocdate: January 5 2014 $
+.Dd $Mdocdate: January 11 2014 $
 .\"
 .\" Check which macro package we use, and do other -mdoc setup.
 .\"
@@ -1693,17 +1693,10 @@ Currently,
 must start with a space, opening parenthesis or digit to be recognised.
 Cannot be applied to a vector.
 .Pp
-.It Xo
+.It Pf ${ Ns Ar name Ns @#}
-.Pf ${ Ar name
+The hash (using the BAFH algorithm) of the expansion of
-.Pf @# Ns Oo Ar seed Oc Ns }
+.Ar name .
-.Xc
+This is also used internally for the shell's hashtables.
-The internal hash of the expansion of
-.Ar name ,
-with an optional (defaulting to zero)
-.Op Ar seed .
-At the moment, this is NZAAT (a 32-bit hash based on
-Bob Jenkins' one-at-a-time hash), but this is not set.
-This is the hash the shell uses internally for its associative arrays.
 .Pp
 .It Pf ${ Ns Ar name Ns @Q}
 A quoted expression safe for re-entry, whose value is the value of the
@@ -3633,7 +3626,7 @@ Since expressions may need to be quoted,
 is syntactic sugar for
 .No let \&" Ns Ar expr Ns \&" .
 .Pp
-.It let]
+.It Ic let]
 Internally used alias for
 .Ic let .
 .Pp

sh.h

@@ -169,9 +169,9 @@
 #endif
 
 #ifdef EXTERN
-__RCSID("$MirOS: src/bin/mksh/sh.h,v 1.683 2014/01/11 16:26:28 tg Exp $");
+__RCSID("$MirOS: src/bin/mksh/sh.h,v 1.684 2014/01/11 18:09:42 tg Exp $");
 #endif
-#define MKSH_VERSION "R49 2014/01/05"
+#define MKSH_VERSION "R49 2014/01/11"
 
 /* arithmetic types: C implementation */
 #if !HAVE_CAN_INTTYPES
@@ -1616,48 +1616,6 @@ EXTERN struct timeval j_usrtime, j_systime;
 } while (/* CONSTCOND */ 0)
 
 
-/* NZAAT hash based on Bob Jenkins' one-at-a-time hash */
-
-/* From: src/kern/include/nzat.h,v 1.2 2011/07/18 00:35:40 tg Exp $ */
-
-#define NZATInit(h) do {					\
-	(h) = 0;						\
-} while (/* CONSTCOND */ 0)
-
-#define NZATUpdateByte(h,b) do {				\
-	(h) += (uint8_t)(b);					\
-	++(h);							\
-	(h) += (h) << 10;					\
-	(h) ^= (h) >> 6;					\
-} while (/* CONSTCOND */ 0)
-
-#define NZATUpdateMem(h,p,z) do {				\
-	register const uint8_t *NZATUpdateMem_p;		\
-	register size_t NZATUpdateMem_z = (z);			\
-								\
-	NZATUpdateMem_p = (const void *)(p);			\
-	while (NZATUpdateMem_z--)				\
-		NZATUpdateByte((h), *NZATUpdateMem_p++);	\
-} while (/* CONSTCOND */ 0)
-
-#define NZATUpdateString(h,s) do {				\
-	register const char *NZATUpdateString_s;		\
-	register uint8_t NZATUpdateString_c;			\
-								\
-	NZATUpdateString_s = (const void *)(s);			\
-	while ((NZATUpdateString_c = *NZATUpdateString_s++))	\
-		NZATUpdateByte((h), NZATUpdateString_c);	\
-} while (/* CONSTCOND */ 0)
-
-#define NZAATFinish(h) do {					\
-	(h) += (h) << 10;					\
-	(h) ^= (h) >> 6;					\
-	(h) += (h) << 3;					\
-	(h) ^= (h) >> 11;					\
-	(h) += (h) << 15;					\
-} while (/* CONSTCOND */ 0)
-
-
 /* lalloc.c */
 void ainit(Area *);
 void afreeall(Area *);
@@ -1895,7 +1853,6 @@ struct tbl **ktsort(struct table *);
 void DF(const char *, ...)
     MKSH_A_FORMAT(__printf__, 1, 2);
 #endif
-uint32_t chvt_rndsetup(const void *, size_t) MKSH_A_PURE;
 /* misc.c */
 void setctypes(const char *, int);
 void initctypes(void);
@@ -2010,8 +1967,10 @@ size_t array_ref_len(const char *) MKSH_A_PURE;
 char *arrayname(const char *);
 mksh_uari_t set_array(const char *, bool, const char **);
 uint32_t hash(const void *) MKSH_A_PURE;
+uint32_t chvt_rndsetup(const void *, size_t) MKSH_A_PURE;
 mksh_ari_t rndget(void);
 void rndset(unsigned long);
+void rndpush(const void *);
 
 enum Test_op {
 	/* non-operator */

var.c

@@ -22,12 +22,13 @@
  */
 
 #include "sh.h"
+#include "mirhash.h"
 
 #if defined(__OpenBSD__)
 #include <sys/sysctl.h>
 #endif
 
-__RCSID("$MirOS: src/bin/mksh/var.c,v 1.176 2014/01/05 21:57:29 tg Exp $");
+__RCSID("$MirOS: src/bin/mksh/var.c,v 1.177 2014/01/11 18:09:43 tg Exp $");
 
 /*-
  * Variables
@@ -40,7 +41,7 @@ __RCSID("$MirOS: src/bin/mksh/var.c,v 1.176 2014/01/05 21:57:29 tg Exp $");
  */
 
 static struct table specials;
-static uint32_t lcg_state = 5381;
+static uint32_t lcg_state = 5381, qh_state = 4711;
 
 static char *formatstr(struct tbl *, const char *);
 static void exportprep(struct tbl *, const char *);
@@ -1490,6 +1491,11 @@ set_array(const char *var, bool reset, const char **vals)
 void
 change_winsz(void)
 {
+	struct timeval tv;
+
+	mksh_TIME(tv);
+	BAFHUpdateMem_mem(qh_state, &tv, sizeof(tv));
+
 #ifdef TIOCGWINSZ
 	/* check if window size has changed */
 	if (tty_init_fd() < 2) {
@@ -1520,9 +1526,28 @@ hash(const void *s)
 {
 	register uint32_t h;
 
-	NZATInit(h);
+	BAFHInit(h);
-	NZATUpdateString(h, s);
+	BAFHUpdateStr_reg(h, s);
-	NZAATFinish(h);
+	BAFHFinish_reg(h);
+	return (h);
+}
+
+uint32_t
+chvt_rndsetup(const void *bp, size_t sz)
+{
+	register uint32_t h;
+
+	/* use LCG as seed but try to get them to deviate immediately */
+	h = lcg_state;
+	(void)rndget();
+	BAFHFinish_reg(h);
+	/* variation through pid, ppid, and the works */
+	BAFHUpdateMem_reg(h, &rndsetupstate, sizeof(rndsetupstate));
+	/* some variation, some possibly entropy, depending on OE */
+	BAFHUpdateMem_reg(h, bp, sz);
+	/* mix them all up */
+	BAFHFinish_reg(h);
+
 	return (h);
 }
 
@@ -1540,28 +1565,62 @@ void
 rndset(unsigned long v)
 {
 	register uint32_t h;
+#if defined(arc4random_pushb_fast) || defined(MKSH_A4PB)
+	register uint32_t t;
+#endif
+	struct {
+		struct timeval tv;
+		void *sp;
+		uint32_t qh;
+		pid_t pp;
+		short r;
+	} z;
 
-	NZATInit(h);
+	h = lcg_state;
-	NZATUpdateMem(h, &lcg_state, sizeof(lcg_state));
+	BAFHFinish_reg(h);
-	NZATUpdateMem(h, &v, sizeof(v));
+	BAFHUpdateMem_reg(h, &v, sizeof(v));
+
+	mksh_TIME(z.tv);
+	z.sp = &lcg_state;
+	z.pp = procpid;
+	z.r = (short)rndget();
 
 #if defined(arc4random_pushb_fast) || defined(MKSH_A4PB)
+	t = qh_state;
+	BAFHFinish_reg(t);
+	z.qh = (t & 0xFFFF8000) | rndget();
+	lcg_state = (t << 15) | rndget();
 	/*
 	 * either we have very chap entropy get and push available,
 	 * with malloc() pulling in this code already anyway, or the
 	 * user requested us to use the old functions
 	 */
-	lcg_state = h;
+	t = h;
-	NZAATFinish(lcg_state);
+	BAFHUpdateMem_reg(t, &lcg_state, sizeof(lcg_state));
+	BAFHFinish_reg(t);
+	lcg_state = t;
 #if defined(arc4random_pushb_fast)
 	arc4random_pushb_fast(&lcg_state, sizeof(lcg_state));
 	lcg_state = arc4random();
 #else
 	lcg_state = arc4random_pushb(&lcg_state, sizeof(lcg_state));
 #endif
-	NZATUpdateMem(h, &lcg_state, sizeof(lcg_state));
+	BAFHUpdateMem_reg(h, &lcg_state, sizeof(lcg_state));
+#else
+	z.qh = qh_state;
 #endif
 
-	NZAATFinish(h);
+	BAFHUpdateMem_reg(h, &z, sizeof(z));
+	BAFHFinish_reg(h);
 	lcg_state = h;
 }
+
+void
+rndpush(const void *s)
+{
+	register uint32_t h = qh_state;
+
+	BAFHUpdateStr_reg(h, s);
+	BAFHUpdateOctet_reg(h, 0);
+	qh_state = h;
+}