initial check-in, version 0.5.0

src/string/bcmp.c

@@ -0,0 +1,7 @@
+#include <string.h>
+#include <strings.h>
+
+int bcmp(const void *s1, const void *s2, size_t n)
+{
+	return memcmp(s1, s2, n);
+}

src/string/bcopy.c

@@ -0,0 +1,7 @@
+#include <string.h>
+#include <strings.h>
+
+void bcopy(const void *s1, void *s2, size_t n)
+{
+	memmove(s2, s1, n);
+}

src/string/bzero.c

@@ -0,0 +1,7 @@
+#include <string.h>
+#include <strings.h>
+
+void bzero(void *s, size_t n)
+{
+	memset(s, 0, n);
+}

src/string/index.c

@@ -0,0 +1,7 @@
+#include <string.h>
+#include <strings.h>
+
+char *index(const char *s, int c)
+{
+	return strchr(s, c);
+}

src/string/memchr.c

@@ -0,0 +1,24 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <limits.h>
+
+#define SS (sizeof(size_t))
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+#define HIGHS (ONES * (UCHAR_MAX/2+1))
+#define HASZERO(x) ((x)-ONES & ~(x) & HIGHS)
+
+void *memchr(const void *src, int c, size_t n)
+{
+	const unsigned char *s = src;
+	c = (unsigned char)c;
+	for (; ((uintptr_t)s & ALIGN) && n && *s != c; s++, n--);
+	if (n && *s != c) {
+		const size_t *w;
+		size_t k = ONES * c;
+		for (w = (const void *)s; n>=SS && !HASZERO(*w^k); w++, n-=SS);
+		for (s = (const void *)w; n && *s != c; s++, n--);
+	}
+	return n ? (void *)s : 0;
+}

src/string/memcmp.c

@@ -0,0 +1,8 @@
+#include <string.h>
+
+int memcmp(const void *vl, const void *vr, size_t n)
+{
+	const unsigned char *l=vl, *r=vr;
+	for (; n && *l == *r; n--, l++, r++);
+	return n ? *l-*r : 0;
+}

src/string/memcpy.c

@@ -0,0 +1,29 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+
+#define SS (sizeof(size_t))
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+
+void *memcpy(void *dest, const void *src, size_t n)
+{
+	unsigned char *d = dest;
+	const unsigned char *s = src;
+
+	if (((uintptr_t)d & ALIGN) != ((uintptr_t)s & ALIGN))
+		goto misaligned;
+
+	for (; ((uintptr_t)d & ALIGN) && n; n--) *d++ = *s++;
+	if (n) {
+		size_t *wd = (void *)d;
+		const size_t *ws = (const void *)s;
+
+		for (; n>=SS; n-=SS) *wd++ = *ws++;
+		d = (void *)wd;
+		s = (const void *)ws;
+misaligned:
+		for (; n; n--) *d++ = *s++;
+	}
+	return dest;
+}

src/string/memmove.c

@@ -0,0 +1,14 @@
+#include <string.h>
+
+void *memmove(void *dest, const void *src, size_t n)
+{
+	char *d = dest;
+	const char *s = src;
+	if (d==s) return d;
+	if ((size_t)(d-s) < n) {
+		while (n--) d[n] = s[n];
+		return dest;
+	}
+	/* Assumes memcpy is overlap-safe when dest < src */
+	return memcpy(d, s, n);
+}

src/string/mempcpy.c

@@ -0,0 +1,7 @@
+#include <string.h>
+
+void *mempcpy(void *dest, void *src, size_t n)
+{
+	memcpy(dest, src, n);
+	return (char *)dest + n;
+}

src/string/memset.c

@@ -0,0 +1,21 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <limits.h>
+
+#define SS (sizeof(size_t))
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+
+void *memset(void *dest, int c, size_t n)
+{
+	unsigned char *s = dest;
+	c = (unsigned char)c;
+	for (; ((uintptr_t)s & ALIGN) && n; n--) *s++ = c;
+	if (n) {
+		size_t *w, k = ONES * c;
+		for (w = (void *)s; n>=SS; n-=SS, w++) *w = k;
+		for (s = (void *)w; n; n--, s++) *s = c;
+	}
+	return dest;
+}

src/string/rindex.c

@@ -0,0 +1,7 @@
+#include <string.h>
+#include <strings.h>
+
+char *rindex(const char *s, int c)
+{
+	return strrchr(s, c);
+}

src/string/stpcpy.c

@@ -0,0 +1,29 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <limits.h>
+#include "libc.h"
+
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+#define HIGHS (ONES * (UCHAR_MAX/2+1))
+#define HASZERO(x) ((x)-ONES & ~(x) & HIGHS)
+
+char *__stpcpy(char *d, const char *s)
+{
+	size_t *wd;
+	const size_t *ws;
+
+	if (((uintptr_t)s & ALIGN) == ((uintptr_t)d & ALIGN)) {
+		for (; (*d=*s) && ((uintptr_t)s & ALIGN); s++, d++);
+		if (!*s) return d;
+		wd=(void *)d; ws=(const void *)s;
+		for (; !HASZERO(*ws); *wd++ = *ws++);
+		d=(void *)wd; s=(const void *)ws;
+	}
+	for (; (*d=*s); s++, d++);
+
+	return d;
+}
+
+weak_alias(__stpcpy, stpcpy);

src/string/stpncpy.c

@@ -0,0 +1,32 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <limits.h>
+#include "libc.h"
+
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+#define HIGHS (ONES * (UCHAR_MAX/2+1))
+#define HASZERO(x) ((x)-ONES & ~(x) & HIGHS)
+
+char *__stpncpy(char *d, const char *s, size_t n)
+{
+	size_t *wd;
+	const size_t *ws;
+
+	if (((uintptr_t)s & ALIGN) != ((uintptr_t)d & ALIGN)) {
+		for (; ((uintptr_t)s & ALIGN) && n && (*d=*s); n--, s++, d++);
+		if (!n || !*s) goto tail;
+		wd=(void *)d; ws=(const void *)s;
+		for (; n>=sizeof(size_t) && !HASZERO(*ws);
+		       n-=sizeof(size_t), ws++, *wd++) *wd = *ws;
+		d=(void *)wd; s=(const void *)ws;
+	}
+	for (; n && (*d=*s); n--, s++, d++);
+tail:
+	memset(d, 0, n);
+	return d;
+}
+
+weak_alias(__stpncpy, stpncpy);
+

src/string/strcasecmp.c

@@ -0,0 +1,9 @@
+#include <strings.h>
+#include <ctype.h>
+
+int strcasecmp(const char *_l, const char *_r)
+{
+	const unsigned char *l=_l, *r=_r;
+	for (; *l && *r && (*l == *r || tolower(*l) == tolower(*r)); l++, r++);
+	return tolower(*l) - tolower(*r);
+}

src/string/strcasestr.c

@@ -0,0 +1,7 @@
+#include <string.h>
+
+char *strcasestr(const char *h, const char *n)
+{
+	//FIXME!
+	return strstr(h, n);
+}

src/string/strcat.c

@@ -0,0 +1,7 @@
+#include <string.h>
+
+char *strcat(char *dest, const char *src)
+{
+	strcpy(dest + strlen(dest), src);
+	return dest;
+}

src/string/strchr.c

@@ -0,0 +1,23 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <limits.h>
+
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+#define HIGHS (ONES * (UCHAR_MAX/2+1))
+#define HASZERO(x) ((x)-ONES & ~(x) & HIGHS)
+
+char *strchr(const char *s, int c)
+{
+	c = (char)c;
+	if (!c) return (char *)s + strlen(s);
+	for (; ((uintptr_t)s & ALIGN) && *s && *s != c; s++);
+	if (*s && *s != c) {
+		const size_t *w;
+		size_t k = ONES * c;
+		for (w = (const void *)s; !HASZERO(*w) && !HASZERO(*w^k); w++);
+		for (s = (const void *)w; *s && *s != c; s++);
+	}
+	return *s ? (char *)s : 0;
+}

src/string/strchrnul.c

@@ -0,0 +1,7 @@
+#include <string.h>
+
+char *strchrnul(const char *s, int c)
+{
+	char *p = strchr(s, c);
+	return p ? p : (char *)s + strlen(s);
+}

src/string/strcmp.c

@@ -0,0 +1,7 @@
+#include <string.h>
+
+int strcmp(const char *l, const char *r)
+{
+	for (; *l==*r && *l && *r; l++, r++);
+	return *(unsigned char *)l - *(unsigned char *)r;
+}

src/string/strcpy.c

@@ -0,0 +1,16 @@
+#include <string.h>
+
+char *__stpcpy(char *, const char *);
+
+char *strcpy(char *dest, const char *src)
+{
+#if 1
+	__stpcpy(dest, src);
+	return dest;
+#else
+	const unsigned char *s = src;
+	unsigned char *d = dest;
+	while ((*d++ = *s++));
+	return dest;
+#endif
+}

src/string/strcspn.c

@@ -0,0 +1,20 @@
+#include <string.h>
+
+#define BITOP(a,b,op) \
+ ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
+
+size_t strcspn(const char *_s, const char *_c)
+{
+	const unsigned char *s = _s;
+	const unsigned char *c = _c;
+	const unsigned char *a = s;
+	size_t byteset[32/sizeof(size_t)];
+
+	if (!c[0]) return strlen(s);
+	if (!c[1]) return (s=strchr(s, *c)) ? s-a : strlen(a);
+
+	memset(byteset, 0, sizeof byteset);
+	for (; *c && BITOP(byteset, *c, |=); c++);
+	for (; *s && !BITOP(byteset, *s, &); s++);
+	return s-a;
+}

src/string/strdup.c

@@ -0,0 +1,13 @@
+#include <stdlib.h>
+#include <string.h>
+#include "libc.h"
+
+char *__strdup(const char *s)
+{
+	size_t l = strlen(s);
+	char *d = malloc(l+1);
+	if (!d) return NULL;
+	return memcpy(d, s, l+1);
+}
+
+weak_alias(__strdup, strdup);

src/string/strerror_r.c

@@ -0,0 +1,11 @@
+#include <string.h>
+#include <errno.h>
+
+int strerror_r(int err, char *buf, size_t buflen)
+{
+	char *msg = strerror(err);
+	if (strlen(msg) >= buflen)
+		return ERANGE;
+	strcpy(buf, msg);
+	return 0;
+}

src/string/strlcat.c

@@ -0,0 +1,8 @@
+#include <string.h>
+
+size_t strlcat(char *d, const char *s, size_t n)
+{
+	size_t l = strnlen(d, n);
+	if (l == n) return l + strlen(s);
+	return l + strlcpy(d+l, s, n-l);
+}

src/string/strlcpy.c

@@ -0,0 +1,32 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <limits.h>
+#include "libc.h"
+
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+#define HIGHS (ONES * (UCHAR_MAX/2+1))
+#define HASZERO(x) ((x)-ONES & ~(x) & HIGHS)
+
+size_t strlcpy(char *d, const char *s, size_t n)
+{
+	char *d0 = d;
+	size_t *wd;
+	const size_t *ws;
+
+	if (!n--) goto finish;
+	if (((uintptr_t)s & ALIGN) != ((uintptr_t)d & ALIGN)) {
+		for (; ((uintptr_t)s & ALIGN) && n && (*d=*s); n--, s++, d++);
+		if (n && *s) {
+			wd=(void *)d; ws=(const void *)s;
+			for (; n>=sizeof(size_t) && !HASZERO(*ws);
+			       n-=sizeof(size_t), ws++, *wd++) *wd = *ws;
+			d=(void *)wd; s=(const void *)ws;
+		}
+	}
+	for (; n && (*d=*s); n--, s++, d++);
+	*d = 0;
+finish:
+	return d-d0 + strlen(s);
+}

src/string/strlen.c

@@ -0,0 +1,21 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <limits.h>
+
+#define ALIGN (sizeof(size_t)-1)
+#define ONES ((size_t)-1/UCHAR_MAX)
+#define HIGHS (ONES * (UCHAR_MAX/2+1))
+#define HASZERO(x) ((x)-ONES & ~(x) & HIGHS)
+
+size_t strlen(const char *s)
+{
+	const char *a = s;
+	const size_t *w;
+	for (; ((uintptr_t)s & ALIGN) && *s; s++);
+	if (*s) {
+		for (w = (const void *)s; !HASZERO(*w); w++);
+		for (s = (const void *)w; *s; s++);
+	}
+	return s-a;
+}

src/string/strncasecmp.c

@@ -0,0 +1,10 @@
+#include <strings.h>
+#include <ctype.h>
+
+int strncasecmp(const char *_l, const char *_r, size_t n)
+{
+	const unsigned char *l=_l, *r=_r;
+	if (!n--) return 0;
+	for (; *l && *r && n && (*l == *r || tolower(*l) == tolower(*r)); l++, r++, n--);
+	return tolower(*l) - tolower(*r);
+}

src/string/strncat.c

@@ -0,0 +1,10 @@
+#include <string.h>
+
+char *strncat(char *d, const char *s, size_t n)
+{
+	char *a = d;
+	d += strlen(d);
+	while (n && (*d++ = *s++)) n--;
+	*d++ = 0;
+	return a;
+}

src/string/strncmp.c

@@ -0,0 +1,9 @@
+#include <string.h>
+
+int strncmp(const char *_l, const char *_r, size_t n)
+{
+	const unsigned char *l=_l, *r=_r;
+	if (!n--) return 0;
+	for (; *l && *r && n && *l == *r ; l++, r++, n--);
+	return *l - *r;
+}

src/string/strncpy.c

@@ -0,0 +1,9 @@
+#include <string.h>
+
+char *__stpncpy(char *, const char *, size_t);
+
+char *strncpy(char *d, const char *s, size_t n)
+{
+	__stpncpy(d, s, n);
+	return d;
+}

src/string/strndup.c

@@ -0,0 +1,12 @@
+#include <stdlib.h>
+#include <string.h>
+
+char *strndup(const char *s, size_t n)
+{
+	size_t l = strnlen(s, n);
+	char *d = malloc(l+1);
+	if (!d) return NULL;
+	memcpy(d, s, l);
+	d[l] = 0;
+	return d;
+}

src/string/strnlen.c

@@ -0,0 +1,7 @@
+#include <string.h>
+
+size_t strnlen(const char *s, size_t n)
+{
+	const char *p = memchr(s, 0, n);
+	return p ? p-s : n;
+}

src/string/strpbrk.c

@@ -0,0 +1,7 @@
+#include <string.h>
+
+char *strpbrk(const char *s, const char *b)
+{
+	s += strcspn(s, b);
+	return *s ? (char *)s : 0;
+}

src/string/strrchr.c

@@ -0,0 +1,9 @@
+#include <string.h>
+
+char *strrchr(const char *s, int c)
+{
+	const char *p;
+	c = (char)c;
+	for (p=s+strlen(s); p>=s && *p!=c; p--);
+	return p>=s ? (char *)p : 0;
+}

src/string/strsep.c

@@ -0,0 +1,12 @@
+#include <string.h>
+
+char *strsep(char **str, const char *sep)
+{
+	char *s = *str, *end;
+	if (!s) return NULL;
+	end = s + strcspn(s, sep);
+	if (*end) *end++ = 0;
+	else end = 0;
+	*str = end;
+	return s;
+}

src/string/strsignal.c

@@ -0,0 +1,98 @@
+#include <signal.h>
+
+#if (SIGHUP == 1) && (SIGINT == 2) && (SIGQUIT == 3) && (SIGILL == 4) \
+ && (SIGTRAP == 5) && (SIGABRT == 6) && (SIGBUS == 7) && (SIGFPE == 8) \
+ && (SIGKILL == 9) && (SIGUSR1 == 10) && (SIGSEGV == 11) && (SIGUSR2 == 12) \
+ && (SIGPIPE == 13) && (SIGALRM == 14) && (SIGTERM == 15) && (SIGSTKFLT == 16) \
+ && (SIGCHLD == 17) && (SIGCONT == 18) && (SIGSTOP == 19) && (SIGTSTP == 20) \
+ && (SIGTTIN == 21) && (SIGTTOU == 22) && (SIGURG == 23) && (SIGXCPU == 24) \
+ && (SIGXFSZ == 25) && (SIGVTALRM == 26) && (SIGPROF == 27) && (SIGWINCH == 28) \
+ && (SIGPOLL == 29) && (SIGPWR == 30) && (SIGSYS == 31)
+
+#define sigmap(x) x
+
+#else
+
+static const char map[] = {
+	[SIGHUP]    = 1,
+	[SIGINT]    = 2,
+	[SIGQUIT]   = 3,
+	[SIGILL]    = 4,
+	[SIGTRAP]   = 5,
+	[SIGABRT]   = 6,
+	[SIGBUS]    = 7,
+	[SIGFPE]    = 8,
+	[SIGKILL]   = 9,
+	[SIGUSR1]   = 10,
+	[SIGSEGV]   = 11,
+	[SIGUSR2]   = 12,
+	[SIGPIPE]   = 13,
+	[SIGALRM]   = 14,
+	[SIGTERM]   = 15,
+	[SIGSTKFLT] = 16,
+	[SIGCHLD]   = 17,
+	[SIGCONT]   = 18,
+	[SIGSTOP]   = 19,
+	[SIGTSTP]   = 20,
+	[SIGTTIN]   = 21,
+	[SIGTTOU]   = 22,
+	[SIGURG]    = 23,
+	[SIGXCPU]   = 24,
+	[SIGXFSZ]   = 25,
+	[SIGVTALRM] = 26,
+	[SIGPROF]   = 27,
+	[SIGWINCH]  = 28,
+	[SIGPOLL]   = 29,
+	[SIGPWR]    = 30,
+	[SIGSYS]    = 31
+};
+
+#define sigmap(x) ((unsigned)(x) > sizeof map ? 0 : map[(unsigned)(x)])
+
+#endif
+
+static const char strings[] =
+	"Unknown signal\0"
+	"Hangup\0"
+	"Interrupt\0"
+	"Quit\0"
+	"Illegal instruction\0"
+	"Trace/breakpoint trap\0"
+	"Aborted\0"
+	"Bus error\0"
+	"Floating point exception\0"
+	"Killed\0"
+	"User defined signal 1\0"
+	"Segmentation fault\0"
+	"User defined signal 2\0"
+	"Broken pipe\0"
+	"Alarm clock\0"
+	"Terminated\0"
+	"Stack fault\0"
+	"Child exited\0"
+	"Continued\0"
+	"Stopped (signal)\0"
+	"Stopped\0"
+	"Stopped (tty input)\0"
+	"Stopped (tty output)\0"
+	"Urgent I/O condition\0"
+	"CPU time limit exceeded\0"
+	"File size limit exceeded\0"
+	"Virtual timer expired\0"
+	"Profiling timer expired\0"
+	"Window changed\0"
+	"I/O possible\0"
+	"Power failure\0"
+	"Bad system call";
+
+char *strsignal(int signum)
+{
+	char *s = (char *)strings;
+
+	signum = sigmap(signum);
+	if ((unsigned)signum - 1 > 31) signum = 0;
+
+	for (; signum--; s++) for (; *s; s++);
+
+	return s;
+}

src/string/strspn.c

@@ -0,0 +1,22 @@
+#include <string.h>
+
+#define BITOP(a,b,op) \
+ ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
+
+size_t strspn(const char *_s, const char *_c)
+{
+	const unsigned char *s = _s;
+	const unsigned char *c = _c;
+	const unsigned char *a = s;
+	size_t byteset[32/sizeof(size_t)] = { 0 };
+
+	if (!c[0]) return 0;
+	if (!c[1]) {
+		for (; *s == *c; s++);
+		return s-a;
+	}
+
+	for (; *c && BITOP(byteset, *c, |=); c++);
+	for (; *s && BITOP(byteset, *s, &); s++);
+	return s-a;
+}

src/string/strstr.c

@@ -0,0 +1,166 @@
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+
+static char *twobyte_strstr(const unsigned char *h, const unsigned char *n)
+{
+	uint16_t nw = n[0]<<8 | n[1], hw = h[0]<<8 | h[1];
+	for (h++; *h && hw != nw; hw = hw<<8 | *++h);
+	return *h ? (char *)h-1 : 0;
+}
+
+static char *threebyte_strstr(const unsigned char *h, const unsigned char *n)
+{
+	uint32_t nw = n[0]<<24 | n[1]<<16 | n[2]<<8;
+	uint32_t hw = h[0]<<24 | h[1]<<16 | h[2]<<8;
+	for (h+=2; *h && hw != nw; hw = (hw|*++h)<<8);
+	return *h ? (char *)h-2 : 0;
+}
+
+static char *fourbyte_strstr(const unsigned char *h, const unsigned char *n)
+{
+	uint32_t nw = n[0]<<24 | n[1]<<16 | n[2]<<8 | n[3];
+	uint32_t hw = h[0]<<24 | h[1]<<16 | h[2]<<8 | h[3];
+	for (h+=3; *h && hw != nw; hw = hw<<8 | *++h);
+	return *h ? (char *)h-3 : 0;
+}
+
+#if 0
+static char *naive_strstr(const char *h, const char *n)
+{
+	size_t i;
+	for (i=0; n[i] && h[i]; i++)
+	for (   ; n[i] != h[i]; h++, i=0);
+	return n[i] ? 0 : (char *)h;
+}
+#endif
+
+#define MAX(a,b) ((a)>(b)?(a):(b))
+#define MIN(a,b) ((a)<(b)?(a):(b))
+
+#define BITOP(a,b,op) \
+ ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
+
+static char *twoway_strstr(const unsigned char *h, const unsigned char *n)
+{
+	const unsigned char *z;
+	size_t l, ip, jp, k, p, ms, p0, mem, mem0;
+	size_t byteset[32 / sizeof(size_t)] = { 0 };
+	size_t shift[256];
+
+	/* Computing length of needle and fill shift table */
+	for (l=0; n[l] && h[l]; l++)
+		BITOP(byteset, n[l], |=), shift[n[l]] = l+1;
+	if (n[l]) return 0; /* hit the end of h */
+
+	/* Compute maximal suffix */
+	ip = -1; jp = 0; k = p = 1;
+	while (jp+k<l) {
+		if (n[ip+k] == n[jp+k]) {
+			if (k == p) {
+				jp += p;
+				k = 1;
+			} else k++;
+		} else if (n[ip+k] > n[jp+k]) {
+			jp += k;
+			k = 1;
+			p = jp - ip;
+		} else {
+			ip = jp++;
+			k = p = 1;
+		}
+	}
+	ms = ip;
+	p0 = p;
+
+	/* And with the opposite comparison */
+	ip = -1; jp = 0; k = p = 1;
+	while (jp+k<l) {
+		if (n[ip+k] == n[jp+k]) {
+			if (k == p) {
+				jp += p;
+				k = 1;
+			} else k++;
+		} else if (n[ip+k] < n[jp+k]) {
+			jp += k;
+			k = 1;
+			p = jp - ip;
+		} else {
+			ip = jp++;
+			k = p = 1;
+		}
+	}
+	if (ip+1 > ms+1) ms = ip;
+	else p = p0;
+
+	/* Periodic needle? */
+	if (memcmp(n, n+p, ms+1)) {
+		mem0 = 0;
+		p = MAX(ms, l-ms-1) + 1;
+	} else mem0 = l-p;
+	mem = 0;
+
+	/* Initialize incremental end-of-haystack pointer */
+	z = h;
+
+	/* Search loop */
+	for (;;) {
+		/* Update incremental end-of-haystack pointer */
+		if (z-h < l) {
+			/* Fast estimate for MIN(l,63) */
+			size_t grow = l | 63;
+			const char *z2 = memchr(z, 0, grow);
+			if (z2) {
+				z = z2;
+				if (z-h < l) return 0;
+			} else z += grow;
+		}
+
+		/* Check last byte first; advance by shift on mismatch */
+		if (BITOP(byteset, h[l-1], &)) {
+			k = l-shift[h[l-1]];
+			//printf("adv by %zu (on %c) at [%s] (%zu;l=%zu)\n", k, h[l-1], h, shift[h[l-1]], l);
+			if (k) {
+				if (mem0 && mem && k < p) k = l-p;
+				h += k;
+				mem = 0;
+				continue;
+			}
+		} else {
+			h += l;
+			mem = 0;
+			continue;
+		}
+
+		/* Compare right half */
+		for (k=MAX(ms+1,mem); n[k] && n[k] == h[k]; k++);
+		if (n[k]) {
+			h += k-ms;
+			mem = 0;
+			continue;
+		}
+		/* Compare left half */
+		for (k=ms+1; k>mem && n[k-1] == h[k-1]; k--);
+		if (k == mem) return (char *)h;
+		h += p;
+		mem = mem0;
+	}
+}
+
+char *strstr(const char *h, const char *n)
+{
+	/* Return immediately on empty needle */
+	if (!n[0]) return (char *)h;
+
+	/* Use faster algorithms for short needles */
+	h = strchr(h, *n);
+	if (!h || !n[1]) return (char *)h;
+	if (!h[1]) return 0;
+	if (!n[2]) return twobyte_strstr(h, n);
+	if (!h[2]) return 0;
+	if (!n[3]) return threebyte_strstr(h, n);
+	if (!h[3]) return 0;
+	if (!n[4]) return fourbyte_strstr(h, n);
+
+	return twoway_strstr(h, n);
+}

src/string/strtok.c

@@ -0,0 +1,13 @@
+#include <string.h>
+
+char *strtok(char *s, const char *sep)
+{
+	static char *p;
+	if (!s && !(s = p)) return NULL;
+	s += strspn(s, sep);
+	if (!*s) return p = 0;
+	p = s + strcspn(s, sep);
+	if (*p) *p++ = 0;
+	else p = 0;
+	return s;
+}

src/string/strtok_r.c

@@ -0,0 +1,12 @@
+#include <string.h>
+
+char *strtok_r(char *s, const char *sep, char **p)
+{
+	if (!s && !(s = *p)) return NULL;
+	s += strspn(s, sep);
+	if (!*s) return *p = 0;
+	*p = s + strcspn(s, sep);
+	if (**p) *(*p)++ = 0;
+	else *p = 0;
+	return s;
+}

src/string/swab.c

@@ -0,0 +1,13 @@
+#include <unistd.h>
+
+void swab(const void *_src, void *_dest, ssize_t n)
+{
+	const char *src = _src;
+	char *dest = _dest;
+	for (; n>0; n-=2) {
+		dest[0] = src[1];
+		dest[1] = src[0];
+		dest += 2;
+		src += 2;
+	}
+}

src/string/wcscat.c

@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+wchar_t *wcscat(wchar_t *dest, const wchar_t *src)
+{
+	wcscpy(dest + wcslen(dest), src);
+	return dest;
+}

src/string/wcschr.c

@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+wchar_t *wcschr(const wchar_t *s, wchar_t c)
+{
+	if (!c) return (wchar_t *)s + wcslen(s);
+	for (; *s && *s != c; s++);
+	return *s ? (wchar_t *)s : 0;
+}

src/string/wcscmp.c

@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+int wcscmp(const wchar_t *l, const wchar_t *r)
+{
+	for (; *l==*r && *l && *r; l++, r++);
+	return *l - *r;
+}

src/string/wcscpy.c

@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+wchar_t *wcscpy(wchar_t *d, const wchar_t *s)
+{
+	wchar_t *a = d;
+	while ((*d++ = *s++));
+	return a;
+}

src/string/wcscspn.c

@@ -0,0 +1,10 @@
+#include <wchar.h>
+
+size_t wcscspn(const wchar_t *s, const wchar_t *c)
+{
+	const wchar_t *a;
+	if (!c[0]) return wcslen(s);
+	if (!c[1]) return (s=wcschr(a=s, *c)) ? s-a : wcslen(a);
+	for (a=s; *s && !wcschr(c, *s); s++);
+	return s-a;
+}

src/string/wcslen.c

@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+size_t wcslen(const wchar_t *s)
+{
+	const wchar_t *a;
+	for (a=s; *s; s++);
+	return s-a;
+}

src/string/wcsncat.c

@@ -0,0 +1,10 @@
+#include <wchar.h>
+
+wchar_t *wcsncat(wchar_t *d, const wchar_t *s, size_t n)
+{
+	wchar_t *a = d;
+	d += wcslen(d);
+	while (n && (*d++ = *s++)) n--;
+	*d++ = 0;
+	return a;
+}

src/string/wcsncmp.c

@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+int wcsncmp(const wchar_t *l, const wchar_t *r, size_t n)
+{
+	for (; n && *l==*r && *l && *r; l++, r++);
+	return n ? *l - *r : 0;
+}

src/string/wcsncpy.c

@@ -0,0 +1,9 @@
+#include <wchar.h>
+
+wchar_t *wcsncpy(wchar_t *d, const wchar_t *s, size_t n)
+{
+	wchar_t *a = d;
+	while (n && (*d++ = *s++)) n--;
+	wmemset(d, 0, n);
+	return a;
+}

src/string/wcspbrk.c

@@ -0,0 +1,7 @@
+#include <wchar.h>
+
+wchar_t *wcspbrk(const wchar_t *s, const wchar_t *b)
+{
+	s += wcscspn(s, b);
+	return *s ? (wchar_t *)s : NULL;
+}

src/string/wcsrchr.c

@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+wchar_t *wcsrchr(const wchar_t *s, wint_t c)
+{
+	const wchar_t *p;
+	for (p=s+wcslen(s); p>=s && *p!=c; p--);
+	return p>=s ? (wchar_t *)p : 0;
+}

src/string/wcsspn.c

@@ -0,0 +1,8 @@
+#include <wchar.h>
+
+size_t wcsspn(const wchar_t *s, const wchar_t *c)
+{
+	const wchar_t *a;
+	for (a=s; *s && wcschr(c, *s); s++);
+	return s-a;
+}

src/string/wcsstr.c

@@ -0,0 +1,117 @@
+#include <wchar.h>
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+
+static wchar_t *naive_wcsstr(const wchar_t *h, const wchar_t *n)
+{
+	size_t i;
+	for (i=0; n[i] && h[i]; i++)
+	for (   ; n[i] != h[i]; h++, i=0);
+	return n[i] ? 0 : (wchar_t *)h;
+}
+
+#define MAX(a,b) ((a)>(b)?(a):(b))
+#define MIN(a,b) ((a)<(b)?(a):(b))
+
+static wchar_t *twoway_wcsstr(const wchar_t *h, const wchar_t *n)
+{
+	const wchar_t *z;
+	size_t l, ip, jp, k, p, ms, p0, mem, mem0;
+
+	/* Computing length of needle */
+	for (l=0; n[l] && h[l]; l++);
+	if (n[l]) return 0; /* hit the end of h */
+
+	/* Compute maximal suffix */
+	ip = -1; jp = 0; k = p = 1;
+	while (jp+k<l) {
+		if (n[ip+k] == n[jp+k]) {
+			if (k == p) {
+				jp += p;
+				k = 1;
+			} else k++;
+		} else if (n[ip+k] > n[jp+k]) {
+			jp += k;
+			k = 1;
+			p = jp - ip;
+		} else {
+			ip = jp++;
+			k = p = 1;
+		}
+	}
+	ms = ip;
+	p0 = p;
+
+	/* And with the opposite comparison */
+	ip = -1; jp = 0; k = p = 1;
+	while (jp+k<l) {
+		if (n[ip+k] == n[jp+k]) {
+			if (k == p) {
+				jp += p;
+				k = 1;
+			} else k++;
+		} else if (n[ip+k] < n[jp+k]) {
+			jp += k;
+			k = 1;
+			p = jp - ip;
+		} else {
+			ip = jp++;
+			k = p = 1;
+		}
+	}
+	if (ip+1 > ms+1) ms = ip;
+	else p = p0;
+
+	/* Periodic needle? */
+	if (wmemcmp(n, n+p, ms+1)) {
+		mem0 = 0;
+		p = MAX(ms, l-ms-1) + 1;
+	} else mem0 = l-p;
+	mem = 0;
+
+	/* Initialize incremental end-of-haystack pointer */
+	z = h;
+
+	/* Search loop */
+	for (;;) {
+		/* Update incremental end-of-haystack pointer */
+		if (z-h < l) {
+			/* Fast estimate for MIN(l,63) */
+			size_t grow = l | 63;
+			const wchar_t *z2 = wmemchr(z, 0, grow);
+			if (z2) {
+				z = z2;
+				if (z-h < l) return 0;
+			} else z += grow;
+		}
+
+		/* Compare right half */
+		for (k=MAX(ms+1,mem); n[k] && n[k] == h[k]; k++);
+		if (n[k]) {
+			h += k-ms;
+			mem = 0;
+			continue;
+		}
+		/* Compare left half */
+		for (k=ms+1; k>mem && n[k-1] == h[k-1]; k--);
+		if (k == mem) return (wchar_t *)h;
+		h += p;
+		mem = mem0;
+	}
+}
+
+wchar_t *wcsstr(const wchar_t *h, const wchar_t *n)
+{
+	/* Return immediately on empty needle or haystack */
+	if (!n[0]) return (wchar_t *)h;
+	if (!h[0]) return 0;
+
+	/* Use faster algorithms for short needles */
+	h = wcschr(h, *n);
+	if (!h || !n[1]) return (wchar_t *)h;
+	if (!h[1]) return 0;
+	if (!n[2] || !n[3] || !n[4]) return naive_wcsstr(h, n);
+
+	return twoway_wcsstr(h, n);
+}

src/string/wcswcs.c

@@ -0,0 +1,6 @@
+#include <wchar.h>
+
+wchar_t *wcswcs(const wchar_t *haystack, const wchar_t *needle)
+{
+	return wcsstr(haystack, needle);
+}

src/string/wmemchr.c

@@ -0,0 +1,8 @@
+#include <string.h>
+#include <wchar.h>
+
+wchar_t *wmemchr(const wchar_t *s, wchar_t c, size_t n)
+{
+	for (; n && *s != c; s++);
+	return n ? (wchar_t *)s : 0;
+}

src/string/wmemcmp.c

@@ -0,0 +1,8 @@
+#include <string.h>
+#include <wchar.h>
+
+int wmemcmp(const wchar_t *l, const wchar_t *r, size_t n)
+{
+	for (; n && *l==*r; n--, l++, r++);
+	return n ? *l-*r : 0;
+}

src/string/wmemcpy.c

@@ -0,0 +1,9 @@
+#include <string.h>
+#include <wchar.h>
+
+wchar_t *wmemcpy(wchar_t *d, const wchar_t *s, size_t n)
+{
+	wchar_t *a = d;
+	while (n--) *d++ = *s++;
+	return a;
+}

src/string/wmemmove.c

@@ -0,0 +1,11 @@
+#include <string.h>
+#include <wchar.h>
+
+wchar_t *wmemmove(wchar_t *d, const wchar_t *s, size_t n)
+{
+	if ((size_t)(d-s) < n) {
+		while (n--) d[n] = s[n];
+		return d;
+	}
+	return wmemcpy(d, s, n);
+}

src/string/wmemset.c

@@ -0,0 +1,9 @@
+#include <string.h>
+#include <wchar.h>
+
+wchar_t *wmemset(wchar_t *d, wchar_t c, size_t n)
+{
+	wchar_t *ret = d;
+	while (n--) *d++ = c;
+	return ret;
+}