summaryrefslogtreecommitdiff
path: root/libc/string
diff options
context:
space:
mode:
authorManuel Novoa III <mjn3@codepoet.org>2002-12-20 19:26:35 +0000
committerManuel Novoa III <mjn3@codepoet.org>2002-12-20 19:26:35 +0000
commitc9d66e44af5c93a1ea5487fd9bff78274be65850 (patch)
tree975bb662b51ac91f8d1434553b936784b175a802 /libc/string
parentbd6e7b0208c9652717bec30abff2d8dc12eaaca7 (diff)
The big thing is locale dependent collation support.
Also added outdigit support and (legacy) YESSTR/NOSTR support.
Diffstat (limited to 'libc/string')
-rw-r--r--libc/string/Makefile5
-rw-r--r--libc/string/wstring.c684
2 files changed, 650 insertions, 39 deletions
diff --git a/libc/string/Makefile b/libc/string/Makefile
index 088058ebe..fe9b64382 100644
--- a/libc/string/Makefile
+++ b/libc/string/Makefile
@@ -33,13 +33,14 @@ MOBJW= basename.o bcopy.o bzero.o dirname.o ffs.o memccpy.o memchr.o memcmp.o \
strspn.o strstr.o strtok.o strtok_r.o strerror.o _susv3_strerror_r.o \
_string_syserrmsgs.o _glibc_strerror_r.o \
_string_syssigmsgs.o sys_siglist.o strsignal.o psignal.o \
- __xpg_basename.o strlcat.o strlcpy.o sys_errlist.o # strcoll.o
+ __xpg_basename.o strlcat.o strlcpy.o sys_errlist.o
MOBJW2= wcscasecmp.o wcscat.o wcschrnul.o wcschr.o wcscmp.o wcscpy.o wcscspn.o \
wcsdup.o wcslen.o wcsncasecmp.o wcsncat.o wcsncmp.o wcsncpy.o \
wcsnlen.o wcspbrk.o wcsrchr.o wcsspn.o wcsstr.o wcstok.o wmemchr.o \
wmemcmp.o wmemcpy.o wmemmove.o wmempcpy.o wmemset.o wcpcpy.o wcpncpy.o \
- wcsxfrm.o # wcscoll
+ __wcslcpy.o \
+ wcsxfrm.o strxfrm.o # wcscoll strcoll.o
OBJS=$(MOBJ) $(MOBJ1) $(MOBJ2) $(MOBJW)
diff --git a/libc/string/wstring.c b/libc/string/wstring.c
index 531b1c9fd..c1ead6e00 100644
--- a/libc/string/wstring.c
+++ b/libc/string/wstring.c
@@ -26,6 +26,13 @@
*
* ATTENTION! ATTENTION! ATTENTION! ATTENTION! ATTENTION! */
+/* Dec 20, 2002
+ *
+ * Initial test implementation of strcoll, strxfrm, wcscoll, and wcsxfrm.
+ * The code needs to be cleaned up a good bit, but I'd like to see people
+ * test it out.
+ */
+
#define _STDIO_UTILITY
#define _GNU_SOURCE
#include <string.h>
@@ -36,11 +43,12 @@
#include <stdlib.h>
#include <errno.h>
#include <signal.h>
+#include <assert.h>
+#include <locale.h>
#ifdef WANT_WIDE
#include <wchar.h>
#include <wctype.h>
-#include <locale.h>
#define Wvoid wchar_t
#define Wchar wchar_t
@@ -627,17 +635,14 @@ int Wmemcmp(const Wvoid *s1, const Wvoid *s2, size_t n)
#ifdef L_strcmp
+#ifdef __LOCALE_C_ONLY
+#warning c only
#ifdef L_wcscmp
-#ifdef __UCLIBC_MJN3_ONLY__
-#warning implement wcscoll and remove weak alias (or enable for C locale only)
-#endif
weak_alias(wcscmp,wcscoll);
#else /* L_wcscmp */
-#ifdef __UCLIBC_MJN3_ONLY__
-#warning implement strcoll and remove weak alias (or enable for C locale only)
-#endif
weak_alias(strcmp,strcoll);
#endif /* L_wcscmp */
+#endif /* __LOCALE_C_ONLY */
int Wstrcmp(register const Wchar *s1, register const Wchar *s2)
{
@@ -661,23 +666,6 @@ int Wstrcmp(register const Wchar *s1, register const Wchar *s2)
}
#endif
/**********************************************************************/
-#ifdef L_strcoll
-#error implement strcoll and remove weak_alias!!
-
-#if 0
-extern unsigned char *_ctype_collate;
-int strcoll(register const char *s1, const char *s2)
-{
- int r;
-
- while (!(r = (_ctype_collate[(int)(*s1++)]-_ctype_collate[(int)(*s2++)])));
-
- return r;
-}
-#endif
-
-#endif
-/**********************************************************************/
#ifdef L_wcsncmp
#define L_strncmp
#define Wstrncmp wcsncmp
@@ -713,11 +701,6 @@ int Wstrncmp(register const Wchar *s1, register const Wchar *s2, size_t n)
#endif
/**********************************************************************/
-#ifdef L_strxfrm
-#error implement strxfrm
-/* size_t strxfrm(char *dst, const char *src, size_t len); */
-#endif
-/**********************************************************************/
#ifdef L_wmemchr
#define L_memchr
#define Wmemchr wmemchr
@@ -1923,28 +1906,37 @@ size_t strlcat(register char *__restrict dst,
#endif
/**********************************************************************/
-#ifdef L_wcsxfrm
+#ifdef WANT_WIDE
+extern size_t __wcslcpy(wchar_t *__restrict dst,
+ const wchar_t *__restrict src,
+ size_t n);
+#endif
+
+
+#ifdef L___wcslcpy
#define L_strlcpy
-#define Wstrlcpy wcsxfrm
+#define Wstrlcpy __wcslcpy
+#ifdef __LOCALE_C_ONLY
+weak_alias(__wcslcpy,wcsxfrm);
+#endif
#endif
#ifdef L_strlcpy
-#ifndef L_wcsxfrm
+#ifndef L___wcslcpy
#define Wstrlcpy strlcpy
-#ifdef __UCLIBC_MJN3_ONLY__
-#warning implement wcscoll and remove weak alias (or enable for C locale only)
-#endif
+#ifdef __LOCALE_C_ONLY
weak_alias(strlcpy,strxfrm);
#endif
+#endif
/* OpenBSD function:
* Copy at most n-1 chars from src to dst and nul-terminate dst.
* Returns strlen(src), so truncation occurred if the return value is >= n. */
size_t Wstrlcpy(register Wchar *__restrict dst,
- register const Wchar *__restrict src,
- size_t n)
+ register const Wchar *__restrict src,
+ size_t n)
{
const Wchar *src0 = src;
Wchar dummy[1];
@@ -2145,3 +2137,621 @@ void psignal(int signum, register const char *message)
#endif
/**********************************************************************/
+#ifndef __LOCALE_C_ONLY
+
+#ifdef L_strxfrm
+#ifndef WANT_WIDE
+#error WANT_WIDE should be defined for L_strxfrm
+#endif
+#ifdef L_wcsxfrm
+#error L_wcsxfrm already defined for L_strxfrm
+#endif
+
+#define wcscoll strcoll
+#define L_wcsxfrm
+#undef WANT_WIDE
+
+#undef Wvoid
+#undef Wchar
+#undef Wuchar
+#undef Wint
+
+#define Wchar char
+
+#endif /* L_strxfrm */
+
+
+
+#ifdef L_wcsxfrm
+
+#define CUR_COLLATE (&__global_locale.collate)
+
+#define MAX_PENDING 8
+
+typedef struct {
+ const Wchar *s;
+ const Wchar *eob; /* end of backward */
+
+ __uwchar_t weight;
+ __uwchar_t ui_weight; /* undefined or invalid */
+ int colitem;
+ int weightidx;
+ int rule;
+ size_t position;
+ /* should be wchar_t. if wchar < 0 do EILSEQ? */
+ __uwchar_t *cip;
+ __uwchar_t ci_pending[MAX_PENDING]; /* nul-terminated */
+
+ char *back_buf;
+ char *bbe; /* end of back_buf (actual last... not 1 past end) */
+ char *bp; /* ptr into backbuf, NULL if not in backward mode */
+ char ibb[128];
+ size_t bb_size;
+
+ int ru_pushed;
+} col_state_t;
+
+
+#define WEIGHT_MASK 0x3fffU
+#define RULE_MASK 0xc000U
+
+#define RULE_FORWARD (1 << 14)
+#define RULE_POSITION (1 << 15)
+
+#define UI_IDX (WEIGHT_MASK-6)
+#define POSIT_IDX (WEIGHT_MASK-5)
+#define RANGE_IDX (WEIGHT_MASK-4)
+#define UNDEF_IDX (WEIGHT_MASK-3)
+#define INVAL_IDX (WEIGHT_MASK-2)
+#define DITTO_IDX (WEIGHT_MASK-1)
+
+
+#undef TRACE
+#if 0
+#define TRACE(X) printf##X
+#else
+#define TRACE(X) ((void)0)
+#endif
+
+static int lookup(wchar_t wc)
+{
+ unsigned int sc, n, i0, i1;
+
+ if (((__uwchar_t) wc) > 0xffffU) {
+ return 0;
+ }
+
+ sc = wc & CUR_COLLATE->ti_mask;
+ wc >>= CUR_COLLATE->ti_shift;
+ n = wc & CUR_COLLATE->ii_mask;
+ wc >>= CUR_COLLATE->ii_shift;
+
+ i0 = CUR_COLLATE->wcs2colidt_tbl[wc];
+ i0 <<= CUR_COLLATE->ii_shift;
+ i1 = CUR_COLLATE->wcs2colidt_tbl[CUR_COLLATE->ii_len + i0 + n];
+ i1 <<= CUR_COLLATE->ti_shift;
+ return CUR_COLLATE->wcs2colidt_tbl[CUR_COLLATE->ii_len + CUR_COLLATE->ti_len + i1 + sc];
+
+}
+
+static void init_col_state(col_state_t *cs, const Wchar *wcs)
+{
+ memset(cs, 0, sizeof(col_state_t));
+ cs->s = wcs;
+ cs->bp = cs->back_buf = cs->ibb;
+ cs->bb_size = 128;
+ cs->bbe = cs->back_buf + (cs->bb_size -1);
+}
+
+static void next_weight(col_state_t *cs, int pass)
+{
+ int r, w, ru, ri, popping_backup_stack;
+ ssize_t n;
+ const uint16_t *p;
+#ifdef WANT_WIDE
+#define WC (*cs->s)
+#define N (1)
+#else /* WANT_WIDE */
+ mbstate_t mbstate;
+ wchar_t WC;
+ size_t n0, nx;
+#define N n0
+
+ mbstate.mask = 0;
+#endif /* WANT_WIDE */
+
+ do {
+
+ if (cs->ru_pushed) {
+ ru = cs->ru_pushed;
+ TRACE(("ru_pushed = %d\n", ru));
+ cs->ru_pushed = 0;
+ goto POSITION_SKIP;
+ }
+
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning should we walk pendings backwards?
+#endif
+ if (cs->cip) { /* possible pending weight */
+ if ((r = *(cs->cip++)) == 0) {
+ cs->cip = NULL;
+ continue;
+ }
+ cs->weightidx = r & WEIGHT_MASK;
+ assert(cs->weightidx);
+/* assert(cs->weightidx != WEIGHT_MASK); */
+ } else { /* get the next collation item from the string */
+ TRACE(("clearing popping flag\n"));
+ popping_backup_stack = 0;
+
+ IGNORE_LOOP:
+ /* keep first pos as 0 for a sentinal */
+ if (*cs->bp) { /* pending backward chars */
+ POP_BACKUP:
+ popping_backup_stack = 1;
+ TRACE(("setting popping flag\n"));
+ n = 0;
+ if (*cs->bp > 0) { /* singles pending */
+ cs->s -= 1;
+ if ((*cs->bp -= 1) == 0) {
+ cs->bp -= 1;
+ }
+ } else { /* last was a multi */
+ cs->s += *cs->bp;
+ cs->bp -= 1;
+ }
+ } else if (!*cs->s) { /* not in backward mode and end of string */
+ cs->weight = 0;
+ return;
+ } else {
+ cs->position += 1;
+ }
+
+ BACK_LOOP:
+#ifdef WANT_WIDE
+ n = 1;
+ cs->colitem = r = lookup(*cs->s);
+#else /* WANT_WIDE */
+ n = n0 = mbrtowc(&WC, cs->s, SIZE_MAX, &mbstate);
+ if (n < 0) {
+ __set_errno(EILSEQ);
+ cs->weight = 0;
+ return;
+ }
+ cs->colitem = r = lookup(WC);
+#endif /* WANT_WIDE */
+
+ TRACE((" r=%d WC=%#lx\n", r, (unsigned long)(WC)));
+
+ if (r > CUR_COLLATE->max_col_index) { /* starting char for one or more sequences */
+ p = CUR_COLLATE->multistart_tbl;
+ p += p[r-CUR_COLLATE->max_col_index -1];
+ do {
+ n = N;
+ r = *p++;
+ do {
+ if (!*p) { /* found it */
+ cs->colitem = r;
+ TRACE((" found multi %d\n", n));
+ goto FOUND;
+ }
+#ifdef WANT_WIDE
+ /* the lookup check here is safe since we're assured that *p is a valid colidx */
+ if (!cs->s[n] || (lookup(cs->s[n]) != *p)) {
+ do {} while (*p++);
+ break;
+ }
+ ++p;
+ ++n;
+#else /* WANT_WIDE */
+ if (cs->s[n]) {
+ nx = mbrtowc(&WC, cs->s + n, SIZE_MAX, &mbstate);
+ if (nx < 0) {
+ __set_errno(EILSEQ);
+ cs->weight = 0;
+ return;
+ }
+ }
+ if (!cs->s[n] || (lookup(WC) != *p)) {
+ do {} while (*p++);
+ break;
+ }
+ ++p;
+ n += nx;
+#endif /* WANT_WIDE */
+ } while (1);
+ } while (1);
+ } else if (r == 0) { /* illegal, undefined, or part of a range */
+ if ((CUR_COLLATE->range_count)
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning .. need to introduce range as a collating item?
+#endif
+ && (((__uwchar_t)(WC - CUR_COLLATE->range_low)) <= CUR_COLLATE->range_count)
+ ) { /* part of a range */
+ /* Note: cs->colitem = 0 already. */
+ TRACE((" found range\n"));
+ ru = CUR_COLLATE->ruletable[CUR_COLLATE->range_rule_offset*CUR_COLLATE->MAX_WEIGHTS + pass];
+ assert((ru & WEIGHT_MASK) != DITTO_IDX);
+ if ((ru & WEIGHT_MASK) == WEIGHT_MASK) {
+ ru = (ru & RULE_MASK) | RANGE_IDX;
+ cs->weight = CUR_COLLATE->range_base_weight + (WC - CUR_COLLATE->range_low);
+ }
+ goto RANGE_SKIP_TO;
+ } else if (((__uwchar_t)(WC)) <= 0x7fffffffUL) { /* legal but undefined */
+ UNDEFINED:
+ /* Note: cs->colitem = 0 already. */
+ ri = CUR_COLLATE->undefined_idx;
+ assert(ri != 0); /* implicit undefined isn't supported */
+
+ TRACE((" found explicit UNDEFINED\n"));
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning right now single weight locales do not support ..
+#endif
+ if (CUR_COLLATE->num_weights == 1) {
+ TRACE((" single weight UNDEFINED\n"));
+ cs->weightidx = RANGE_IDX;
+ cs->weight = ri;
+ cs->s += n;
+ goto PROCESS_WEIGHT;
+ }
+
+ ri = CUR_COLLATE->index2ruleidx[ri - 1];
+ ru = CUR_COLLATE->ruletable[ri * CUR_COLLATE->MAX_WEIGHTS + pass];
+ assert((ru & WEIGHT_MASK) != WEIGHT_MASK); /* TODO: handle ".." */
+ if ((ru & WEIGHT_MASK) == DITTO_IDX) {
+ cs->colitem = CUR_COLLATE->undefined_idx;
+ }
+ goto RANGE_SKIP_TO;
+ } else { /* illegal */
+ TRACE((" found illegal\n"));
+ __set_errno(EINVAL);
+ /* We put all illegals in the same equiv class with maximal weight,
+ * and ignore them after the first pass. */
+ if (pass > 0) {
+ cs->s += n;
+ goto IGNORE_LOOP;
+ }
+ ru = (RULE_FORWARD | RANGE_IDX);
+ cs->weight = 0xffffU;
+ goto RANGE_SKIP_TO;
+ }
+ } else if (CUR_COLLATE->num_weights == 1) {
+ TRACE((" single weight\n"));
+ cs->weightidx = RANGE_IDX;
+ cs->weight = cs->colitem;
+ cs->s += n;
+ goto PROCESS_WEIGHT;
+ } else {
+ TRACE((" normal\n"));
+ }
+
+ /* if we get here, it is a normal char either singlely weighted, undefined, or in a range */
+ FOUND:
+ ri = CUR_COLLATE->index2ruleidx[cs->colitem - 1];
+ TRACE((" ri=%d ", ri));
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning make sure this is correct
+#endif
+ if (!ri) {
+ TRACE(("NOT IN THIS LOCALE\n"));
+ goto UNDEFINED;
+ }
+ ru = CUR_COLLATE->ruletable[ri * CUR_COLLATE->MAX_WEIGHTS + pass];
+
+ RANGE_SKIP_TO:
+
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning ignoreables probably should not interrupt backwards processing, but this is wrong
+#endif
+/* if (!(ru & WEIGHT_MASK)) { */
+/* TRACE(("IGNORE\n")); */
+/* cs->s += n; */
+/* continue; */
+/* } */
+
+
+ TRACE((" rule = %#x weight = %#x popping = %d s = %p eob = %p\n",
+ ru & RULE_MASK, ru & WEIGHT_MASK, popping_backup_stack,
+ cs->s, cs->eob));
+ /* now we need to check if we're going backwards... */
+
+ if (!popping_backup_stack) {
+ if (!(ru & RULE_MASK)) { /* backward */
+ TRACE(("backwards\n"));
+ assert(cs->bp <= cs->bbe);
+ if (cs->bp == cs->bbe) {
+ if (cs->back_buf == cs->ibb) { /* was using internal buffer */
+ cs->bp = malloc(cs->bb_size + 128);
+ if (!cs->bp) {
+ __set_errno(ENOMEM);
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning what to do here?
+#endif
+ cs->weight = 0;
+ return;
+ }
+ memcpy(cs->bp, cs->back_buf, cs->bb_size);
+
+ } else {
+ cs->bp = realloc(cs->back_buf, cs->bb_size + 128);
+ if (!cs->bp) {
+ __set_errno(ENOMEM);
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning what to do here?
+#endif
+ cs->weight = 0;
+ return;
+ }
+ }
+ cs->bb_size += 128;
+ cs->bbe = cs->bp + (cs->bbe - cs->back_buf);
+ cs->back_buf = cs->bp;
+ cs->bp = cs->bbe;
+
+ }
+ if (n==1) { /* single char */
+ if (*cs->bp && (((unsigned char)(*cs->bp)) < CHAR_MAX)) {
+ *cs->bp += 1; /* increment last single's count */
+ } else { /* last was a multi, or just starting */
+ if (!cs->bp) {
+ cs->bp = cs->back_buf;
+ } else {
+ assert(cs->bp < cs->bbe);
+ ++cs->bp;
+ }
+ *cs->bp = 1;
+ }
+ } else { /* multichar */
+ assert(n>1);
+ assert(cs->bp < cs->bbe);
+ *++cs->bp = -n;
+ }
+ cs->s += n;
+ if (*cs->s) {
+ goto BACK_LOOP;
+ }
+ /* end-of-string so start popping */
+ cs->eob = cs->s;
+ TRACE(("popping\n"));
+ goto POP_BACKUP;
+ } else if (*cs->bp) { /* was going backward but this element isn't */
+ /* discard current and use previous backward element */
+ assert(!cs->cip);
+ cs->eob = cs->s;
+ TRACE(("popping\n"));
+ goto POP_BACKUP;
+ } else { /* was and still going forward */
+ TRACE(("forwards\n"));
+ if ((ru & (RULE_POSITION|WEIGHT_MASK)) > RULE_POSITION) {
+ assert(ru & WEIGHT_MASK);
+ cs->ru_pushed = ru;
+ cs->weight = cs->position;
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning devel code
+#endif
+ cs->position = 0; /* reset to reduce size for strcoll? */
+ cs->s += n;
+ cs->weightidx = RANGE_IDX;
+ goto PROCESS_WEIGHT;
+ }
+ }
+ } else { /* popping backwards stack */
+ TRACE(("popping (continued)\n"));
+ if (!*cs->bp) {
+ cs->s = cs->eob;
+ }
+ cs->s -= n;
+ }
+
+ cs->s += n;
+ POSITION_SKIP:
+ cs->weightidx = ru & WEIGHT_MASK;
+ cs->rule = ru & RULE_MASK;
+ }
+
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning for pending we only want the weight... _not_ the rule
+#endif
+ if (!cs->weightidx) { /* ignore */
+ continue;
+ }
+
+ PROCESS_WEIGHT:
+ assert(cs->weightidx);
+
+
+ if (((unsigned int)(cs->weightidx - UI_IDX)) <= (INVAL_IDX-UI_IDX)) {
+ if (cs->weightidx == UI_IDX) {
+ cs->weight = cs->ui_weight;
+ }
+ return;
+ }
+
+ assert(cs->weightidx != WEIGHT_MASK);
+ if (cs->weightidx == DITTO_IDX) { /* want the weight of the current collating item */
+ TRACE(("doing ditto\n"));
+ w = CUR_COLLATE->index2weight[cs->colitem -1];
+ } else if (cs->weightidx <= CUR_COLLATE->max_col_index) { /* normal */
+ TRACE(("doing normal\n"));
+ w = CUR_COLLATE->index2weight[cs->weightidx -1];
+ } else { /* a string */
+ TRACE(("doing string\n"));
+ assert(!(cs->weightidx & RULE_MASK));
+ /* note: iso14561 allows null string here */
+ p = CUR_COLLATE->weightstr + (cs->weightidx - (CUR_COLLATE->max_col_index + 2));
+ if (*p & WEIGHT_MASK) {
+ r = 0;
+ do {
+ assert(r < MAX_PENDING);
+ cs->ci_pending[r++] = *p++;
+ } while (*p & WEIGHT_MASK);
+ cs->cip = cs->ci_pending;
+ }
+ continue;
+ }
+
+ cs->weight = w;
+ return;
+ } while (1);
+}
+
+int wcscoll (const Wchar *s0, const Wchar *s1)
+{
+ col_state_t ws[2];
+ int pass;
+
+ if (!CUR_COLLATE->num_weights) { /* C locale */
+#ifdef WANT_WIDE
+ return wcscmp(s0, s1);
+#else /* WANT_WIDE */
+ return strcmp(s0, s1);
+#endif /* WANT_WIDE */
+ }
+
+ pass = 0;
+ do { /* loop through the weights levels */
+ init_col_state(ws, s0);
+ init_col_state(ws+1, s1);
+ do { /* loop through the strings */
+ /* for each string, get the next weight */
+ next_weight(ws, pass);
+ next_weight(ws+1, pass);
+ TRACE(("w0=%lu w1=%lu\n",
+ (unsigned long) ws[0].weight,
+ (unsigned long) ws[1].weight));
+
+ if (ws[0].weight != ws[1].weight) {
+ return ws[0].weight - ws[1].weight;
+ }
+ } while (ws[0].weight);
+ } while (++pass < CUR_COLLATE->num_weights);
+
+ return 0;
+}
+
+#ifdef WANT_WIDE
+
+size_t wcsxfrm(wchar_t *__restrict ws1, const wchar_t *__restrict ws2, size_t n)
+{
+ col_state_t cs;
+ size_t count;
+ int pass;
+
+ if (!CUR_COLLATE->num_weights) { /* C locale */
+ return __wcslcpy(ws1, ws2, n);
+ }
+
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning handle empty string as a special case
+#endif
+
+ count = pass = 0;
+ do { /* loop through the weights levels */
+ init_col_state(&cs, ws2);
+ do { /* loop through the string */
+ next_weight(&cs, pass);
+ TRACE(("weight=%lu (%#lx)\n", (unsigned long) cs.weight, (unsigned long) cs.weight));
+ if (count < n) {
+ ws1[count] = cs.weight +1;
+ }
+ ++count;
+ TRACE(("--------------------------------------------\n"));
+ } while (cs.weight);
+ if (count <= n) { /* overwrite the trailing 0 end-of-pass marker */
+ ws1[count-1] = 1;
+ }
+ TRACE(("-------------------- pass %d --------------------\n", pass));
+ } while (++pass < CUR_COLLATE->num_weights);
+ if (count <= n) { /* oops... change it back */
+ ws1[count-1] = 0;
+ }
+ return count-1;
+}
+
+#else /* WANT_WIDE */
+
+static const unsigned long bound[] = {
+ 1UL << 7,
+ 1UL << 11,
+ 1UL << 16,
+ 1UL << 21,
+ 1UL << 26,
+};
+
+static unsigned char first[] = {
+ 0x0, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc
+};
+
+/* Use an extension of UTF-8 to store a 32 bit val in max 6 bytes. */
+
+static size_t store(unsigned char *s, size_t count, size_t n, __uwchar_t weight)
+{
+ int i, r;
+
+ i = 0;
+ do {
+ if (weight < bound[i]) {
+ break;
+ }
+ } while (++i < sizeof(bound)/sizeof(bound[0]));
+
+ r = i+1;
+ if (i + count < n) {
+ s += count;
+ s[0] = first[i];
+ while (i) {
+ s[i] = 0x80 | (weight & 0x3f);
+ weight >>= 6;
+ --i;
+ }
+ s[0] |= weight;
+ }
+
+ return r;
+}
+
+size_t strxfrm(char *__restrict ws1, const char *__restrict ws2, size_t n)
+{
+ col_state_t cs;
+ size_t count, inc;
+ int pass;
+
+ if (!CUR_COLLATE->num_weights) { /* C locale */
+ return strlcpy(ws1, ws2, n);
+ }
+
+#ifdef __UCLIBC_MJN3_ONLY__
+#warning handle empty string as a special case
+#endif
+
+ inc = count = pass = 0;
+ do { /* loop through the weights levels */
+ init_col_state(&cs, ws2);
+ do { /* loop through the string */
+ next_weight(&cs, pass);
+ TRACE(("weight=%lu (%#lx)\n", (unsigned long) cs.weight, (unsigned long) cs.weight));
+ inc = store((unsigned char *)ws1, count, n, cs.weight + 1);
+ count += inc;
+ TRACE(("--------------------------------------------\n"));
+ } while (cs.weight);
+ /* overwrite the trailing 0 end-of-pass marker */
+ assert(inc == 1);
+ if (count <= n) {
+ ws1[count-1] = 1;
+ }
+ TRACE(("-------------------- pass %d --------------------\n", pass));
+ } while (++pass < CUR_COLLATE->num_weights);
+ if (count <= n) { /* oops... change it back */
+ ws1[count-1] = 0;
+ }
+ return count-1;
+}
+
+
+#endif /* WANT_WIDE */
+
+#endif /* wcscoll */
+
+#endif /* __LOCALE_C_ONLY */
+/**********************************************************************/
+