diff options
author | Manuel Novoa III <mjn3@codepoet.org> | 2002-05-23 20:33:53 +0000 |
---|---|---|
committer | Manuel Novoa III <mjn3@codepoet.org> | 2002-05-23 20:33:53 +0000 |
commit | d730fb4d646ac74ee6bbd227411329090fb3206c (patch) | |
tree | 8f843bd6aabc8a9ea20ae708a64177e0c363f37b /libc/misc | |
parent | 50566667587e92dcf900e2d53204d972b5650a38 (diff) |
Multibyte and wide char conversion functions. Some work still to do, but
they're quite solid now and Erik needs them for the gcc port. Comments at
the head of wchar.c.
Diffstat (limited to 'libc/misc')
-rw-r--r-- | libc/misc/wchar/Makefile | 47 | ||||
-rw-r--r-- | libc/misc/wchar/wchar.c | 723 |
2 files changed, 770 insertions, 0 deletions
diff --git a/libc/misc/wchar/Makefile b/libc/misc/wchar/Makefile new file mode 100644 index 000000000..23a1e9bba --- /dev/null +++ b/libc/misc/wchar/Makefile @@ -0,0 +1,47 @@ +# Makefile for uClibc +# +# Copyright (C) 2000 by Lineo, inc. +# Copyright (C) 2000,2001 Erik Andersen <andersen@uclibc.org> +# +# This program is free software; you can redistribute it and/or modify it under +# the terms of the GNU Library General Public License as published by the Free +# Software Foundation; either version 2 of the License, or (at your option) any +# later version. +# +# This program is distributed in the hope that it will be useful, but WITHOUT +# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS +# FOR A PARTICULAR PURPOSE. See the GNU Library General Public License for more +# details. +# +# You should have received a copy of the GNU Library General Public License +# along with this program; if not, write to the Free Software Foundation, Inc., +# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA +# +# Derived in part from the Linux-8086 C library, the GNU C Library, and several +# other sundry sources. Files within this library are copyright by their +# respective copyright holders. + +TOPDIR=../../../ +include $(TOPDIR)Rules.mak + +MSRC= wchar.c +MOBJ= btowc.o wctob.o mbsinit.o mbrlen.o mbrtowc.o wcrtomb.o mbsrtowcs.o \ + wcsrtombs.o _wchar_utf8sntowcs.o _wchar_wcstoutf8s.o \ + __mbsnrtowcs.o __wcsnrtombs.o + +OBJS=$(MOBJ) + +all: $(OBJS) $(LIBC) + +$(LIBC): ar-target + +ar-target: $(OBJS) + $(AR) $(ARFLAGS) $(LIBC) $(OBJS) + +$(MOBJ): $(MSRC) + $(CC) $(CFLAGS) -DL_$* $< -c -o $*.o + $(STRIPTOOL) -x -R .note -R .comment $*.o + +clean: + rm -f *.[oa] *~ core + diff --git a/libc/misc/wchar/wchar.c b/libc/misc/wchar/wchar.c new file mode 100644 index 000000000..f2d9f4a7d --- /dev/null +++ b/libc/misc/wchar/wchar.c @@ -0,0 +1,723 @@ +/* Copyright (C) 2002 Manuel Novoa III + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Library General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Library General Public License for more details. + * + * You should have received a copy of the GNU Library General Public + * License along with this library; if not, write to the Free + * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +/* ATTENTION! ATTENTION! ATTENTION! ATTENTION! ATTENTION! + * + * Besides uClibc, I'm using this code in my libc for elks, which is + * a 16-bit environment with a fairly limited compiler. It would make + * things much easier for me if this file isn't modified unnecessarily. + * In particular, please put any new or replacement functions somewhere + * else, and modify the makefile to use your version instead. + * Thanks. Manuel + * + * ATTENTION! ATTENTION! ATTENTION! ATTENTION! ATTENTION! */ + + +/* May 23, 2002 Initial Notes: + * + * I'm still tweaking this stuff, but it passes the tests I've thrown + * at it, and Erik needs it for the gcc port. The glibc extension + * __wcsnrtombs() hasn't been tested, as I didn't find a test for it + * in the glibc source. I also need to fix the behavior of + * _wchar_utf8sntowcs() if the max number of wchars to convert is 0. + * + * UTF-8 -> wchar -> UTF-8 conversion tests on Markus Kuhn's UTF-8-demo.txt + * file on my platform (x86) show about 5-10% faster conversion speed than + * glibc with mbsrtowcs()/wcsrtombs() and almost twice as fast as glibc with + * individual mbrtowc()/wcrtomb() calls. + * + * If 'DECODER' is defined, then _wchar_utf8sntowcs() will be compiled + * as a fail-safe UTF-8 decoder appropriate for a terminal, etc. which + * needs to deal gracefully with whatever is sent to it. In that mode, + * it passes Markus Kuhn's UTF-8-test.txt stress test. I plan to add + * an arg to force that behavior, so the interface will be changing. + * + * I need to fix the error checking for 16-bit wide chars. This isn't + * an issue for uClibc, but may be for ELKS. I'm currently not sure + * if I'll use 16-bit, 32-bit, or configureable wchars in ELKS. + * + * Manuel + */ + +#define _GNU_SOURCE +#define _ISOC99_SOURCE +#include <errno.h> +#include <stddef.h> +#include <limits.h> +#include <stdint.h> +#include <inttypes.h> +#include <stdlib.h> +#include <stdio.h> +#include <assert.h> +#include <locale.h> +#include <wchar.h> + +#define ENCODING (__global_locale.encoding) + +#if WCHAR_MAX > 0xffffU +#define UTF_8_MAX_LEN 6 +#else +#define UTF_8_MAX_LEN 3 +#endif + +/* #define KUHN */ + +#warning implement __CTYPE_HAS_UTF_8_LOCALES! +#define __CTYPE_HAS_UTF_8_LOCALES + +/* Implementation-specific work functions. */ + +extern size_t _wchar_utf8sntowcs(wchar_t *__restrict pwc, size_t wn, + const char **__restrict src, size_t n, + mbstate_t *ps, int allow_continuation); + +extern size_t _wchar_wcsntoutf8s(char *__restrict s, size_t n, + const wchar_t **__restrict src, size_t wn); + +/* glibc extensions. */ + +extern size_t __mbsnrtowcs(wchar_t *__restrict dst, + const char **__restrict src, + size_t NMC, size_t len, mbstate_t *__restrict ps); + +extern size_t __wcsnrtombs(char *__restrict dst, + const wchar_t **__restrict src, + size_t NWC, size_t len, mbstate_t *__restrict ps); + +/**********************************************************************/ +#ifdef L_btowc + +wint_t btowc(int c) +{ + wchar_t wc; + unsigned char buf[1]; + mbstate_t mbstate; + + if (c != EOF) { + *buf = (unsigned char) c; + mbstate.mask = 0; /* Initialize the mbstate. */ + if (mbrtowc(&wc, buf, 1, &mbstate) == 1) { + return wc; + } + } + return WEOF; +} + +#endif +/**********************************************************************/ +#ifdef L_wctob + +/* Note: We completely ignore ps in all currently supported conversions. */ + +int wctob(wint_t c) +{ + unsigned char buf[MB_LEN_MAX]; + + return (wcrtomb(buf, c, NULL) == 1) ? *buf : EOF; +} + +#endif +/**********************************************************************/ +#ifdef L_mbsinit + +int mbsinit(const mbstate_t *ps) +{ + return !ps || !ps->mask; +} + +#endif +/**********************************************************************/ +#ifdef L_mbrlen + +size_t mbrlen(const char *__restrict s, size_t n, mbstate_t *__restrict ps) +{ + static mbstate_t mbstate; /* Rely on bss 0-init. */ + + return mbrtowc(NULL, s, n, (ps != NULL) ? ps : &mbstate); +} + +#endif +/**********************************************************************/ +#ifdef L_mbrtowc + +size_t mbrtowc(wchar_t *__restrict pwc, const char *__restrict s, + size_t n, mbstate_t *__restrict ps) +{ + static mbstate_t mbstate; /* Rely on bss 0-init. */ + wchar_t wcbuf[1]; + const char *p; + size_t r; + char empty_string[1]; /* Avoid static to be fPIC friendly. */ + + if (!ps) { + ps = &mbstate; + } + + if (!s) { + pwc = (wchar_t *) s; /* NULL */ + empty_string[0] = 0; /* Init the empty string when necessary. */ + s = empty_string; + n = 1; + } else if (!n) { + return (ps->mask && (ps->wc == 0xffffU)) /* TODO: change error code? */ + ? ((size_t) -1) : ((size_t) -2); + } + + p = s; + +#ifdef __CTYPE_HAS_UTF_8_LOCALES + /* Need to do this here since mbsrtowcs doesn't allow incompletes. */ + if (ENCODING == __ctype_encoding_utf8) { + r = _wchar_utf8sntowcs(pwc, 1, &p, n, ps, 1); + return (r == 1) ? (p-s) : r; + } +#endif + + r = __mbsnrtowcs(wcbuf, &p, SIZE_MAX, 1, ps); + + if (((ssize_t) r) >= 0) { + if (pwc) { + *pwc = *wcbuf; + } + } + return (size_t) r; +} + +#endif +/**********************************************************************/ +#ifdef L_wcrtomb + +/* Note: We completely ignore ps in all currently supported conversions. */ +/* TODO: Check for valid state anyway? */ + +size_t wcrtomb(register char *__restrict s, wchar_t wc, + mbstate_t *__restrict ps) +{ + wchar_t wcbuf[2]; + const wchar_t *pwc; + size_t r; + char buf[MB_LEN_MAX]; + + if (!s) { + s = buf; + wc = 0; + } + + pwc = wcbuf; + wcbuf[0] = wc; + wcbuf[1] = 0; + + r = __wcsnrtombs(s, &pwc, SIZE_MAX, MB_LEN_MAX, ps); + return (r != 0) ? r : 1; +} + +#endif +/**********************************************************************/ +#ifdef L_mbsrtowcs + +size_t mbsrtowcs(wchar_t *__restrict dst, const char **__restrict src, + size_t len, mbstate_t *__restrict ps) +{ + static mbstate_t mbstate; /* Rely on bss 0-init. */ + + return __mbsnrtowcs(dst, src, SIZE_MAX, len, + ((ps != NULL) ? ps : &mbstate)); +} + +#endif +/**********************************************************************/ +#ifdef L_wcsrtombs + +/* Note: We completely ignore ps in all currently supported conversions. + + * TODO: Check for valid state anyway? */ + +size_t wcsrtombs(char *__restrict dst, const wchar_t **__restrict src, + size_t len, mbstate_t *__restrict ps) +{ + return __wcsnrtombs(dst, src, SIZE_MAX, len, ps); +} + +#endif +/**********************************************************************/ +#ifdef L__wchar_utf8sntowcs + +/* Define DECODER to generate a UTF-8 decoder which passes Markus Kuhn's + * UTF-8-test.txt strss test. + */ +/* #define DECODER */ + +#ifdef DECODER +#ifndef KUHN +#define KUHN +#endif +#endif + +size_t _wchar_utf8sntowcs(wchar_t *__restrict pwc, size_t wn, + const char **__restrict src, size_t n, + mbstate_t *ps, int allow_continuation) +{ + register const char *s; + __uwchar_t mask; + __uwchar_t wc; + wchar_t wcbuf[1]; + size_t count; + int incr; + + s = *src; + + assert(s != NULL); + assert(ps != NULL); + + incr = 1; + if (!pwc) { + pwc = wcbuf; + wn = SIZE_MAX; + incr = 0; + } +#warning fix _wchar_utf8sntowcs to allow wn == 0! + assert(wn > 0); /* TODO: fix this!! */ + + count = wn; + + if ((mask = (__uwchar_t) ps->mask) != 0) { /* A continuation... */ +#ifdef DECODER + wc = (__uwchar_t) ps->wc; + if (n) { + goto CONTINUE; + } + goto DONE; +#else + if ((wc = (__uwchar_t) ps->wc) != 0xffffU) { + /* TODO: change error code here and below? */ + if (n) { + goto CONTINUE; + } + goto DONE; + } + return (size_t) -1; /* We're in an error state. */ +#endif + } + + do { + if (!n) { + goto DONE; + } + --n; + if ((wc = ((unsigned char) *s++)) >= 0x80) { /* Not ASCII... */ + mask = 0x40; +#warning fix range for 16 bit wides + if ( ((unsigned char)(s[-1] - 0xc0)) < (0xfe - 0xc0) ) { + goto START; + } + BAD: +#ifdef DECODER + wc = 0xfffd; + goto COMPLETE; +#else + ps->mask = mask; + ps->wc = 0xffffU; + return (size_t) -1; /* Illegal start byte! */ +#endif + + CONTINUE: + while (n) { + --n; + if ((*s & 0xc0) != 0x80) { + goto BAD; + } + mask <<= 5; + wc <<= 6; + wc += (*s & 0x3f); /* keep seperate for bcc (smaller code) */ + ++s; + START: + wc &= ~(mask << 1); + + if ((wc & mask) == 0) { /* Character completed. */ + if ((mask >>= 5) == 0x40) { + mask += mask; + } + /* Check for invalid sequences (longer than necessary) + * and invalid chars. */ + if ( (wc < mask) /* Sequence not minimal length. */ +#ifdef KUHN +#if UTF_8_MAX_LEN == 3 +#error broken since mask can overflow!! + /* For plane 0, these are the only defined values.*/ + || (wc > 0xfffdU) +#else + /* Note that we don't need to worry about exceeding */ + /* 31 bits as that is the most that UTF-8 provides. */ + || ( ((__uwchar_t)(wc - 0xfffeU)) < 2) +#endif + || ( ((__uwchar_t)(wc - 0xd800U)) < (0xe000U - 0xd800U) ) +#endif /* KUHN */ + ) { + goto BAD; + } + goto COMPLETE; + } + } + /* Character potentially valid but incomplete. */ + if (!allow_continuation) { + if (count != wn) { + return 0; + } + /* NOTE: The following can fail if you allow and then disallow + * continuation!!! */ +#if UTF_8_MAX_LEN == 3 +#error broken since mask can overflow!! +#endif + /* Need to back up... */ + do { + --s; + } while ((mask >>= 5) >= 0x40); + goto DONE; + } + ps->mask = (wchar_t) mask; + ps->wc = (wchar_t) wc; + *src = s; + return (size_t) -2; + } + COMPLETE: + *pwc = wc; + pwc += incr; + + } +#ifdef DECODER + while (--count); +#else + while (wc && --count); + + if (!wc) { + s = NULL; + } +#endif + + DONE: + /* ps->wc is irrelavent here. */ + ps->mask = 0; + if (pwc != wcbuf) { + *src = s; + } + + return wn - count; +} + +#endif +/**********************************************************************/ +#ifdef L__wchar_wcstoutf8s + +size_t _wchar_wcsntoutf8s(char *__restrict s, size_t n, + const wchar_t **__restrict src, size_t wn) +{ + register char *p; + size_t len, t; + __uwchar_t wc; + const __uwchar_t *swc; + int store; + char buf[MB_LEN_MAX]; + char m; + + store = 1; + if (!s) { + s = buf; + n = SIZE_MAX; + store = 0; + } + + t = n; + swc = (const __uwchar_t *) *src; + + assert(swc != NULL); + + while (wn && t) { + wc = *swc; + + *s = wc; + len = 1; + + if (wc >= 0x80) { +#ifdef KUHN + if ( +#if UTF_8_MAX_LEN == 3 + /* For plane 0, these are the only defined values.*/ + /* Note that we don't need to worry about exceeding */ + /* 31 bits as that is the most that UTF-8 provides. */ + (wc > 0xfffdU) +#else + /* UTF_8_MAX_LEN == 6 */ + (wc > 0x7fffffffUL) + || ( ((__uwchar_t)(wc - 0xfffeU)) < 2) +#endif + || ( ((__uwchar_t)(wc - 0xd800U)) < (0xe000U - 0xd800U) ) + ) { + return (size_t) -1; + } +#else /* KUHN */ +#if UTF_8_MAX_LEN != 3 + if (wc > 0x7fffffffUL) { /* Value too large. */ + return (size_t) -1; + } +#endif +#endif /* KUHN */ + + wc >>= 1; + p = s; + do { + ++p; + } while (wc >>= 5); + wc = *swc; + if ((len = p - s) > t) { /* Not enough space. */ + break; + } + + m = 0x80; + while( p>s ) { + m = (m >> 1) | 0x80; + *--p = (wc & 0x3f) | 0x80; + wc >>= 6; + } + *s |= (m << 1); + } else if (wc == 0) { /* End of string. */ + swc = NULL; + break; + } + + ++swc; + --wn; + t -= len; + if (store) { + s += len; + } + } + + *src = (const wchar_t *) swc; + + return n - t; +} + + +#endif +/**********************************************************************/ +#ifdef L___mbsnrtowcs + +/* WARNING: We treat len as SIZE_MAX when dst is NULL! */ + +size_t mbsnrtowcs(wchar_t *__restrict dst, const char **__restrict src, + size_t NMC, size_t len, mbstate_t *__restrict ps) + __attribute__ ((__weak__, __alias__("__mbsnrtowcs"))); + +size_t __mbsnrtowcs(wchar_t *__restrict dst, const char **__restrict src, + size_t NMC, size_t len, mbstate_t *__restrict ps) +{ + static mbstate_t mbstate; /* Rely on bss 0-init. */ + wchar_t wcbuf[1]; + const char *s; + size_t count, r; + int incr; + + if (!ps) { + ps = &mbstate; + } + +#ifdef __CTYPE_HAS_UTF_8_LOCALES + if (ENCODING == __ctype_encoding_utf8) { + return ((r = _wchar_utf8sntowcs(dst, len, src, NMC, ps, 1)) + != (size_t) -2) ? r : 0; + } +#endif + incr = 1; + if (!dst) { + dst = wcbuf; + len = SIZE_MAX; + incr = 0; + } + + /* Since all the following encodings are single-byte encodings... */ + if (len > NMC) { + len = NMC; + } + + count = len; + s = *src; + +#ifdef __CTYPE_HAS_8_BIT_LOCALES + if (ENCODING == __ctype_encoding_8_bit) { + wchar_t wc; + while (count) { + if ((wc = ((unsigned char)(*s))) >= 0x80) { /* Non-ASCII... */ + wc -= 0x80; + wc = __global_locale.tbl8c2wc[ + (__global_locale.idx8c2wc[wc >> Cc2wc_IDX_SHIFT] + << Cc2wc_IDX_SHIFT) + (wc & (Cc2wc_ROW_LEN - 1))]; + if (!wc) { + goto BAD; + } + } else if (!wc) { + s = NULL; + break; + } + ++s; + *dst = wc; + dst += incr; + --count; + } + if (dst != wcbuf) { + *src = s; + } + return len - count; + } +#endif + + assert(ENCODING == __ctype_encoding_7_bit); + + while (count) { + if ((*dst = (unsigned char) *s) == 0) { + s = NULL; + break; + } + if (*dst >= 0x80) { +#ifdef __CTYPE_HAS_8_BIT_LOCALES + BAD: +#endif + __set_errno(EILSEQ); + return (size_t) -1; + } + ++s; + dst += incr; + --count; + } + if (dst != wcbuf) { + *src = s; + } + return len - count; +} + +#endif +/**********************************************************************/ +#ifdef L___wcsnrtombs + +/* WARNING: We treat len as SIZE_MAX when dst is NULL! */ + +/* Note: We completely ignore ps in all currently supported conversions. + * TODO: Check for valid state anyway? */ + +size_t wcsnrtombs(char *__restrict dst, const wchar_t **__restrict src, + size_t NWC, size_t len, mbstate_t *__restrict ps) + __attribute__ ((__weak__, __alias__("__wcsnrtombs"))); + +size_t __wcsnrtombs(char *__restrict dst, const wchar_t **__restrict src, + size_t NWC, size_t len, mbstate_t *__restrict ps) +{ + const __uwchar_t *s; + size_t count; + int incr; + char buf[MB_LEN_MAX]; + +#ifdef __CTYPE_HAS_UTF_8_LOCALES + if (ENCODING == __ctype_encoding_utf8) { + return _wchar_wcsntoutf8s(dst, len, src, NWC); + } +#endif + + incr = 1; + if (!dst) { + dst = buf; + len = SIZE_MAX; + incr = 0; + } + + /* Since all the following encodings are single-byte encodings... */ + if (len > NWC) { + len = NWC; + } + + count = len; + s = (const __uwchar_t *) *src; + +#ifdef __CTYPE_HAS_8_BIT_LOCALES + if (ENCODING == __ctype_encoding_8_bit) { + __uwchar_t wc; + __uwchar_t u; + while (count) { + if ((wc = *s) <= 0x7f) { + if (!(*dst = (unsigned char) wc)) { + s = NULL; + break; + } + } else { + u = 0; + if (wc <= Cwc2c_DOMAIN_MAX) { + u = __global_locale.idx8wc2c[wc >> (Cwc2c_TI_SHIFT + + Cwc2c_TT_SHIFT)]; + u = __global_locale.tbl8wc2c[(u << Cwc2c_TI_SHIFT) + + ((wc >> Cwc2c_TT_SHIFT) + & ((1 << Cwc2c_TI_SHIFT)-1))]; + u = __global_locale.tbl8wc2c[Cwc2c_TI_LEN + + (u << Cwc2c_TT_SHIFT) + + (wc & ((1 << Cwc2c_TT_SHIFT)-1))]; + } + +/* #define __WCHAR_REPLACEMENT_CHAR '?' */ +#ifdef __WCHAR_REPLACEMENT_CHAR + *dst = (unsigned char) ( u ? u : __WCHAR_REPLACEMENT_CHAR ); +#else + if (!u) { + goto BAD; + } + *dst = (unsigned char) u; +#endif + } + ++s; + dst += incr; + --count; + } + if (dst != buf) { + *src = (const wchar_t *) s; + } + return len - count; + } +#endif + + assert(ENCODING == __ctype_encoding_7_bit); + + while (count) { + if (*s >= 0x80) { +#if defined(__CTYPE_HAS_8_BIT_LOCALES) && !defined(__WCHAR_REPLACEMENT_CHAR) + BAD: +#endif + __set_errno(EILSEQ); + return (size_t) -1; + + } + if ((*dst = (unsigned char) *s) == 0) { + s = NULL; + break; + } + ++s; + dst += incr; + --count; + } + if (dst != buf) { + *src = (const wchar_t *) s; + } + return len - count; +} + +#endif +/**********************************************************************/ |