/* Convert multibyte character to wide character.
- Copyright (C) 1999-2002, 2005-2016 Free Software Foundation, Inc.
+ Copyright (C) 1999-2002, 2005-2019 Free Software Foundation, Inc.
Written by Bruno Haible <bruno@clisp.org>, 2008.
This program is free software: you can redistribute it and/or modify
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
- along with this program. If not, see <http://www.gnu.org/licenses/>. */
+ along with this program. If not, see <https://www.gnu.org/licenses/>. */
#include <config.h>
# include "localcharset.h"
# include "streq.h"
# include "verify.h"
+# include "glthread/lock.h"
+
+# ifndef FALLTHROUGH
+# if __GNUC__ < 7
+# define FALLTHROUGH ((void) 0)
+# else
+# define FALLTHROUGH __attribute__ ((__fallthrough__))
+# endif
+# endif
+
+/* Returns a classification of special values of the encoding of the current
+ locale. */
+typedef enum {
+ enc_other, /* other */
+ enc_utf8, /* UTF-8 */
+ enc_eucjp, /* EUC-JP */
+ enc_94, /* EUC-KR, GB2312, BIG5 */
+ enc_euctw, /* EUC-TW */
+ enc_gb18030, /* GB18030 */
+ enc_sjis /* SJIS */
+} enc_t;
+static inline enc_t
+locale_enc (void)
+{
+ const char *encoding = locale_charset ();
+ if (STREQ_OPT (encoding, "UTF-8", 'U', 'T', 'F', '-', '8', 0, 0, 0, 0))
+ return enc_utf8;
+ if (STREQ_OPT (encoding, "EUC-JP", 'E', 'U', 'C', '-', 'J', 'P', 0, 0, 0))
+ return enc_eucjp;
+ if (STREQ_OPT (encoding, "EUC-KR", 'E', 'U', 'C', '-', 'K', 'R', 0, 0, 0)
+ || STREQ_OPT (encoding, "GB2312", 'G', 'B', '2', '3', '1', '2', 0, 0, 0)
+ || STREQ_OPT (encoding, "BIG5", 'B', 'I', 'G', '5', 0, 0, 0, 0, 0))
+ return enc_94;
+ if (STREQ_OPT (encoding, "EUC-TW", 'E', 'U', 'C', '-', 'T', 'W', 0, 0, 0))
+ return enc_euctw;
+ if (STREQ_OPT (encoding, "GB18030", 'G', 'B', '1', '8', '0', '3', '0', 0, 0))
+ return enc_gb18030;
+ if (STREQ_OPT (encoding, "SJIS", 'S', 'J', 'I', 'S', 0, 0, 0, 0, 0))
+ return enc_sjis;
+ return enc_other;
+}
+
+# if GNULIB_WCHAR_SINGLE
+/* When we know that the locale does not change, provide a speedup by
+ caching the value of locale_enc. */
+static int cached_locale_enc = -1;
+static inline enc_t
+locale_enc_cached (void)
+{
+ if (cached_locale_enc < 0)
+ cached_locale_enc = locale_enc ();
+ return cached_locale_enc;
+}
+# else
+/* By default, don't make assumptions, hence no caching. */
+# define locale_enc_cached locale_enc
+# endif
+/* This lock protects the internal state of mbtowc against multiple simultaneous
+ calls of mbrtowc. */
+gl_lock_define_initialized(static, mbtowc_lock)
verify (sizeof (mbstate_t) >= 4);
char buf[4];
const char *p;
size_t m;
+ enc_t enc;
+ int res;
switch (nstate)
{
break;
case 3:
buf[2] = pstate[3];
- /*FALLTHROUGH*/
+ FALLTHROUGH;
case 2:
buf[1] = pstate[2];
- /*FALLTHROUGH*/
+ FALLTHROUGH;
case 1:
buf[0] = pstate[1];
p = buf;
/* Here m > 0. */
-# if __GLIBC__ || defined __UCLIBC__
- /* Work around bug <http://sourceware.org/bugzilla/show_bug.cgi?id=9674> */
- mbtowc (NULL, NULL, 0);
-# endif
- {
- int res = mbtowc (pwc, p, m);
-
- if (res >= 0)
- {
- if (pwc != NULL && ((*pwc == 0) != (res == 0)))
- abort ();
- if (nstate >= (res > 0 ? res : 1))
- abort ();
- res -= nstate;
- pstate[0] = 0;
- return res;
- }
+ enc = locale_enc_cached ();
- /* mbtowc does not distinguish between invalid and incomplete multibyte
- sequences. But mbrtowc needs to make this distinction.
- There are two possible approaches:
- - Use iconv() and its return value.
- - Use built-in knowledge about the possible encodings.
- Given the low quality of implementation of iconv() on the systems that
- lack mbrtowc(), we use the second approach.
- The possible encodings are:
- - 8-bit encodings,
- - EUC-JP, EUC-KR, GB2312, EUC-TW, BIG5, GB18030, SJIS,
- - UTF-8.
- Use specialized code for each. */
- if (m >= 4 || m >= MB_CUR_MAX)
- goto invalid;
- /* Here MB_CUR_MAX > 1 and 0 < m < 4. */
+ if (enc == enc_utf8) /* UTF-8 */
{
- const char *encoding = locale_charset ();
+ /* Achieve multi-thread safety by not calling mbtowc() at all. */
+ /* Cf. unistr/u8-mbtouc.c. */
+ unsigned char c = (unsigned char) p[0];
- if (STREQ_OPT (encoding, "UTF-8", 'U', 'T', 'F', '-', '8', 0, 0, 0, 0))
+ if (c < 0x80)
{
- /* Cf. unistr/u8-mblen.c. */
- unsigned char c = (unsigned char) p[0];
-
- if (c >= 0xc2)
- {
- if (c < 0xe0)
- {
- if (m == 1)
- goto incomplete;
- }
- else if (c < 0xf0)
- {
- if (m == 1)
- goto incomplete;
- if (m == 2)
- {
- unsigned char c2 = (unsigned char) p[1];
-
- if ((c2 ^ 0x80) < 0x40
- && (c >= 0xe1 || c2 >= 0xa0)
- && (c != 0xed || c2 < 0xa0))
- goto incomplete;
- }
- }
- else if (c <= 0xf4)
- {
- if (m == 1)
- goto incomplete;
- else /* m == 2 || m == 3 */
- {
- unsigned char c2 = (unsigned char) p[1];
-
- if ((c2 ^ 0x80) < 0x40
- && (c >= 0xf1 || c2 >= 0x90)
- && (c < 0xf4 || (c == 0xf4 && c2 < 0x90)))
- {
- if (m == 2)
- goto incomplete;
- else /* m == 3 */
- {
- unsigned char c3 = (unsigned char) p[2];
-
- if ((c3 ^ 0x80) < 0x40)
- goto incomplete;
- }
- }
- }
- }
- }
- goto invalid;
+ if (pwc != NULL)
+ *pwc = c;
+ res = (c == 0 ? 0 : 1);
+ goto success;
}
-
- /* As a reference for this code, you can use the GNU libiconv
- implementation. Look for uses of the RET_TOOFEW macro. */
-
- if (STREQ_OPT (encoding,
- "EUC-JP", 'E', 'U', 'C', '-', 'J', 'P', 0, 0, 0))
+ if (c >= 0xc2)
{
- if (m == 1)
+ if (c < 0xe0)
{
- unsigned char c = (unsigned char) p[0];
-
- if ((c >= 0xa1 && c < 0xff) || c == 0x8e || c == 0x8f)
+ if (m == 1)
goto incomplete;
- }
- if (m == 2)
- {
- unsigned char c = (unsigned char) p[0];
-
- if (c == 0x8f)
+ else /* m >= 2 */
{
unsigned char c2 = (unsigned char) p[1];
- if (c2 >= 0xa1 && c2 < 0xff)
- goto incomplete;
+ if ((c2 ^ 0x80) < 0x40)
+ {
+ if (pwc != NULL)
+ *pwc = ((unsigned int) (c & 0x1f) << 6)
+ | (unsigned int) (c2 ^ 0x80);
+ res = 2;
+ goto success;
+ }
}
}
- goto invalid;
- }
- if (STREQ_OPT (encoding,
- "EUC-KR", 'E', 'U', 'C', '-', 'K', 'R', 0, 0, 0)
- || STREQ_OPT (encoding,
- "GB2312", 'G', 'B', '2', '3', '1', '2', 0, 0, 0)
- || STREQ_OPT (encoding,
- "BIG5", 'B', 'I', 'G', '5', 0, 0, 0, 0, 0))
- {
- if (m == 1)
+ else if (c < 0xf0)
{
- unsigned char c = (unsigned char) p[0];
-
- if (c >= 0xa1 && c < 0xff)
+ if (m == 1)
goto incomplete;
- }
- goto invalid;
- }
- if (STREQ_OPT (encoding,
- "EUC-TW", 'E', 'U', 'C', '-', 'T', 'W', 0, 0, 0))
- {
- if (m == 1)
- {
- unsigned char c = (unsigned char) p[0];
+ else
+ {
+ unsigned char c2 = (unsigned char) p[1];
- if ((c >= 0xa1 && c < 0xff) || c == 0x8e)
- goto incomplete;
- }
- else /* m == 2 || m == 3 */
- {
- unsigned char c = (unsigned char) p[0];
+ if ((c2 ^ 0x80) < 0x40
+ && (c >= 0xe1 || c2 >= 0xa0)
+ && (c != 0xed || c2 < 0xa0))
+ {
+ if (m == 2)
+ goto incomplete;
+ else /* m >= 3 */
+ {
+ unsigned char c3 = (unsigned char) p[2];
- if (c == 0x8e)
- goto incomplete;
+ if ((c3 ^ 0x80) < 0x40)
+ {
+ if (pwc != NULL)
+ *pwc = ((unsigned int) (c & 0x0f) << 12)
+ | ((unsigned int) (c2 ^ 0x80) << 6)
+ | (unsigned int) (c3 ^ 0x80);
+ res = 3;
+ goto success;
+ }
+ }
+ }
+ }
}
- goto invalid;
- }
- if (STREQ_OPT (encoding,
- "GB18030", 'G', 'B', '1', '8', '0', '3', '0', 0, 0))
- {
- if (m == 1)
+ else if (c <= 0xf4)
{
- unsigned char c = (unsigned char) p[0];
-
- if ((c >= 0x90 && c <= 0xe3) || (c >= 0xf8 && c <= 0xfe))
+ if (m == 1)
goto incomplete;
- }
- else /* m == 2 || m == 3 */
- {
- unsigned char c = (unsigned char) p[0];
-
- if (c >= 0x90 && c <= 0xe3)
+ else
{
unsigned char c2 = (unsigned char) p[1];
- if (c2 >= 0x30 && c2 <= 0x39)
+ if ((c2 ^ 0x80) < 0x40
+ && (c >= 0xf1 || c2 >= 0x90)
+ && (c < 0xf4 || (c == 0xf4 && c2 < 0x90)))
{
if (m == 2)
goto incomplete;
- else /* m == 3 */
+ else
{
unsigned char c3 = (unsigned char) p[2];
- if (c3 >= 0x81 && c3 <= 0xfe)
- goto incomplete;
+ if ((c3 ^ 0x80) < 0x40)
+ {
+ if (m == 3)
+ goto incomplete;
+ else /* m >= 4 */
+ {
+ unsigned char c4 = (unsigned char) p[3];
+
+ if ((c4 ^ 0x80) < 0x40)
+ {
+ if (pwc != NULL)
+ *pwc = ((unsigned int) (c & 0x07) << 18)
+ | ((unsigned int) (c2 ^ 0x80) << 12)
+ | ((unsigned int) (c3 ^ 0x80) << 6)
+ | (unsigned int) (c4 ^ 0x80);
+ res = 4;
+ goto success;
+ }
+ }
+ }
}
}
}
}
- goto invalid;
}
- if (STREQ_OPT (encoding, "SJIS", 'S', 'J', 'I', 'S', 0, 0, 0, 0, 0))
- {
- if (m == 1)
- {
- unsigned char c = (unsigned char) p[0];
+ goto invalid;
+ }
+ else
+ {
+ /* The hidden internal state of mbtowc would make this function not
+ multi-thread safe. Achieve multi-thread safety through a lock. */
+ gl_lock_lock (mbtowc_lock);
- if ((c >= 0x81 && c <= 0x9f) || (c >= 0xe0 && c <= 0xea)
- || (c >= 0xf0 && c <= 0xf9))
- goto incomplete;
- }
- goto invalid;
- }
+ /* Put the hidden internal state of mbtowc into its initial state.
+ This is needed at least with glibc, uClibc, and MSVC CRT.
+ See <https://sourceware.org/bugzilla/show_bug.cgi?id=9674>. */
+ mbtowc (NULL, NULL, 0);
- /* An unknown multibyte encoding. */
- goto incomplete;
- }
+ res = mbtowc (pwc, p, m);
- incomplete:
- {
- size_t k = nstate;
- /* Here 0 <= k < m < 4. */
- pstate[++k] = s[0];
- if (k < m)
+ gl_lock_unlock (mbtowc_lock);
+
+ if (res >= 0)
+ {
+ if (pwc != NULL && ((*pwc == 0) != (res == 0)))
+ abort ();
+ goto success;
+ }
+
+ /* mbtowc does not distinguish between invalid and incomplete multibyte
+ sequences. But mbrtowc needs to make this distinction.
+ There are two possible approaches:
+ - Use iconv() and its return value.
+ - Use built-in knowledge about the possible encodings.
+ Given the low quality of implementation of iconv() on the systems
+ that lack mbrtowc(), we use the second approach.
+ The possible encodings are:
+ - 8-bit encodings,
+ - EUC-JP, EUC-KR, GB2312, EUC-TW, BIG5, GB18030, SJIS,
+ - UTF-8 (already handled above).
+ Use specialized code for each. */
+ if (m >= 4 || m >= MB_CUR_MAX)
+ goto invalid;
+ /* Here MB_CUR_MAX > 1 and 0 < m < 4. */
+ switch (enc)
{
- pstate[++k] = s[1];
- if (k < m)
- pstate[++k] = s[2];
+ /* As a reference for this code, you can use the GNU libiconv
+ implementation. Look for uses of the RET_TOOFEW macro. */
+
+ case enc_eucjp: /* EUC-JP */
+ {
+ if (m == 1)
+ {
+ unsigned char c = (unsigned char) p[0];
+
+ if ((c >= 0xa1 && c < 0xff) || c == 0x8e || c == 0x8f)
+ goto incomplete;
+ }
+ if (m == 2)
+ {
+ unsigned char c = (unsigned char) p[0];
+
+ if (c == 0x8f)
+ {
+ unsigned char c2 = (unsigned char) p[1];
+
+ if (c2 >= 0xa1 && c2 < 0xff)
+ goto incomplete;
+ }
+ }
+ goto invalid;
+ }
+
+ case enc_94: /* EUC-KR, GB2312, BIG5 */
+ {
+ if (m == 1)
+ {
+ unsigned char c = (unsigned char) p[0];
+
+ if (c >= 0xa1 && c < 0xff)
+ goto incomplete;
+ }
+ goto invalid;
+ }
+
+ case enc_euctw: /* EUC-TW */
+ {
+ if (m == 1)
+ {
+ unsigned char c = (unsigned char) p[0];
+
+ if ((c >= 0xa1 && c < 0xff) || c == 0x8e)
+ goto incomplete;
+ }
+ else /* m == 2 || m == 3 */
+ {
+ unsigned char c = (unsigned char) p[0];
+
+ if (c == 0x8e)
+ goto incomplete;
+ }
+ goto invalid;
+ }
+
+ case enc_gb18030: /* GB18030 */
+ {
+ if (m == 1)
+ {
+ unsigned char c = (unsigned char) p[0];
+
+ if ((c >= 0x90 && c <= 0xe3) || (c >= 0xf8 && c <= 0xfe))
+ goto incomplete;
+ }
+ else /* m == 2 || m == 3 */
+ {
+ unsigned char c = (unsigned char) p[0];
+
+ if (c >= 0x90 && c <= 0xe3)
+ {
+ unsigned char c2 = (unsigned char) p[1];
+
+ if (c2 >= 0x30 && c2 <= 0x39)
+ {
+ if (m == 2)
+ goto incomplete;
+ else /* m == 3 */
+ {
+ unsigned char c3 = (unsigned char) p[2];
+
+ if (c3 >= 0x81 && c3 <= 0xfe)
+ goto incomplete;
+ }
+ }
+ }
+ }
+ goto invalid;
+ }
+
+ case enc_sjis: /* SJIS */
+ {
+ if (m == 1)
+ {
+ unsigned char c = (unsigned char) p[0];
+
+ if ((c >= 0x81 && c <= 0x9f) || (c >= 0xe0 && c <= 0xea)
+ || (c >= 0xf0 && c <= 0xf9))
+ goto incomplete;
+ }
+ goto invalid;
+ }
+
+ default:
+ /* An unknown multibyte encoding. */
+ goto incomplete;
}
- if (k != m)
- abort ();
}
- pstate[0] = m;
- return (size_t)(-2);
- invalid:
- errno = EILSEQ;
- /* The conversion state is undefined, says POSIX. */
- return (size_t)(-1);
+ success:
+ /* res >= 0 is the corrected return value of mbtowc (pwc, p, m). */
+ if (nstate >= (res > 0 ? res : 1))
+ abort ();
+ res -= nstate;
+ pstate[0] = 0;
+ return res;
+
+ incomplete:
+ {
+ size_t k = nstate;
+ /* Here 0 <= k < m < 4. */
+ pstate[++k] = s[0];
+ if (k < m)
+ {
+ pstate[++k] = s[1];
+ if (k < m)
+ pstate[++k] = s[2];
+ }
+ if (k != m)
+ abort ();
}
+ pstate[0] = m;
+ return (size_t)(-2);
+
+ invalid:
+ errno = EILSEQ;
+ /* The conversion state is undefined, says POSIX. */
+ return (size_t)(-1);
}
}
projects / deliverable / binutils-gdb.git / blobdiff