bug-gnulib
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

new module 'mbrtowc'


From: Bruno Haible
Subject: new module 'mbrtowc'
Date: Fri, 19 Dec 2008 03:30:13 +0100
User-agent: KMail/1.9.9

The 'mbrtowc' module emulation, based on the system's mbtowc(), is more
interesting:

  - sizeof(mbstate_t) has to be >= 4 for UTF-8, but on IRIX 6.5 mbstate_t is
    defined to 'char'. (Thanks to Paul for idea to verify this assumption :-)).
    So we have to override this type.

  - mbtowc() does not report to the caller a distinction between incomplete and
    invalid multibyte sequence. But mbrtowc() has to do so. How to do so, is
    tricky.

  - What should the return value of  mbrtowc(&wc, s, 0)  be? According to the
    spec, I would think (size_t)(-2). But glibc returns 0. Hmm...

  - glibc's mbtowc() implementation for UTF-8 is buggy.


2008-12-18  Bruno Haible  <address@hidden>

        New module 'mbrtowc'.
        * lib/wchar.in.h (mbstate_t): Override when the system does not have
        mbsinit and mbrtowc.
        (mbrtowc): New declaration.
        * lib/mbrtowc.c: New file.
        * m4/mbrtowc.m4 (gl_FUNC_MBRTOWC, gl_PREREQ_MBRTOWC): New macros.
        * modules/mbrtowc: New file.
        * m4/wchar.m4 (gl_WCHAR_H_DEFAULTS): Initialize GNULIB_MBRTOWC and
        HAVE_MBRTOWC.
        * modules/wchar (Makefile.am): Substitute GNULIB_MBRTOWC and
        HAVE_MBRTOWC.
        * doc/posix-functions/mbrtowc.texi: Document the new module.

================================ lib/mbrtowc.c ================================
/* Convert multibyte character to wide character.
   Copyright (C) 1999-2002, 2005-2008 Free Software Foundation, Inc.
   Written by Bruno Haible <address@hidden>, 2008.

   This program is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 3 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 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/>.  */

#include <config.h>

/* Specification.  */
#include <wchar.h>

#include <errno.h>
#include <stdlib.h>

#include "localcharset.h"
#include "streq.h"


static char internal_state[4];

size_t
mbrtowc (wchar_t *pwc, const char *s, size_t n, mbstate_t *ps)
{
  char *pstate = (char *)ps;

  if (pstate == NULL)
    pstate = internal_state;

  if (s == NULL)
    {
      pwc = NULL;
      s = "";
      n = 1;
    }

  if (n == 0)
    return (size_t)(-2);

  /* Here n > 0.  */
  {
    size_t nstate = pstate[0];
    char buf[4];
    const char *p;
    size_t m;

    switch (nstate)
      {
      case 0:
        p = s;
        m = n;
        break;
      case 3:
        buf[2] = pstate[3];
        /*FALLTHROUGH*/
      case 2:
        buf[1] = pstate[2];
        /*FALLTHROUGH*/
      case 1:
        buf[0] = pstate[1];
        p = buf;
        m = nstate;
        buf[m++] = s[0];
        if (n >= 2 && m < 4)
          {
            buf[m++] = s[1];
            if (n >= 3 && m < 4)
              buf[m++] = s[2];
          }
        break;
      default:
        errno = EINVAL;
        return (size_t)(-1);
      }

    /* Here 0 < m ≤ 4.  */

#if __GLIBC__
    /* 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;
        }

      /* 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, 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.  */
      {
        const char *encoding = locale_charset ();

        if (STREQ (encoding, "UTF-8", 'U', 'T', 'F', '-', '8', 0, 0, 0, 0))
          {
            /* 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;
          }

        /* As a reference for this code, you can use the GNU libiconv
           implementation.  Look for uses of the RET_TOOFEW macro.  */

        if (STREQ (encoding, "EUC-JP", 'E', 'U', 'C', '-', 'J', 'P', 0, 0, 0))
          {
            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;
          }
        if (STREQ (encoding, "EUC-KR", 'E', 'U', 'C', '-', 'K', 'R', 0, 0, 0)
            || STREQ (encoding, "GB2312", 'G', 'B', '2', '3', '1', '2', 0, 0, 0)
            || STREQ (encoding, "BIG5", 'B', 'I', 'G', '5', 0, 0, 0, 0, 0))
          {
            if (m == 1)
              {
                unsigned char c = (unsigned char) p[0];

                if (c >= 0xa1 && c < 0xff)
                  goto incomplete;
              }
            goto invalid;
          }
        if (STREQ (encoding, "EUC-TW", 'E', 'U', 'C', '-', 'T', 'W', 0, 0, 0))
          {
            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;
          }
        if (STREQ (encoding, "SJIS", 'S', 'J', 'I', 'S', 0, 0, 0, 0, 0))
          {
            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;
          }

        /* An unknown multibyte encoding.  */
        goto incomplete;
      }

     incomplete:
      {
        size_t k = nstate;
        /* Here 0 < k < m < 4.  */
        pstate[++k] = s[0];
        if (k < m)
          pstate[++k] = s[1];
        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);
    }
  }
}
=============================== modules/mbrtowc ===============================
Description:
mbrtowc() function: convert multibyte character to wide character.

Files:
lib/mbrtowc.c
m4/mbrtowc.m4
m4/mbstate_t.m4

Depends-on:
wchar
localcharset
streq

configure.ac:
gl_FUNC_MBRTOWC
gl_WCHAR_MODULE_INDICATOR([mbrtowc])

Makefile.am:

Include:
<wchar.h>

License:
LGPL

Maintainer:
Bruno Haible

===============================================================================
--- lib/wchar.in.h.orig 2008-12-19 03:15:03.000000000 +0100
+++ lib/wchar.in.h      2008-12-19 03:13:37.000000000 +0100
@@ -71,6 +71,16 @@
 #endif
 
 
+/* Override mbstate_t if it is too small.
+   On IRIX 6.5, sizeof (mbstate_t) == 1, which is not sufficient for
+   implementing mbrtowc for encodings like UTF-8.  */
+#if !(@HAVE_MBSINIT@ && @HAVE_MBRTOWC@)
+typedef int rpl_mbstate_t;
+# undef mbstate_t
+# define mbstate_t rpl_mbstate_t
+#endif
+
+
 /* Convert a single-byte character to a wide character.  */
 #if @GNULIB_BTOWC@
 # if address@hidden@
@@ -113,6 +123,20 @@
 #endif
 
 
+/* Convert a multibyte character to a wide character.  */
+#if @GNULIB_MBRTOWC@
+# if address@hidden@
+extern size_t mbrtowc (wchar_t *pwc, const char *s, size_t n, mbstate_t *ps);
+# endif
+#elif defined GNULIB_POSIXCHECK
+# undef mbrtowc
+# define mbrtowc(w,s,n,p) \
+    (GL_LINK_WARNING ("mbrtowc is unportable - " \
+                      "use gnulib module mbrtowc for portability"), \
+     mbrtowc (w, s, n, p))
+#endif
+
+
 /* Return the number of screen columns needed for WC.  */
 #if @GNULIB_WCWIDTH@
 # if @REPLACE_WCWIDTH@
--- m4/mbrtowc.m4.orig  2008-12-19 03:15:03.000000000 +0100
+++ m4/mbrtowc.m4       2008-12-19 03:13:37.000000000 +0100
@@ -1,9 +1,29 @@
-# mbrtowc.m4 serial 9
+# mbrtowc.m4 serial 10
 dnl Copyright (C) 2001-2002, 2004-2005, 2008 Free Software Foundation, Inc.
 dnl This file is free software; the Free Software Foundation
 dnl gives unlimited permission to copy and/or distribute it,
 dnl with or without modifications, as long as this notice is preserved.
 
+AC_DEFUN([gl_FUNC_MBRTOWC],
+[
+  AC_REQUIRE([gl_WCHAR_H_DEFAULTS])
+
+  AC_REQUIRE([AC_TYPE_MBSTATE_T])
+  AC_CHECK_FUNCS_ONCE([mbrtowc])
+  if test $ac_cv_func_mbrtowc = no; then
+    HAVE_MBRTOWC=0
+    gl_REPLACE_WCHAR_H
+    AC_LIBOBJ([mbrtowc])
+    gl_PREREQ_MBRTOWC
+  fi
+])
+
+# Prerequisites of lib/mbrtowc.c.
+AC_DEFUN([gl_PREREQ_MBRTOWC], [
+  :
+])
+
+
 dnl From Paul Eggert
 
 dnl This override of an autoconf macro can be removed when autoconf 2.60 or
--- m4/wchar.m4.orig    2008-12-19 03:15:03.000000000 +0100
+++ m4/wchar.m4 2008-12-19 03:13:37.000000000 +0100
@@ -64,11 +64,13 @@
   GNULIB_BTOWC=0;   AC_SUBST([GNULIB_BTOWC])
   GNULIB_WCTOB=0;   AC_SUBST([GNULIB_WCTOB])
   GNULIB_MBSINIT=0; AC_SUBST([GNULIB_MBSINIT])
+  GNULIB_MBRTOWC=0; AC_SUBST([GNULIB_MBRTOWC])
   GNULIB_WCWIDTH=0; AC_SUBST([GNULIB_WCWIDTH])
   dnl Assume proper GNU behavior unless another module says otherwise.
   HAVE_BTOWC=1;        AC_SUBST([HAVE_BTOWC])
   HAVE_WCTOB=1;        AC_SUBST([HAVE_WCTOB])
   HAVE_MBSINIT=1;      AC_SUBST([HAVE_MBSINIT])
+  HAVE_MBRTOWC=1;      AC_SUBST([HAVE_MBRTOWC])
   HAVE_DECL_WCWIDTH=1; AC_SUBST([HAVE_DECL_WCWIDTH])
   REPLACE_WCWIDTH=0;   AC_SUBST([REPLACE_WCWIDTH])
   WCHAR_H='';          AC_SUBST([WCHAR_H])
--- modules/wchar.orig  2008-12-19 03:15:03.000000000 +0100
+++ modules/wchar       2008-12-19 03:13:37.000000000 +0100
@@ -28,11 +28,13 @@
              -e 's|@''GNULIB_BTOWC''@|$(GNULIB_BTOWC)|g' \
              -e 's|@''GNULIB_WCTOB''@|$(GNULIB_WCTOB)|g' \
              -e 's|@''GNULIB_MBSINIT''@|$(GNULIB_MBSINIT)|g' \
+             -e 's|@''GNULIB_MBRTOWC''@|$(GNULIB_MBRTOWC)|g' \
              -e 's|@''GNULIB_WCWIDTH''@|$(GNULIB_WCWIDTH)|g' \
              -e 's|@''HAVE_WINT_T''@|$(HAVE_WINT_T)|g' \
              -e 's|@''HAVE_BTOWC''@|$(HAVE_BTOWC)|g' \
              -e 's|@''HAVE_WCTOB''@|$(HAVE_WCTOB)|g' \
              -e 's|@''HAVE_MBSINIT''@|$(HAVE_MBSINIT)|g' \
+             -e 's|@''HAVE_MBRTOWC''@|$(HAVE_MBRTOWC)|g' \
              -e 's|@''HAVE_DECL_WCWIDTH''@|$(HAVE_DECL_WCWIDTH)|g' \
              -e 's|@''REPLACE_WCWIDTH''@|$(REPLACE_WCWIDTH)|g' \
              -e '/definition of GL_LINK_WARNING/r $(LINK_WARNING_H)' \
--- doc/posix-functions/mbrtowc.texi.orig       2008-12-19 03:15:03.000000000 
+0100
+++ doc/posix-functions/mbrtowc.texi    2008-12-19 03:13:37.000000000 +0100
@@ -4,18 +4,18 @@
 
 POSIX specification: 
@url{http://www.opengroup.org/onlinepubs/9699919799/functions/mbrtowc.html}
 
-Gnulib module: ---
+Gnulib module: mbrtowc
 
 Portability problems fixed by Gnulib:
 @itemize
address@hidden
+This function is missing on some platforms:
+HP-UX 11, IRIX 6.5, Solaris 2.6, mingw, Interix 3.5.
 @end itemize
 
 Portability problems not fixed by Gnulib:
 @itemize
 @item
-This function is missing on some platforms:
-HP-UX 11, IRIX 6.5, Solaris 2.6, mingw, Interix 3.5.
address@hidden
 On Windows platforms, @code{wchar_t} is a 16-bit type and therefore cannot
 accommodate all Unicode characters.
 @end itemize




reply via email to

[Prev in Thread] Current Thread [Next in Thread]