Re: wide char consistency in the C locale

[Bruno Haible via Gnulib discussion list]

I did:
>       * lib/mbrtowc.c (rpl_mbrtowc) [MBRTOWC_IN_C_LOCALE_MAYBE_EILSEQ]: In the
>       C locale, map 0x80..0xFF to 0xDF80..0xDFFF instead of 0x0080..0x00FF.
>       * lib/wcrtomb.c: Include hard-locale.h, <locale.h>.
>       (wcrtomb) [MBRTOWC_IN_C_LOCALE_MAYBE_EILSEQ]: In the C locale, map
>       0xDF80..0xDFFF to 0x80..0xFF.

This causes test failures among the *c32* functions. Fixed as follows:


2026-09-20  Bruno Haible  <br...@clisp.org>

        mbrtoc32, c32rtomb: Make consistent with mbrtowc on glibc (regr. today).
        * lib/mbrtoc32.c (mbrtoc32): In the C locale, map 0x80..0xFF to
        U+DF80..U+DFFF instead of U+0080..U+00FF.
        * lib/c32rtomb.c (c32rtomb): In the C locale, map U+DF80..U+DFFF to
        0x80..0xFF.

diff --git a/lib/c32rtomb.c b/lib/c32rtomb.c
index 9da0abb6be..1503fe2bc0 100644
--- a/lib/c32rtomb.c
+++ b/lib/c32rtomb.c
@@ -54,6 +54,7 @@ c32rtomb (char *s, char32_t wc, mbstate_t *ps)
 # endif
 
 # if MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859 /* OpenBSD */ \
+     || MBRTOC32_IN_C_LOCALE_MAYBE_EILSEQ /* glibc */ \
      || !_GL_WCHAR_T_IS_UCS4 /* NetBSD ≥ 11 */
   if ((wc >= 0xDF80 && wc <= 0xDFFF) && !hard_locale (LC_CTYPE))
     {
@@ -64,16 +65,6 @@ c32rtomb (char *s, char32_t wc, mbstate_t *ps)
     }
 # endif
 
-# if MBRTOC32_IN_C_LOCALE_MAYBE_EILSEQ
-  if ((wc >= 0x0080 && wc <= 0x00FF) && !hard_locale (LC_CTYPE))
-    {
-      /* In the "C" locale, map the code points U+0080..U+00FF back to the 
bytes
-         0x80..0xFF, for consistency with the mbrtoc32 and btoc32 functions.  
*/
-      s[0] = (unsigned char) wc;
-      return 1;
-    }
-# endif
-
   return c32rtomb (s, wc, ps);
 
 #elif !GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T /* Cygwin, mingw, MSVC */
diff --git a/lib/mbrtoc32.c b/lib/mbrtoc32.c
index 00de09a258..57d7fc2fca 100644
--- a/lib/mbrtoc32.c
+++ b/lib/mbrtoc32.c
@@ -304,11 +304,11 @@ mbrtoc32 (char32_t *pwc, const char *s, size_t n, 
mbstate_t *ps)
     abort ();
 #  endif
 
-#  if MBRTOC32_IN_C_LOCALE_MAYBE_EILSEQ
+#  if MBRTOC32_IN_C_LOCALE_MAYBE_EILSEQ /* glibc */
   if ((size_t) -2 <= ret && n != 0 && ! hard_locale (LC_CTYPE))
     {
       if (pwc != NULL)
-        *pwc = (unsigned char) *s;
+        *pwc = 0xDF00 + (unsigned char) *s;
       return 1;
     }
 #  endif