Re: new test failures on Cygwin and OpenBSD

Bruno Haible via Gnulib discussion list <[email protected]>
Newsgroups gmane.comp.lib.gnulib.bugs
Message-ID <2821197.7ZnGzRAi6d@cagnes> 
Paul Eggert wrote:
> > So, for me, (3) is the right approach: In the "C" locale, map 0x80..0xFF not
> > to U+0080..U+00FF but to U+DF80..U+DFFF, like musl libc does it. This will
> > ensure that c32isprint returns 0 on these inputs, like isprint does.
> 
> This sounds good to me as well. It is unfortunate that the "C" locale, which is supposed to be fast, will have a small performance penalty on OpenBSD and Cygwin (and I assume native MS-Windows?) due to Gnulib having to replace mbrtoc32. But that's better than incorrect behavior.

Yes, I agree.

I used the program below to look at the behaviour of various platforms:

Result:

                       wchar_t world                           char32_t world

glibc                  mbrtowc failed                          OK
musl                   OK (...wc=0xDFFF)                       OK
macOS                  OK (...wc=0x00FF)                       OK
FreeBSD                OK (...wc=0x00FF)                       OK
NetBSD                 OK (...wc=0x00FF)                       OK
OpenBSD ≤ 7.3          OK (...wc=0x00FF)                       OK
OpenBSD                OK (...wc=0x00FF)                       FAIL
AIX 32-bit             OK (...wc=0x00FF)                       OK
AIX 64-bit             OK (...wc=0x00FF)                       OK
Solaris 11             OK (...wc=0x00FF)                       OK
Solaris OpenIndiana    OK (...wc=0x00FF)                       OK
Cygwin                 mbrtowc failed                          FAIL
mingw                  isprint=0 -- ...wc=0x00FF iswprint=1    FAIL
MSVC                   isprint=0 -- ...wc=0x00FF iswprint=1    FAIL
Android                mbrtowc failed                          OK ("C" is UTF-8)
Haiku                  mbrtowc failed                          OK

From this list, I infer that the correct condition
  - for the Cygwin, mingw, MSVC case is:
    _GL_SMALL_WCHAR_T && !GNULIB_defined_mbstate_t
  - for the OpenBSD case is:
    !MBRTOWC_IN_C_LOCALE_MAYBE_EILSEQ && HAVE_MBRTOC32 && !MUSL_LIBC

Implemented through the three attached patches.

Bruno

================================================================================
#include <locale.h>
#include <wchar.h>
#include <ctype.h>
#include <wctype.h>
#include <stdio.h>

int main ()
{
  int result;
  if (setlocale (LC_ALL, "C") == NULL)
    return 2;
  result = 0;
  {
    int i;
    for (i = 0xC0; i <= 0xFF; i++)
      {
        char c = i;
        wchar_t wc;
        mbstate_t mbs = { 0, };
        if (mbrtowc (&wc, &c, 1, &mbs) > 1)
          {
            printf ("mbrtowc failed for c=0x%02X\n", i);
            return 3;
          }
        int is1 = !!isprint ((unsigned char) c);
        int is2 = !!iswprint (wc);
        printf ("c=0x%02X isprint=%d -- wc=0x%04X iswprint=%d\n", i, is1, wc, is2);
        if (is1 != is2)
          result = 1;
      }
  }
  if (result == 0)
    printf ("OK\n");
  return result;
}
================================================================================
0001-mbrtoc32-Make-c32isprint-compatible-with-isprint-in-.patch (text/x-patch, 4.7 KB) 
From d05f95bec14dcd5d58aef358a520f15ec64c381b Mon Sep 17 00:00:00 2001
From: Bruno Haible <[email protected]>
Date: Thu, 27 Aug 2026 23:53:07 +0200
Subject: [PATCH 1/3] mbrtoc32: Make c32isprint compatible with isprint in the
 C locale.

Reported by Paul Eggert in
<https://lists.gnu.org/archive/html/bug-gnulib/2026-08/msg00222.html>.

* lib/mbrtoc32.c (mbrtoc32)
[!GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T]: On Cygwin and native
Windows, in the C locale, map 0x80..0xFF differently.
* lib/btoc32.c: Include hard-locale.h.
(btoc32): [!GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T]: On Cygwin
and native Windows, in the C locale, map 0x80..0xFF differently.
* modules/btoc32 (Depends-on): Add hard-locale.
---
 ChangeLog      | 15 ++++++++++++++-
 lib/btoc32.c   | 17 +++++++++++++++++
 lib/mbrtoc32.c | 14 ++++++++++++--
 modules/btoc32 |  1 +
 4 files changed, 44 insertions(+), 3 deletions(-)

diff --git a/ChangeLog b/ChangeLog
index a02b8878b4..53aa25ecb2 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,16 @@
+2026-08-27  Bruno Haible  <[email protected]>
+
+	mbrtoc32: Make c32isprint compatible with isprint in the C locale.
+	Reported by Paul Eggert in
+	<https://lists.gnu.org/archive/html/bug-gnulib/2026-08/msg00222.html>.
+	* lib/mbrtoc32.c (mbrtoc32)
+	[!GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T]: On Cygwin and native
+	Windows, in the C locale, map 0x80..0xFF differently.
+	* lib/btoc32.c: Include hard-locale.h.
+	(btoc32): [!GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T]: On Cygwin
+	and native Windows, in the C locale, map 0x80..0xFF differently.
+	* modules/btoc32 (Depends-on): Add hard-locale.
+
 2026-08-27  Bruno Haible  <[email protected]>
 
 	getlogin: Fix link error with option --with-systemd (regr. 2025-03-09).
@@ -1862,7 +1875,7 @@
 	go back to the old way of doing things, without the recently-added
 	USE_C_LOCALE stuff.  Then make the following changes instead:
 	* lib/quotearg.c [_QUOTEARG_AVOID_UCHAR_H]: Don’t include <uchar.h>.
-	(mbrto2c, mbstate_t) [_QUOTEARG_AVOID_UCHAR_H && defined __UCLIBC__
+	(mbrtowc, mbstate_t) [_QUOTEARG_AVOID_UCHAR_H && defined __UCLIBC__
 	&& !defined __UCLIBC_HAS_WCHAR__]: New macros.
 	Also, don’t include <wchar.h> or <wctype.h>.
 	(char32_t, c32isprint, mbrtoc32, GNULIB_MBRTOC32_REGULAR)
diff --git a/lib/btoc32.c b/lib/btoc32.c
index e4ca972db0..9af58b2582 100644
--- a/lib/btoc32.c
+++ b/lib/btoc32.c
@@ -26,6 +26,11 @@
 #include <string.h>
 #include <wchar.h>
 
+#if !GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T
+# include "hard-locale.h"
+# include <locale.h>
+#endif
+
 #if GL_CHAR32_T_IS_UNICODE
 # include "lc-charset-unicode.h"
 #endif
@@ -51,6 +56,18 @@ btoc32 (int c)
     }
   return WEOF;
 #else
+# if !GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T /* Cygwin, mingw, MSVC */
+  if (!hard_locale (LC_CTYPE))
+    {
+      /* In the "C" locale, map the bytes 0x80..0xFF to U+DF80..U+DFFF, so that
+         the c32is* functions return false on them, for consistency with the
+         <ctype.h> is* functions.  */
+      if (c != EOF)
+        return (c < 0x80 ? c : 0xDF00 + c);
+      else
+        return WEOF;
+    }
+# endif
   /* In all known locale encodings, unibyte characters correspond only to
      characters in the BMP.  */
   wint_t wc = btowc (c);
diff --git a/lib/mbrtoc32.c b/lib/mbrtoc32.c
index 5f89e05460..ff4799ded1 100644
--- a/lib/mbrtoc32.c
+++ b/lib/mbrtoc32.c
@@ -92,7 +92,7 @@ mbrtoc32 (char32_t *pwc, const char *s, size_t n, mbstate_t *ps)
 # include "localcharset.h"
 # include "streq-opt.h"
 
-# if MBRTOC32_IN_C_LOCALE_MAYBE_EILSEQ
+# if MBRTOC32_IN_C_LOCALE_MAYBE_EILSEQ || _GL_SMALL_WCHAR_T
 #  include "hard-locale.h"
 #  include <locale.h>
 # endif
@@ -291,7 +291,7 @@ mbrtoc32 (char32_t *pwc, const char *s, size_t n, mbstate_t *ps)
 
   return ret;
 
-# elif _GL_SMALL_WCHAR_T
+# elif _GL_SMALL_WCHAR_T /* Cygwin, mingw, MSVC */
 
   /* Special-case all encodings that may produce wide character values
      > WCHAR_MAX.  */
@@ -386,6 +386,16 @@ mbrtoc32 (char32_t *pwc, const char *s, size_t n, mbstate_t *ps)
       /* The conversion state is undefined, says POSIX.  */
       return (size_t)(-1);
     }
+  else if (!hard_locale (LC_CTYPE))
+    {
+      /* In the "C" locale, map the bytes 0x80..0xFF to U+DF80..U+DFFF, so that
+         the c32is* functions return false on them, for consistency with the
+         <ctype.h> is* functions.  */
+      unsigned char c = (unsigned char) s[0];
+      if (pwc != NULL)
+        *pwc = (c < 0x80 ? c : 0xDF00 + c);
+      return (c == 0 ? 0 : 1);
+    }
   else
     {
       wchar_t wc;
diff --git a/modules/btoc32 b/modules/btoc32
index e2acebb402..d3e13be578 100644
--- a/modules/btoc32
+++ b/modules/btoc32
@@ -8,6 +8,7 @@ Depends-on:
 uchar-h
 mbszero
 mbrtoc32
+hard-locale
 btowc
 
 configure.ac:
-- 
2.53.0
0002-mbrtoc32-Make-c32isprint-compatible-with-isprint-in-.patch (text/x-patch, 9.4 KB) 
From 27353a8efc89559b855c9688a8c32183418267b7 Mon Sep 17 00:00:00 2001
From: Bruno Haible <[email protected]>
Date: Fri, 28 Aug 2026 00:41:53 +0200
Subject: [PATCH 2/3] mbrtoc32: Make c32isprint compatible with isprint in the
 C locale.

Reported by Paul Eggert in
<https://lists.gnu.org/archive/html/bug-gnulib/2026-08/msg00222.html>.

* m4/mbrtoc32.m4 (gl_MBRTOC32_C_LOCALE_LIKE_ISO_8859): New macro.
(gl_FUNC_MBRTOC32): Define MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859 and
set REPLACE_MBRTOC32 on platforms such as OpenBSD.
* lib/mbrtoc32.c (mbrtoc32) [MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859]:
In the C locale, map 0x80..0xFF differently.
* lib/btoc32.c (btoc32) [MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859]:
Likewise.
* doc/posix-functions/mbrtoc32.texi: Mention the OpenBSD problem.
---
 ChangeLog                         | 12 +++++++
 doc/posix-functions/mbrtoc32.texi |  5 +++
 lib/btoc32.c                      | 28 ++++++++--------
 lib/mbrtoc32.c                    | 23 +++++++++++--
 m4/mbrtoc32.m4                    | 55 +++++++++++++++++++++++++++++--
 5 files changed, 105 insertions(+), 18 deletions(-)

diff --git a/ChangeLog b/ChangeLog
index 53aa25ecb2..b829acc302 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,17 @@
 2026-08-27  Bruno Haible  <[email protected]>
 
+	mbrtoc32: Make c32isprint compatible with isprint in the C locale.
+	Reported by Paul Eggert in
+	<https://lists.gnu.org/archive/html/bug-gnulib/2026-08/msg00222.html>.
+	* m4/mbrtoc32.m4 (gl_MBRTOC32_C_LOCALE_LIKE_ISO_8859): New macro.
+	(gl_FUNC_MBRTOC32): Define MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859 and
+	set REPLACE_MBRTOC32 on platforms such as OpenBSD.
+	* lib/mbrtoc32.c (mbrtoc32) [MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859]:
+	In the C locale, map 0x80..0xFF differently.
+	* lib/btoc32.c (btoc32) [MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859]:
+	Likewise.
+	* doc/posix-functions/mbrtoc32.texi: Mention the OpenBSD problem.
+
 	mbrtoc32: Make c32isprint compatible with isprint in the C locale.
 	Reported by Paul Eggert in
 	<https://lists.gnu.org/archive/html/bug-gnulib/2026-08/msg00222.html>.
diff --git a/doc/posix-functions/mbrtoc32.texi b/doc/posix-functions/mbrtoc32.texi
index 629a75dec4..72a373b3e6 100644
--- a/doc/posix-functions/mbrtoc32.texi
+++ b/doc/posix-functions/mbrtoc32.texi
@@ -22,6 +22,11 @@
 @c https://sourceware.org/PR29511
 glibc 2.35.
 @item
+In the C or POSIX locales,
+this function returns values like in ISO-8859-1 locales,
+which leads to @code{c32isprint} etc.@: being inconsistent with @code{isprint}:
+OpenBSD 7.9.
+@item
 This function returns 0 instead of @code{(size_t) -2} when the input
 is empty:
 @c https://sourceware.org/PR16950
diff --git a/lib/btoc32.c b/lib/btoc32.c
index 9af58b2582..31b3890723 100644
--- a/lib/btoc32.c
+++ b/lib/btoc32.c
@@ -26,7 +26,8 @@
 #include <string.h>
 #include <wchar.h>
 
-#if !GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T
+#if MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859 \
+    || (!GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T)
 # include "hard-locale.h"
 # include <locale.h>
 #endif
@@ -41,6 +42,19 @@ _GL_EXTERN_INLINE
 wint_t
 btoc32 (int c)
 {
+#if MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859 /* OpenBSD */ \
+    || (!GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T) /* Cygwin, mingw, MSVC */
+  if (!hard_locale (LC_CTYPE))
+    {
+      /* In the "C" locale, map the bytes 0x80..0xFF to U+DF80..U+DFFF, so that
+         the c32is* functions return false on them, for consistency with the
+         <ctype.h> is* functions.  */
+      if (c != EOF)
+        return (c < 0x80 ? c : 0xDF00 + c);
+      else
+        return WEOF;
+    }
+#endif
 #if HAVE_WORKING_MBRTOC32 && HAVE_WORKING_C32RTOMB && !_GL_WCHAR_T_IS_UCS4
   /* The char32_t encoding of a multibyte character may be different than its
      wchar_t encoding.  */
@@ -56,18 +70,6 @@ btoc32 (int c)
     }
   return WEOF;
 #else
-# if !GNULIB_defined_mbstate_t && _GL_SMALL_WCHAR_T /* Cygwin, mingw, MSVC */
-  if (!hard_locale (LC_CTYPE))
-    {
-      /* In the "C" locale, map the bytes 0x80..0xFF to U+DF80..U+DFFF, so that
-         the c32is* functions return false on them, for consistency with the
-         <ctype.h> is* functions.  */
-      if (c != EOF)
-        return (c < 0x80 ? c : 0xDF00 + c);
-      else
-        return WEOF;
-    }
-# endif
   /* In all known locale encodings, unibyte characters correspond only to
      characters in the BMP.  */
   wint_t wc = btowc (c);
diff --git a/lib/mbrtoc32.c b/lib/mbrtoc32.c
index ff4799ded1..826f367708 100644
--- a/lib/mbrtoc32.c
+++ b/lib/mbrtoc32.c
@@ -92,7 +92,9 @@ mbrtoc32 (char32_t *pwc, const char *s, size_t n, mbstate_t *ps)
 # include "localcharset.h"
 # include "streq-opt.h"
 
-# if MBRTOC32_IN_C_LOCALE_MAYBE_EILSEQ || _GL_SMALL_WCHAR_T
+# if (MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859 \
+      || MBRTOC32_IN_C_LOCALE_MAYBE_EILSEQ \
+      || _GL_SMALL_WCHAR_T)
 #  include "hard-locale.h"
 #  include <locale.h>
 # endif
@@ -134,8 +136,10 @@ mbrtoc32 (char32_t *pwc, const char *s, size_t n, mbstate_t *ps)
       n = 1;
     }
 
-# if (MBRTOC32_EMPTY_INPUT_BUG || _GL_SMALL_WCHAR_T \
-      || (GNULIB_WCHAR_SINGLE_LOCALE && __GLIBC__ >= 2 && !__UCLIBC__))
+# if (MBRTOC32_EMPTY_INPUT_BUG \
+      || (GNULIB_WCHAR_SINGLE_LOCALE && __GLIBC__ >= 2 && !__UCLIBC__) \
+      || MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859 \
+      || _GL_SMALL_WCHAR_T)
   if (n == 0)
     return (size_t) -2;
 # endif
@@ -262,6 +266,19 @@ mbrtoc32 (char32_t *pwc, const char *s, size_t n, mbstate_t *ps)
   /* mbrtoc32() may produce different values for wc than mbrtowc().  Therefore
      use mbrtoc32().  */
 
+#  if MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859 /* OpenBSD */
+  if (!hard_locale (LC_CTYPE))
+    {
+      /* In the "C" locale, map the bytes 0x80..0xFF to U+DF80..U+DFFF, so that
+         the c32is* functions return false on them, for consistency with the
+         <ctype.h> is* functions.  */
+      unsigned char c = (unsigned char) s[0];
+      if (pwc != NULL)
+        *pwc = (c < 0x80 ? c : 0xDF00 + c);
+      return (c == 0 ? 0 : 1);
+    }
+#  endif
+
 #  if defined _WIN32 && !defined __CYGWIN__
   char32_t wc;
   size_t ret = mbrtoc32 (&wc, s, n, ps);
diff --git a/m4/mbrtoc32.m4 b/m4/mbrtoc32.m4
index 9eb9a7e223..2631695713 100644
--- a/m4/mbrtoc32.m4
+++ b/m4/mbrtoc32.m4
@@ -1,5 +1,5 @@
 # mbrtoc32.m4
-# serial 25
+# serial 26
 dnl Copyright (C) 2014-2026 Free Software Foundation, Inc.
 dnl This file is free software; the Free Software Foundation
 dnl gives unlimited permission to copy and/or distribute it,
@@ -9,6 +9,7 @@
 AC_DEFUN([gl_FUNC_MBRTOC32],
 [
   AC_REQUIRE([gl_UCHAR_H_DEFAULTS])
+  AC_REQUIRE([AC_CANONICAL_HOST])
 
   AC_REQUIRE([AC_TYPE_MBSTATE_T])
   dnl Determine REPLACE_MBSTATE_T, from which GNULIB_defined_mbstate_t is
@@ -28,6 +29,7 @@ AC_DEFUN([gl_FUNC_MBRTOC32]
     else
       gl_MBRTOC32_EMPTY_INPUT
       gl_MBRTOC32_C_LOCALE
+      gl_MBRTOC32_C_LOCALE_LIKE_ISO_8859
       gl_MBRTOC32_UTF8_LOCALE
       case "$gl_cv_func_mbrtoc32_empty_input" in
         *yes) ;;
@@ -43,6 +45,17 @@ AC_DEFUN([gl_FUNC_MBRTOC32]
            REPLACE_MBRTOC32=1
            ;;
       esac
+      case "$gl_cv_func_mbrtoc32_C_locale_sans_EILSEQ" in
+        *yes)
+          case "$gl_cv_func_mbrtoc32_C_locale_like_iso_8859" in
+            *yes)
+              AC_DEFINE([MBRTOC32_IN_C_LOCALE_MAYBE_LIKE_ISO_8859], [1],
+                [Define if the mbrtoc32 function in the C locale may work like in an ISO-8859-1 locale.])
+              REPLACE_MBRTOC32=1
+              ;;
+          esac
+          ;;
+      esac
       case "$gl_cv_func_mbrtoc32_utf8_locale_works" in
         *yes) ;;
         *) AC_DEFINE([MBRTOC32_MULTIBYTE_LOCALE_BUG], [1],
@@ -167,7 +180,7 @@ AC_DEFUN([gl_MBRTOC32_EMPTY_INPUT]
 AC_DEFUN([gl_MBRTOC32_C_LOCALE],
 [
   AC_REQUIRE([AC_CANONICAL_HOST]) dnl for cross-compiles
-  AC_CACHE_CHECK([whether the C locale is free of encoding errors],
+  AC_CACHE_CHECK([whether mbrtoc32 in the C locale is free of encoding errors],
     [gl_cv_func_mbrtoc32_C_locale_sans_EILSEQ],
     [AC_RUN_IFELSE(
        [AC_LANG_PROGRAM(
@@ -204,6 +217,44 @@ AC_DEFUN([gl_MBRTOC32_C_LOCALE]
     ])
 ])
 
+dnl Test whether mbrtoc32 in the C locale works like in an ISO-8859-1 locale.
+
+AC_DEFUN([gl_MBRTOC32_C_LOCALE_LIKE_ISO_8859],
+[
+  AC_REQUIRE([AC_CANONICAL_HOST]) dnl for cross-compiles
+  AC_CACHE_CHECK([whether mbrtoc32 in the C locale works like in an ISO-8859-1 locale],
+    [gl_cv_func_mbrtoc32_C_locale_like_iso_8859],
+    [AC_RUN_IFELSE(
+       [AC_LANG_PROGRAM(
+          [[#include <limits.h>
+            #include <locale.h>
+            #ifdef __HAIKU__
+             #include <stdint.h>
+            #endif
+            #include <uchar.h>
+          ]], [[
+            int i;
+            const char *locale = setlocale (LC_ALL, "C");
+            if (! locale)
+              return 2;
+            char c = 0xE0;
+            char32_t wc;
+            mbstate_t mbs = { 0, };
+            size_t ss = mbrtoc32 (&wc, &c, 1, &mbs);
+            if (ss == 1 && wc == 0x00E0)
+              return 3;
+            return 0;
+          ]])],
+       [gl_cv_func_mbrtoc32_C_locale_like_iso_8859=no],
+       [gl_cv_func_mbrtoc32_C_locale_like_iso_8859=yes],
+       [case "$host_os" in
+          *-musl* | midipix*) gl_cv_func_mbrtoc32_C_locale_like_iso_8859="no" ;;
+          *)                  gl_cv_func_mbrtoc32_C_locale_like_iso_8859="possibly yes" ;;
+        esac
+       ])
+    ])
+])
+
 dnl Test whether mbrtoc32 works when it's fed the bytes one-by-one in an UTF-8
 dnl locale.
 
-- 
2.53.0
0003-tests-Check-that-c32isprint-is-consistent-with-ispri.patch (text/x-patch, 2.1 KB) 
From de054f5dc8906dbac0054e113d7d4fd0430a3398 Mon Sep 17 00:00:00 2001
From: Bruno Haible <[email protected]>
Date: Fri, 28 Aug 2026 00:43:56 +0200
Subject: [PATCH 3/3] tests: Check that c32isprint is consistent with isprint
 in the C locale.

* tests/test-c32isprint.c: Include <ctype.h>.
(main): In the C locale, check that c32isprint is consistent with
isprint.
---
 ChangeLog               |  5 +++++
 tests/test-c32isprint.c | 15 +++++++++++++++
 2 files changed, 20 insertions(+)

diff --git a/ChangeLog b/ChangeLog
index b829acc302..2254b198b1 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,10 @@
 2026-08-27  Bruno Haible  <[email protected]>
 
+	tests: Check that c32isprint is consistent with isprint in the C locale.
+	* tests/test-c32isprint.c: Include <ctype.h>.
+	(main): In the C locale, check that c32isprint is consistent with
+	isprint.
+
 	mbrtoc32: Make c32isprint compatible with isprint in the C locale.
 	Reported by Paul Eggert in
 	<https://lists.gnu.org/archive/html/bug-gnulib/2026-08/msg00222.html>.
diff --git a/tests/test-c32isprint.c b/tests/test-c32isprint.c
index 9d91ef3640..42a5092d4f 100644
--- a/tests/test-c32isprint.c
+++ b/tests/test-c32isprint.c
@@ -21,6 +21,7 @@
 #include "signature.h"
 SIGNATURE_CHECK (c32isprint, int, (wint_t));
 
+#include <ctype.h>
 #include <locale.h>
 #include <stdlib.h>
 #include <string.h>
@@ -106,6 +107,20 @@ main (int argc, char *argv[])
       {
       case '0':
         /* C locale; tested above.  */
+#if !defined __ANDROID__
+        /* On Android ≥ 5.0, the default locale is the "C.UTF-8" locale, not the
+           "C" locale.  Furthermore, when you attempt to set the "C" or "POSIX"
+           locale via setlocale(), what you get is a "C" locale with UTF-8
+           encoding, that is, effectively the "C.UTF-8" locale.  */
+        /* Check that c32isprint is consistent with isprint.  */
+        for (int c = 0x80; c < 0x100; c++)
+          {
+            ASSERT (!isprint ((unsigned char) c));
+            buf[0] = (unsigned char) c;
+            is = for_character (buf, 1);
+            ASSERT (is == 0);
+          }
+#endif
         return test_exit_status;
 
       case '1':
-- 
2.53.0
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.