Android and the C locale

[Bruno Haible]

Android < 5.0 had only dummy locales.

Starting with Android 5.0 (according to the Android libc's git history),
they have locales. But there are two problems:

1) The default locale (i.e. the locale in use when setlocale was not called)
is the "C.UTF-8" locale, not the "C" locale.

Test case:
================================================================================
#include <stdio.h>
#include <stdlib.h>
#include <locale.h>

int main ()
{
  printf ("Locale=|%s| LC_CTYPE=|%s| MB_CUR_MAX=%d\n",
          setlocale (LC_ALL, NULL), setlocale (LC_CTYPE, NULL), (int) 
MB_CUR_MAX);
}
================================================================================
prints
  Locale=|C.UTF-8| LC_CTYPE=|C.UTF-8| MB_CUR_MAX=4
rather than the expected
  Locale=|C| LC_CTYPE=|C| MB_CUR_MAX=1

POSIX 
<https://pubs.opengroup.org/onlinepubs/9699919799.2018edition/basedefs/V1_chap07.html#tag_07_02>
says that the default locale should be the "C"/"POSIX" locale.

2) A setlocale call that is meant to set the "C" or "POSIX" locale actually
sets a locale with UTF-8 encoding.

Test case 1:
================================================================================
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <locale.h>
#include <wchar.h>

int main ()
{
  mbstate_t state;
  if (setlocale (LC_ALL, "") == NULL)
    return 1;
  memset (&state, '\0', sizeof (state));
  printf ("Locale=|%s| LC_CTYPE=|%s| MB_CUR_MAX=%d mbrtowc(0xC0)=%d\n",
          setlocale (LC_ALL, NULL), setlocale (LC_CTYPE, NULL), (int) 
MB_CUR_MAX,
          (int) mbrtowc (NULL, "\xC0", 1, &state));
}
================================================================================
$ LC_ALL=C ./a.out
and
$ LC_ALL=POSIX ./a.out
print
  Locale=|C.UTF-8| LC_CTYPE=|C.UTF-8| MB_CUR_MAX=4 mbrtowc(0xC0)=-2
rather than the expected
  Locale=|C| LC_CTYPE=|C| MB_CUR_MAX=1 mbrtowc(0xC0)=-1

Test case 2:
================================================================================
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <locale.h>
#include <wchar.h>

int main ()
{
  mbstate_t state;
  if (setlocale (LC_ALL, "C") == NULL)
    return 1;
  memset (&state, '\0', sizeof (state));
  printf ("Locale=|%s| LC_CTYPE=|%s| MB_CUR_MAX=%d mbrtowc(0xC0)=%d\n",
          setlocale (LC_ALL, NULL), setlocale (LC_CTYPE, NULL), (int) 
MB_CUR_MAX,
          (int) mbrtowc (NULL, "\xC0", 1, &state));
  if (setlocale (LC_ALL, "POSIX") == NULL)
    return 1;
  memset (&state, '\0', sizeof (state));
  printf ("Locale=|%s| LC_CTYPE=|%s| MB_CUR_MAX=%d mbrtowc(0xC0)=%d\n",
          setlocale (LC_ALL, NULL), setlocale (LC_CTYPE, NULL), (int) 
MB_CUR_MAX,
          (int) mbrtowc (NULL, "\xC0", 1, &state));
}
================================================================================
prints
  Locale=|C| LC_CTYPE=|C| MB_CUR_MAX=4 mbrtowc(0xC0)=-2
  Locale=|C| LC_CTYPE=|C| MB_CUR_MAX=4 mbrtowc(0xC0)=-2
rather than the expected
  Locale=|C| LC_CTYPE=|C| MB_CUR_MAX=1 mbrtowc(0xC0)=-1
  Locale=|C| LC_CTYPE=|C| MB_CUR_MAX=1 mbrtowc(0xC0)=-1

One of the consequences are these two test failures:


FAIL: test-mbrtoc32-5.sh
========================

../../gltests/test-mbrtoc32.c:105: assertion 'ret == 1' failed
Aborted
FAIL test-mbrtoc32-5.sh (exit status: 134)

FAIL: test-mbrtowc5.sh
======================

../../gltests/test-mbrtowc.c:105: assertion 'ret == 1' failed
Aborted
FAIL test-mbrtowc5.sh (exit status: 134)


As a workaround, I'm applying these two patches.

2023-01-16  Bruno Haible  <bruno@clisp.org>

        mbrtowc, mbrtoc32 tests: Avoid test failure on Android ≥ 5.0.
        * tests/test-mbrtowc.c (main): On Android 5.0 or newer, when testing
        the "C" locale, verify that the encoding is UTF-8.
        * tests/test-mbrtoc32.c (main): Likewise.
        * doc/posix-functions/setlocale.texi: Mention the Android problems.

        mbrtowc, mbrtoc32 tests: Refactor.
        * tests/test-mbrtowc.c (main): Straighten convoluted code.
        * tests/test-mbrtoc32.c (main): Likewise.

0001-mbrtowc-mbrtoc32-tests-Refactor.patch
Description: Text Data

0002-mbrtowc-mbrtoc32-tests-Avoid-test-failure-on-Android.patch
Description: Text Data