bpo-30647: Check nl_langinfo(CODESET) in locale coercion (GH-2374)

- On some versions of FreeBSD, setting the "UTF-8" locale
  succeeds, but a subsequent "nl_langinfo(CODESET)" fails
- adding a check for this in the coercion logic means that
  coercion will happen on systems where this check succeeds,
  and will be skipped otherwise
- that way CPython should automatically adapt to changes in
  platform behaviour, rather than needing a new release to
  enable coercion at build time
- this also allows UTF-8 to be re-enabled as a coercion
  target, restoring the locale coercion behaviour on Mac OS X
This commit is contained in:
Nick Coghlan 2017-06-30 00:48:14 +10:00 committed by GitHub
parent f7d090c165
commit 18974c35ad
2 changed files with 29 additions and 15 deletions

View File

@ -1,6 +1,7 @@
# Tests the attempted automatic coercion of the C locale to a UTF-8 locale
import unittest
import locale
import os
import sys
import sysconfig
@ -32,24 +33,34 @@ else:
# In order to get the warning messages to match up as expected, the candidate
# order here must much the target locale order in Python/pylifecycle.c
_C_UTF8_LOCALES = ("C.UTF-8", "C.utf8") #, "UTF-8")
# XXX (ncoghlan): Using UTF-8 as a target locale is currently disabled due to
# problems encountered on *BSD systems with those test cases
# For additional details see:
# nl_langinfo CODESET error: https://bugs.python.org/issue30647
# locale handling differences: https://bugs.python.org/issue30672
_C_UTF8_LOCALES = ("C.UTF-8", "C.utf8", "UTF-8")
# There's no reliable cross-platform way of checking locale alias
# lists, so the only way of knowing which of these locales will work
# is to try them with locale.setlocale(). We do that in a subprocess
# to avoid altering the locale of the test runner.
#
# If the relevant locale module attributes exist, and we're not on a platform
# where we expect it to always succeed, we also check that
# `locale.nl_langinfo(locale.CODESET)` works, as if it fails, the interpreter
# will skip locale coercion for that particular target locale
_check_nl_langinfo_CODESET = bool(
sys.platform not in ("darwin", "linux") and
hasattr(locale, "nl_langinfo") and
hasattr(locale, "CODESET")
)
def _set_locale_in_subprocess(locale_name):
cmd_fmt = "import locale; print(locale.setlocale(locale.LC_CTYPE, '{}'))"
if _check_nl_langinfo_CODESET:
# If there's no valid CODESET, we expect coercion to be skipped
cmd_fmt += "; import sys; sys.exit(not locale.nl_langinfo(locale.CODESET))"
cmd = cmd_fmt.format(locale_name)
result, py_cmd = run_python_until_end("-c", cmd, __isolated=True)
return result.rc == 0
_fields = "fsencoding stdin_info stdout_info stderr_info lang lc_ctype lc_all"
_EncodingDetails = namedtuple("EncodingDetails", _fields)

View File

@ -399,17 +399,10 @@ typedef struct _CandidateLocale {
static _LocaleCoercionTarget _TARGET_LOCALES[] = {
{"C.UTF-8"},
{"C.utf8"},
/* {"UTF-8"}, */
{"UTF-8"},
{NULL}
};
/* XXX (ncoghlan): Using UTF-8 as a target locale is currently disabled due to
* problems encountered on *BSD systems with those test cases
* For additional details see:
* nl_langinfo CODESET error: https://bugs.python.org/issue30647
* locale handling differences: https://bugs.python.org/issue30672
*/
static char *
get_default_standard_stream_error_handler(void)
{
@ -490,6 +483,16 @@ _Py_CoerceLegacyLocale(void)
const char *new_locale = setlocale(LC_CTYPE,
target->locale_name);
if (new_locale != NULL) {
#if !defined(__APPLE__) && defined(HAVE_LANGINFO_H) && defined(CODESET)
/* Also ensure that nl_langinfo works in this locale */
char *codeset = nl_langinfo(CODESET);
if (!codeset || *codeset == '\0') {
/* CODESET is not set or empty, so skip coercion */
new_locale = NULL;
setlocale(LC_CTYPE, "");
continue;
}
#endif
/* Successfully configured locale, so make it the default */
_coerce_default_locale_settings(target);
return;