Matthias Klose <d...@debian.org> writes:
> It would be good to have a self-contained example to show the
> exact issue.
Enable the "de_AT.UTF-8" locale in "dpkg-reconfigure locales",
copy the attached files to a directory, and run "make check"
there. The C version outputs "test" with glibc 2.19, but the
Python version fails with Python 3.4.2:
gcc -Wall -Wextra -o 840610 840610.c
mkdir -p de/LC_MESSAGES
msgfmt -o de/LC_MESSAGES/840610.mo de.po
LANGUAGE=de_AT.UTF-8 LC_ALL=C ./840610
test
LANGUAGE=de_AT.UTF-8 LC_ALL=C python3 ./840610.py
Traceback (most recent call last):
File "./840610.py", line 8, in <module>
print(gettext.dgettext('840610', 'test'))
UnicodeEncodeError: 'ascii' codec can't encode character '\xfc' in position 2:
ordinal not in range(128)
Makefile:21: recipe for target 'check' failed
make: *** [check] Error 1
You can also play with commands like
strace -E LANGUAGE=de_AT.UTF-8 -E LC_ALL=C -e file ./840610
strace -E LANGUAGE=de_AT.UTF-8 -E LC_ALL=C.UTF-8 -e file ./840610
which show that, if LC_ALL=C exactly, then glibc doesn't even
attempt to open any 840610.mo file. This is the special case
that was added in glibc 2.2.1 and has not been implemented
in the Python gettext library, as of Python 3.4.2.
#define _GNU_SOURCE 1
#include <errno.h>
#include <error.h>
#include <libintl.h>
#include <locale.h>
#include <stdio.h>
int
main(void)
{
if (setlocale(LC_ALL, "") == NULL)
error(1, errno, "setlocale");
if (bindtextdomain("840610", ".") == NULL)
error(1, errno, "bindtextdomain");
if (puts(dgettext("840610", "test")) == EOF)
error(1, errno, "puts");
return 0;
}
#! /usr/bin/python3
import locale
import gettext
locale.setlocale(locale.LC_ALL, '')
gettext.bindtextdomain('840610', '.')
print(gettext.dgettext('840610', 'test'))
msgid ""
msgstr ""
"Language: de\n"
"MIME-Version: 1.0\n"
"Content-Type: text/plain; charset=UTF-8\n"
"Content-Transfer-Encoding: 8bit\n"
msgid "test"
msgstr "Prüfung"
#! /usr/bin/make -f
PYTHON = python3
CC = gcc
CFLAGS = -Wall -Wextra
all: 840610 de/LC_MESSAGES/840610.mo
clean:
rm -f 840610
rm -rf de
840610: 840610.c
$(CC) $(CFLAGS) -o 840610 840610.c
de/LC_MESSAGES/840610.mo: de.po
mkdir -p de/LC_MESSAGES
msgfmt -o de/LC_MESSAGES/840610.mo de.po
check: all
LANGUAGE=de_AT.UTF-8 LC_ALL=C ./840610
LANGUAGE=de_AT.UTF-8 LC_ALL=C $(PYTHON) ./840610.py
