Bug #107905 for Encode: netbsd: encoding::_get_locale_encoding returns invalid encoding name (Encode

Wed Oct 21 11:18:33 2015 dolmen [...] cpan.org - Ticket created

Subject:

netbsd: encoding::_get_locale_encoding returns invalid encoding name (Encode < 2.73)

On netbsd with Encode 2.60:
    perl -w '-MEncode 2.05' -mencoding -e 'print encoding::_get_locale_encoding, $/'
    646

The "646" value comes directly from I18N::Langinfo

    perl -MI18N::Langinfo=langinfo,CODESET -e 'print langinfo(CODESET), $/'
    646

I think this is an issue that has been fixed in 2.73 (with my rewrite of encoding::_get_locale_encoding), however this is not enforced by tests.
https://github.com/dankogai/p5-encode/commit/4ed9ed08d2943ffc222863bc353a4a6e484c97db

Unfortunately I don't have direct access to a netbsd system to fully understand the issue (I found the issue through CPAN Testers).

So the next step would be to add a test that checks encoding::_get_locale_encoding returns either undef or the name of an encoding known to Encode.
That test would also dump the result of langinfo, the current locale (POSIX::setlocale) and locale related environment variables for diagnostics.

(this is something that I intend to do myself, but that may take months because of my lack of tuits)

--
Olivier Mengué - http://perlresume.org/DOLMEN

Thu Oct 22 10:10:06 2015 dolmen [...] cpan.org - Correspondence added

Here is some C code to check if the issue comes directly from nl_langinfo(CODESET) .

--
Olivier Mengué - http://perlresume.org/DOLMEN

Subject:

codeset.c

/* cc -o codeset codeset.c */ #include <stdio.h> #include <stdlib.h> #include <locale.h> #include <langinfo.h> void print_codeset(const char *locale_name) { printf("locale=%s\nCODESET=%s\n", locale_name, nl_langinfo(CODESET)); } int main() { const char *locale; printf("LC_ALL=%s\nLC_CTYPE=%s\n\n", getenv("LC_ALL"), getenv("LC_CTYPE")); locale = setlocale(LC_CTYPE, NULL); print_codeset(locale); puts("\nUsing locale...\n"); locale = setlocale(LC_CTYPE, ""); print_codeset(locale); }

Thu Oct 22 10:18:22 2015 dolmen [...] cpan.org - Correspondence added

Le 2015-10-21 17:18:33, DOLMEN a écrit :
Show quoted text

> perl -w '-MEncode 2.05' -mencoding -e 'print encoding::_get_locale_encoding, $/'
> 646

646 refers to ISO/IEC 646 which is similar to ASCII
https://en.wikipedia.org/wiki/ISO/IEC_646

Encode is already aware of this:
perl -MEncode=find_encoding -E 'say find_encoding("646")->name'
ascii

So _get_locale_encoding in Encode 2.73+ will properly return 'ascii'.

--
Olivier Mengué - http://perlresume.org/DOLMEN

Thu Oct 22 10:30:36 2015 dolmen [...] cpan.org - Correspondence added

Le 2015-10-21 17:18:33, DOLMEN a écrit :
Show quoted text

>
> So the next step would be to add a test that checks
> encoding::_get_locale_encoding returns either undef or the name of an
> encoding
> known to Encode.
> That test would also dump the result of langinfo, the current locale
> (POSIX::setlocale) and locale related environment variables for
> diagnostics.
>
> (this is something that I intend to do myself, but that may take
> months because
> of my lack of tuits)

I have a very bad memory.
I have already written this test myself some months ago and it is in the Encode distribution:
https://github.com/dankogai/p5-encode/blob/master/t/encoding-locale.t

So I'm closing the issue.

--
Olivier Mengué - http://perlresume.org/DOLMEN

Thu Oct 22 10:30:41 2015 dolmen [...] cpan.org - Status changed from 'new' to 'resolved'

Thu Oct 22 10:30:42 2015 dolmen [...] cpan.org - Fixed in 2.73 added

Bug #107905 for Encode: netbsd: encoding::_get_locale_encoding returns invalid encoding name (Encode < 2.73)