[gentoo-dev] et_EE locale and language of error messages

[gentoo-dev] et_EE locale and language of error messages

[Stefan Schweizer]

Hi,

there are at least two problems with how portage currently handles locales:

- Firstly some packages fail to build with obscure LC_* settings
The continuous stream of et_EE bugs is annoying: http://tinyurl.com/jsqzb

- and secondly I get my gcc output in german when I have a german locale
set. This makes it really hard to report bugs or the bugreports are useless
for most developers that do not understand the language.

Those problems cannot be easily workarounded since portage does not use
LC_ALL and LANG settings from /etc/make.conf

I propose to have the portage build environment set the language to English
or LC_ALL=C by default. That would significantly reduce the bugs with
unreadable error messages+ solve all the et_EE bugs at once.

One problem could be that packages depend on LC_* to install the correct
language. But that is a real bug then in my opinion, because ebuilds should
only honour LINGUAS and not LC_* during build time. Those bugs should be
detected and fixed.

What do you think? LC_ALL=C in portage or not?

- Stefan

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] et_EE locale and language of error messages

[Marc Hildebrand]

> What do you think? LC_ALL=C in portage or not?
> 
> - Stefan
> 

Well this problem (localized error messages) exists since I know linux 
and the solution has always been "use per user locale settings and keep 
LC_ALL=POSIX or =C as a system default".
Maybe we should just update the docs?

Otoh LC_ALL=C could help if you intend to use a .utf-8 locale as root, 
though. So if it does help solving bugs and causes no trouble, why not.

Cheers,

Marc
-- 
gentoo-dev@gentoo.org mailing list

[gentoo-dev] Re: et_EE locale and language of error messages

[Stefan Schweizer]

Marc Hildebrand wrote:
> Otoh LC_ALL=C could help if you intend to use a .utf-8 locale as root,
> though. So if it does help solving bugs and causes no trouble, why not.


ok, we have prepared a patch now, so everyone can have a look at it.
http://dev.gentoo.org/~zmedico/tmp/portage_lc_all.patch

the default LC_ALL=C will be overwriteable by PORTAGE_LC_ALL= in make.conf.
Of course broken settings like et_EE will not be supported.

If no other objections come up against patch will be added to portage within
a day. So, if you have any problem with it, speak up now.

- Stefan

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: et_EE locale and language of error messages

[Harald van Dijk]

On Fri, May 19, 2006 at 03:13:48PM +0200, Stefan Schweizer wrote:
> Marc Hildebrand wrote:
> > Otoh LC_ALL=C could help if you intend to use a .utf-8 locale as root,
> > though. So if it does help solving bugs and causes no trouble, why not.
> 
> 
> ok, we have prepared a patch now, so everyone can have a look at it.
> http://dev.gentoo.org/~zmedico/tmp/portage_lc_all.patch
> 
> the default LC_ALL=C will be overwriteable by PORTAGE_LC_ALL= in make.conf.
> Of course broken settings like et_EE will not be supported.

Different != broken. It's the packages that are broken with unwarranted
assumptions.

> If no other objections come up against patch will be added to portage within
> a day. So, if you have any problem with it, speak up now.

Yes, please do ignore my current objections.
-- 
gentoo-dev@gentoo.org mailing list

[gentoo-dev] Re: et_EE locale and language of error messages

[Duncan]

Stefan Schweizer <genstef@gentoo.org> posted e4kgai$j3n$1@sea.gmane.org,
excerpted below, on  Fri, 19 May 2006 15:13:48 +0200:

> Marc Hildebrand wrote:
>> Otoh LC_ALL=C could help if you intend to use a .utf-8 locale as root,
>> though. So if it does help solving bugs and causes no trouble, why not.
> 
> 
> ok, we have prepared a patch now, so everyone can have a look at it.
> http://dev.gentoo.org/~zmedico/tmp/portage_lc_all.patch
> 
> the default LC_ALL=C will be overwriteable by PORTAGE_LC_ALL= in
> make.conf. Of course broken settings like et_EE will not be supported.
> 
> If no other objections come up against patch will be added to portage
> within a day. So, if you have any problem with it, speak up now.

While I cautiously support the idea, limiting objections to a day appears
to be ramming something thru, and beyond that, simply isn't practical for
a number of folks who may have either valid objections or valid
suggestions that could ease potential future issues.  This is
/particularly/ the case when some may be ignoring the list ATM due to the
thread getting the most activity ATM.  If folks aren't interested in that,
they may be avoiding the list ATM as a simple matter of efficiency and
personal sanity.

Also, it's Friday.  What about those that have taken off for a long
weekend?

Please... Tuesday at the very earliest seems reasonable, and I'd suggest
an entire week from first post, so next Friday.  Also note that portage is
in 2.1-rc freeze ATM, and while this might be considered a bug-fix
exception, /some/ sort of debate, more than a day, seems warranted. 
Otherwise, 2.2 would look like a more reasonable target.

Again, this is from someone that otherwise favors the idea.  A day's
debate is simply not enough, and looks like you are just trying to ram it
thru without /letting/ there be debate.  Please reconsider at least on the
timing.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: et_EE locale and language of error messages

[Marius Mauch]

[-- Attachment #1: Type: text/plain, Size: 1245 bytes --]

On Fri, 19 May 2006 15:13:48 +0200
Stefan Schweizer <genstef@gentoo.org> wrote:

> Marc Hildebrand wrote:
> > Otoh LC_ALL=C could help if you intend to use a .utf-8 locale as
> > root, though. So if it does help solving bugs and causes no
> > trouble, why not.
> 
> 
> ok, we have prepared a patch now, so everyone can have a look at it.
> http://dev.gentoo.org/~zmedico/tmp/portage_lc_all.patch
> 
> the default LC_ALL=C will be overwriteable by PORTAGE_LC_ALL= in
> make.conf. Of course broken settings like et_EE will not be supported.
> 
> If no other objections come up against patch will be added to portage
> within a day. So, if you have any problem with it, speak up now.

Zac, you creating a 2.2 branch or what?
Why does this have to be fast tracked all of a sudden?

Oh, and just for completeness a few (historic) bug references:
http://bugs.gentoo.org/show_bug.cgi?id=9901
http://bugs.gentoo.org/show_bug.cgi?id=38418
http://bugs.gentoo.org/show_bug.cgi?id=57973
http://bugs.gentoo.org/show_bug.cgi?id=133758

-- 
Public Key at http://www.genone.de/info/gpg-key.pub

In the beginning, there was nothing. And God said, 'Let there be
Light.' And there was still nothing, but you could see a bit better.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[gentoo-dev] Re: Re: et_EE locale and language of error messages

[Stefan Schweizer]

Marius Mauch wrote:
> Why does this have to be fast tracked all of a sudden?

Because someone took care of it eventually and it will fix all the estonian
bugs at once + allow other-LC-people like me to file bugs without having to
run emerge again with LC_ALL=C.

And I think it should go in as soon as possible to ensure it gets enough
testing, when you do not like it you can always overwrite it :)

Also I would want to have it in the stable branch anyway because of
bugreports by first-time users who do not use the latest version of
portage. It is better to add it now while in pre-release phase than after
that when it is stable.

- Stefan

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: Re: et_EE locale and language of error messages

[Marius Mauch]

[-- Attachment #1: Type: text/plain, Size: 608 bytes --]

On Fri, 19 May 2006 19:27:18 +0200
Stefan Schweizer <genstef@gentoo.org> wrote:

> Also I would want to have it in the stable branch anyway because of
> bugreports by first-time users who do not use the latest version of
> portage. It is better to add it now while in pre-release phase than
> after that when it is stable.

We aren't in the pre-release phase anymore but in release testing.

Marius

-- 
Public Key at http://www.genone.de/info/gpg-key.pub

In the beginning, there was nothing. And God said, 'Let there be
Light.' And there was still nothing, but you could see a bit better.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] Re: et_EE locale and language of error messages

[Zac Medico]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Marius Mauch wrote:
> On Fri, 19 May 2006 15:13:48 +0200
> Stefan Schweizer <genstef@gentoo.org> wrote:
> 
>> Marc Hildebrand wrote:
>>> Otoh LC_ALL=C could help if you intend to use a .utf-8 locale as
>>> root, though. So if it does help solving bugs and causes no
>>> trouble, why not.
>>
>> ok, we have prepared a patch now, so everyone can have a look at it.
>> http://dev.gentoo.org/~zmedico/tmp/portage_lc_all.patch
>>
>> the default LC_ALL=C will be overwriteable by PORTAGE_LC_ALL= in
>> make.conf. Of course broken settings like et_EE will not be supported.
>>
>> If no other objections come up against patch will be added to portage
>> within a day. So, if you have any problem with it, speak up now.
> 
> Zac, you creating a 2.2 branch or what?
> Why does this have to be fast tracked all of a sudden?

I wouldn't want something like this to be fast tracked unless there is practically unanimous support for it.  There seems to be a reasonable argument against it though, so fast tracking would be inappropriate.  As truedfx noted on bug 133758, LC_ALL=C can be set in the ebuild environment via an entry in /etc/env.d/, so people already have an avenue to get the requested behavior if they want it.

Zac
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (GNU/Linux)

iD8DBQFEbgRJ/ejvha5XGaMRAk2vAKDMlGVV/CDGXPr73imPCZ5XjEgYYACff5ho
Zb9yFW3w7xazllBz/zkJdUE=
=0VAi
-----END PGP SIGNATURE-----
-- 
gentoo-dev@gentoo.org mailing list

[gentoo-dev] Re: et_EE locale and language of error messages

[Wiktor Wandachowicz]

Marc Hildebrand <zypher <at> gentoo.org> writes:

> > What do you think? LC_ALL=C in portage or not?
> > 
> > - Stefan
> 
> Well this problem (localized error messages) exists since I know linux 
> and the solution has always been "use per user locale settings and keep 
> LC_ALL=POSIX or =C as a system default".
> Maybe we should just update the docs?

A bug regarding this issue has been filled:
http://bugs.gentoo.org/show_bug.cgi?id=134032

>From the current state of portage code, the patch in question while useful is
not necessary. Portage already defines in /usr/lib/portage/portage_const.py:

USER_CONFIG_PATH        = "/etc/portage"
EBUILD_SH_ENV_FILE      = USER_CONFIG_PATH+"/bashrc"

which is then used in /usr/lib/pym/portage.py:

mysettings["PORTAGE_BASHRC"] = EBUILD_SH_ENV_FILE

and consequently in /usr/lib/bin/ebuild.sh:

if [ -f "${PORTAGE_BASHRC}" ]; then
        source "${PORTAGE_BASHRC}"
fi

So, if the documentation is updated (and possibly advertized in next GWN), no
change (even small) is necessary.

And while I can voice my (user's) opinion, I *really* like the idea of forcing
the sane emerge environment by this one-liner. So in fact I'm torn apart...

Wiktor Wandachowicz
(SirYes)


-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] et_EE locale and language of error messages

[Harald van Dijk]

On Fri, May 19, 2006 at 11:38:06AM +0200, Stefan Schweizer wrote:
> Hi,
> 
> there are at least two problems with how portage currently handles locales:
> 
> - Firstly some packages fail to build with obscure LC_* settings
> The continuous stream of et_EE bugs is annoying: http://tinyurl.com/jsqzb
> 
> - and secondly I get my gcc output in german when I have a german locale
> set. This makes it really hard to report bugs or the bugreports are useless
> for most developers that do not understand the language.
> 
> Those problems cannot be easily workarounded since portage does not use
> LC_ALL and LANG settings from /etc/make.conf
> 
> I propose to have the portage build environment set the language to English
> or LC_ALL=C by default. That would significantly reduce the bugs with
> unreadable error messages+ solve all the et_EE bugs at once.
> 
> One problem could be that packages depend on LC_* to install the correct
> language. But that is a real bug then in my opinion, because ebuilds should
> only honour LINGUAS and not LC_* during build time. Those bugs should be
> detected and fixed.
> 
> What do you think? LC_ALL=C in portage or not?

No, it's needlessly unfriendly to users, and encourages broken packages.
et_EE breakage should be fixed, and slowly but surely is, and as for
unreadable error messages, getting German gcc output in a German locale
is a feature, not a bug. It can indeed be a problem in bugreports, but
it's a much milder one, since it's trivial to look up what any
particular message is translated from.
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] et_EE locale and language of error messages

[Jakub Moc]

[-- Attachment #1: Type: text/plain, Size: 928 bytes --]

Harald van Dijk wrote:

> is a feature, not a bug. It can indeed be a problem in bugreports, but
> it's a much milder one, since it's trivial to look up what any
> particular message is translated from.

Well no, I completely disagree. Error output in $random language makes
searching for duplicate bugs very difficult/impossible, also doesn't
make fixing the bugs any easier....

When people report bugs, they need to report the errors in English. We
also don't allow bugs to be filed in $random language, I don't see why
error messages should be an exception.

Feel free to write a translation tool for error messages, until then
this is a no go... ;)

-- 
Best regards,

 Jakub Moc
 mailto:jakub@gentoo.org
 GPG signature:
 http://subkeys.pgp.net:11371/pks/lookup?op=get&search=0xCEBA3D9E
 Primary key fingerprint: D2D7 933C 9BA1 C95B 2C95  B30F 8717 D5FD CEBA 3D9E

 ... still no signature   ;)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] et_EE locale and language of error messages

[Patrick McLean]

> No, it's needlessly unfriendly to users, and encourages broken packages.
> et_EE breakage should be fixed, and slowly but surely is, and as for
> unreadable error messages, getting German gcc output in a German locale
> is a feature, not a bug. It can indeed be a problem in bugreports, but
> it's a much milder one, since it's trivial to look up what any
> particular message is translated from.

It's not trivial if you don't have the German locales installed.
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] et_EE locale and language of error messages

[Harald van Dijk]

On Fri, May 19, 2006 at 09:09:08AM -0400, Patrick McLean wrote:
> > No, it's needlessly unfriendly to users, and encourages broken packages.
> > et_EE breakage should be fixed, and slowly but surely is, and as for
> > unreadable error messages, getting German gcc output in a German locale
> > is a feature, not a bug. It can indeed be a problem in bugreports, but
> > it's a much milder one, since it's trivial to look up what any
> > particular message is translated from.
> 
> It's not trivial if you don't have the German locales installed.

grep through gcc/po/*, which doesn't require installation of the
locales, only that you didn't delete the gcc tarball after fetching it.
(Yes, I do it myself, and no, I don't normally have the gcc sources
unpacked.) Jakub, probably not very seriously, mentioned automating such
a lookup. Well, a simple shell script should do it, and if there is
actual interest in it, I might write it myself.
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] et_EE locale and language of error messages

[Carsten Lohrke]

[-- Attachment #1: Type: text/plain, Size: 403 bytes --]

On Friday 19 May 2006 15:24, Harald van Dijk wrote:
> grep through gcc/po/*, which doesn't require installation of the
> locales

Providing the error messages in english is part of what I consider the users 
job when filing a bug report. Having to grep for the strings is wasted time. 
I'm not so sure, if hiding compilation problems with other locales is a good 
idea, though.


Carsten



[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] et_EE locale and language of error messages

[Daniel Drake]

Harald van Dijk wrote:
> and as for
> unreadable error messages, getting German gcc output in a German locale
> is a feature, not a bug. 

I agree - but only when you use gcc on the command line, or in a 
Makefile, or in some other normal usage scenario.

I think Stefan is suggesting just using the standard English locale 
*during portage merges* rather than as a system-wide thing.

Does that change your view at all?

Daniel

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] et_EE locale and language of error messages

[Harald van Dijk]

On Fri, May 19, 2006 at 02:44:03PM +0100, Daniel Drake wrote:
> Harald van Dijk wrote:
> >and as for
> >unreadable error messages, getting German gcc output in a German locale
> >is a feature, not a bug. 
> 
> I agree - but only when you use gcc on the command line, or in a 
> Makefile, or in some other normal usage scenario.
> 
> I think Stefan is suggesting just using the standard English locale 
> *during portage merges* rather than as a system-wide thing.
> 
> Does that change your view at all?

I know that's what he's suggesting, so no, it doesn't change my view.
-- 
gentoo-dev@gentoo.org mailing list

[gentoo-dev] Re: et_EE locale and language of error messages

[Stefan Schweizer]

Harald van Dijk wrote:
> [..] encourages broken packages.
> et_EE breakage should be fixed, and slowly but surely is[..]


That is your main problem here and I have discussed this in IRC with you and
it is true in my opinion that it does not improve gentoo or make the
distribution any better to close et_EE bugs as WONTFIX.

But the default should be C to make sure everyone can compile their system
without an error, and if people want locale output and help with bugfixing
they can of course enable it in their make.conf and keep the bugs coming :)

Also developers wont have to translate that often or ask for translations
when the default locale is C. Developers are more likely to fix a bug
quickly if they do not have to ask twice :)

Kind regards,
Stefan

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: et_EE locale and language of error messages

[Harald van Dijk]

On Fri, May 19, 2006 at 05:44:34PM +0200, Stefan Schweizer wrote:
> Harald van Dijk wrote:
> > [..] encourages broken packages.
> > et_EE breakage should be fixed, and slowly but surely is[..]
> 
> 
> That is your main problem here and I have discussed this in IRC with you and
> it is true in my opinion that it does not improve gentoo or make the
> distribution any better to close et_EE bugs as WONTFIX.
> 
> But the default should be C to make sure everyone can compile their system
> without an error, and if people want locale output and help with bugfixing
> they can of course enable it in their make.conf and keep the bugs coming :)
> 
> Also developers wont have to translate that often or ask for translations
> when the default locale is C. Developers are more likely to fix a bug
> quickly if they do not have to ask twice :)

Indeed, this is less than ideal, but not unacceptable to me :)
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] et_EE locale and language of error messages

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 378 bytes --]

On Friday 19 May 2006 05:38, Stefan Schweizer wrote:
> I propose to have the portage build environment set the language to English
> or LC_ALL=C by default. That would significantly reduce the bugs with
> unreadable error messages+ solve all the et_EE bugs at once.

this is complete garbage

you're simply ignoring the problem because it's easier than fixing it properly
-mike

[-- Attachment #2: Type: application/pgp-signature, Size: 827 bytes --]

Re: [gentoo-dev] et_EE locale and language of error messages

[Kevin F. Quinn]

[-- Attachment #1: Type: text/plain, Size: 2651 bytes --]

On Fri, 19 May 2006 11:38:06 +0200
Stefan Schweizer <genstef@gentoo.org> wrote:

> Hi,
> 
> there are at least two problems with how portage currently handles
> locales:
> 
> - Firstly some packages fail to build with obscure LC_* settings

Anything that expects ordering etc from a particular locale, but does
not set the locale, is broken.  It's nothing to do with obscure; I've
seen bugs occur due to people setting their default locale to
en_US.UTF-8 which can hardly be called obscure.  Also calling the
Estonian locale "obscure" is just rude.

> The continuous stream of et_EE bugs is annoying:
> http://tinyurl.com/jsqzb

These should be fixed by setting the correct locale prior to the
commands that are sensitive to it.  Such fixes should be sent or
negotiated with upstream.

> - and secondly I get my gcc output in german when I have a german
> locale set. This makes it really hard to report bugs or the
> bugreports are useless for most developers that do not understand the
> language.

My preferred  approach here, is that if the people working the bug don't
understand the report (including wranglers of course) simply resolve it
"NEEDINFO" requesting the user translate the error messages if
they can, or perhaps do 'export LC_ALL=C' before emerge and
repost the results.  Note that setting the locale to C may actually
cause a change in behaviour, perhaps even preventing the bug from
occurring if the bug is locale-specific, so obtaining a translation is
better.

> Those problems cannot be easily workarounded since portage does not
> use LC_ALL and LANG settings from /etc/make.conf
>
> I propose to have the portage build environment set the language to
> English or LC_ALL=C by default. That would significantly reduce the
> bugs with unreadable error messages+ solve all the et_EE bugs at once.

The 'et_EE' bugs would not be solved, they would be hidden.  It's much
better to inform upstream that they've assumed the C locale; those I've
interacted with on this issue so far have been happy to be informed and
have set the locale where they need to.

> One problem could be that packages depend on LC_* to install the
> correct language. But that is a real bug then in my opinion, because
> ebuilds should only honour LINGUAS and not LC_* during build time.
> Those bugs should be detected and fixed.

I disagree.  LINGUAS is a Gentoo-specific thing, so is only relevant to
ebuilds.  If a package uses LC_* to determine the user's locale
preferences, I see no problem with that.

> What do you think? LC_ALL=C in portage or not?

I vote not.

-- 
Kevin F. Quinn

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [gentoo-dev] et_EE locale and language of error messages

[Harald van Dijk]

On Wed, May 24, 2006 at 12:36:17PM +0200, Kevin F. Quinn wrote:
> On Fri, 19 May 2006 11:38:06 +0200
> Stefan Schweizer <genstef@gentoo.org> wrote:
> > One problem could be that packages depend on LC_* to install the
> > correct language. But that is a real bug then in my opinion, because
> > ebuilds should only honour LINGUAS and not LC_* during build time.
> > Those bugs should be detected and fixed.
> 
> I disagree.  LINGUAS is a Gentoo-specific thing, so is only relevant to
> ebuilds.  If a package uses LC_* to determine the user's locale
> preferences, I see no problem with that.

LINGUAS is not Gentoo-specific at all. Gentoo didn't even come up with
it. It's a gettext variable, which is re-used (with an annoyingly
slightly different meaning) by Gentoo.

That said, if a package can only install one translation at a time, and
it detects this at compile-time via LC_*, it's sort of braindead at
least. (A creates a package for B. B would like the Italian version. A
does not know any Italian. There is a build error. Because the system
forced LC_* to be set to Italian, A has no idea what the errors mean.)
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] et_EE locale and language of error messages

[Kevin F. Quinn]

[-- Attachment #1: Type: text/plain, Size: 833 bytes --]

On Wed, 24 May 2006 13:17:54 +0200
Harald van Dijk <truedfx@gentoo.org> wrote:

> LINGUAS is not Gentoo-specific at all. Gentoo didn't even come up with
> it. It's a gettext variable, which is re-used (with an annoyingly
> slightly different meaning) by Gentoo.

Thanks for the clarification - I only looked through the gnu libc info
pages for it (since we had been talking about locales) so I didn't
find it.

> That said, if a package can only install one translation at a time,
> and it detects this at compile-time via LC_*, it's sort of braindead
> at least. (A creates a package for B. B would like the Italian
> version. A does not know any Italian. There is a build error. Because
> the system forced LC_* to be set to Italian, A has no idea what the
> errors mean.)

Fair enough.  

-- 
Kevin F. Quinn

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [gentoo-dev] et_EE locale and language of error messages

[Jakub Moc]

[-- Attachment #1: Type: text/plain, Size: 3982 bytes --]

Kevin F. Quinn wrote:
> On Fri, 19 May 2006 11:38:06 +0200
> Stefan Schweizer <genstef@gentoo.org> wrote:
> 
>> Hi,
>>
>> there are at least two problems with how portage currently handles
>> locales:
>>
>> - Firstly some packages fail to build with obscure LC_* settings
> 
> Anything that expects ordering etc from a particular locale, but does
> not set the locale, is broken.  It's nothing to do with obscure; I've
> seen bugs occur due to people setting their default locale to
> en_US.UTF-8 which can hardly be called obscure.  Also calling the
> Estonian locale "obscure" is just rude.
> 
>> The continuous stream of et_EE bugs is annoying:
>> http://tinyurl.com/jsqzb
> 
> These should be fixed by setting the correct locale prior to the
> commands that are sensitive to it.  Such fixes should be sent or
> negotiated with upstream.
> 
>> - and secondly I get my gcc output in german when I have a german
>> locale set. This makes it really hard to report bugs or the
>> bugreports are useless for most developers that do not understand the
>> language.
> 
> My preferred  approach here, is that if the people working the bug don't
> understand the report (including wranglers of course) simply resolve it
> "NEEDINFO" requesting the user translate the error messages if
> they can, or perhaps do 'export LC_ALL=C' before emerge and
> repost the results.  Note that setting the locale to C may actually
> cause a change in behaviour, perhaps even preventing the bug from
> occurring if the bug is locale-specific, so obtaining a translation is
> better.

Doing that all the time. Bug reports with errors in any other language
than English are genuine PITA when searching for duplicates. Bugs with
errors in Russian/Estonian/Swedish/Chinese/... (add your favorite
language you can't even say 'good morning' in here) are useless for most
developers out there. Even if you can understand the language, the
translation often sucks, also - you'll get way more references when
googling for some error in English compared to any other locale out there.

Every other user that is asked to provide the messages with locales set
to C comes back asking how do I do it. Then they come back saying, oh
LC_ALL=C, or LC_MESSAGES=C didn't work, where do I put it exactly? Then
they come back and say, oh, it still didn't work... Also, asking someone
to provide errors in English when they occured say 10 hours into OO.org
compile makes the user really happy, of course. :P

Please, make portage spit out the errors in English by *default* to not
waste people's time. If someone insists on overriding it, then let them
do it in make.conf, but the default should be English and English should
be required when filing bugs (unless they are locale-specific ones).

I don't care about the amount of et_EE bugs, however they are almost
always upstream ones and I don't see that Gentoo devs would be keen on
fixing them - of course it's very low priority and there's more
important stuff to do in the limited time devs can spend on Gentoo
stuff, also users can emerge the thing just fine just by unsetting that
locale.

>> One problem could be that packages depend on LC_* to install the
>> correct language. But that is a real bug then in my opinion, because
>> ebuilds should only honour LINGUAS and not LC_* during build time.
>> Those bugs should be detected and fixed.
> 
> I disagree.  LINGUAS is a Gentoo-specific thing, so is only relevant to
> ebuilds.  If a package uses LC_* to determine the user's locale
> preferences, I see no problem with that.

Nope, LINGUAS is not Gentoo-specific, it's pretty much standard thing
for anything gettext-based.



-- 
Best regards,

 Jakub Moc
 mailto:jakub@gentoo.org
 GPG signature:
 http://subkeys.pgp.net:11371/pks/lookup?op=get&search=0xCEBA3D9E
 Primary key fingerprint: D2D7 933C 9BA1 C95B 2C95  B30F 8717 D5FD CEBA 3D9E

 ... still no signature   ;)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] et_EE locale and language of error messages

[Kevin F. Quinn]

[-- Attachment #1: Type: text/plain, Size: 1153 bytes --]

On Wed, 24 May 2006 13:42:37 +0200
Jakub Moc <jakub@gentoo.org> wrote:

> Please, make portage spit out the errors in English by *default* to
> not waste people's time. If someone insists on overriding it, then
> let them do it in make.conf, but the default should be English and
> English should be required when filing bugs (unless they are
> locale-specific ones).

ok.  I know we're repeating stuff discussed on IRC last week, but I
think it's useful to put this sort of thing on the record.

I'll go with having English as a default (rather than environment
locale) with an override settable in make.conf - I think it's important
that people have at least the option of emerge logs providing error
messages in their native language.  Also there could well be subtle
bugs that trip on a non-C locale, that we never notice; that the ones
highlighted when the locale is et_EE are fatal is just luck.


On the subject of timing elsewhere in the thread, I don't see there's
any hurry, as things have been as they are for a long time.  However
that's a decision for the people who plan out portage releases.

-- 
Kevin F. Quinn

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [gentoo-dev] et_EE locale and language of error messages

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 538 bytes --]

On Wednesday 24 May 2006 07:42, Jakub Moc wrote:
> Every other user that is asked to provide the messages with locales set
> to C comes back asking how do I do it. Then they come back saying, oh
> LC_ALL=C, or LC_MESSAGES=C didn't work, where do I put it exactly? Then
> they come back and say, oh, it still didn't work... Also, asking someone
> to provide errors in English when they occured say 10 hours into OO.org
> compile makes the user really happy, of course. :P

update the bugzilla howto guide then and point them to that
-mike

[-- Attachment #2: Type: application/pgp-signature, Size: 827 bytes --]

Re: [gentoo-dev] et_EE locale and language of error messages

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 822 bytes --]

On Wednesday 24 May 2006 07:42, Jakub Moc wrote:
> Please, make portage spit out the errors in English by *default* to not
> waste people's time. If someone insists on overriding it, then let them
> do it in make.conf

and where do you propose we document said override ?  in the locale guide ?  
then the only we've accomplished is making users set the samething in two 
different places

or perhaps we obscure it in a manpage (or even better, dont document it at 
all) ?  then we field bugs/forums/mailing lists with confused users who cant 
get portage output in their native language

just get Chris White to update his little bugzilla document with some steps on 
how to get english output and we close bugs as NEEDINFO telling them to 
review the document and open with the proper information
-mike

[-- Attachment #2: Type: application/pgp-signature, Size: 827 bytes --]