Re: [PATCH] treewide: Dodge dynamic memory allocation in strerror() from glibc > 2.40

[Stefano Brivio <sbrivio@redhat.com>]

On Wed, 11 Dec 2024 11:36:20 +0100
Paul Holzinger <pholzing@redhat.com> wrote:

> On 11/12/2024 04:55, David Gibson wrote:
> >> I wonder if ignoring the locale is really that bad. After all, we print
> >> all the messages in English, without localisation. Printing the error
> >> description in other languages is arguably inconsistent.  
> > Hm, I guess.  Maybe we can tackle respecting locale for the kernel
> > library errors when/if we add localisation support for our own
> > messages.
>
> As German who has a few systems with a German locale configured I prefer 
> it when the locale is ignored in such cases. Getting a mixed 
> English/German error text always looks wrong and I much rather have it 
> only in English. But then again I understand English and these error 
> messages well enough already so users who do not might feel differently.

I'm a native speaker of Italian not picking Italian locales (so that
"others" understand, and to avoid witnessing abuses on Italian), but
setting LC_PAPER="en_GB.UTF-8" and LC_MEASUREMENT="en_GB.UTF-8" which
are otherwise wrong (nobody should send envelopes with fingers inside).

I find two small advantages of translated locales in error reports: in
some cases they might give you a rough indication of the user's
timezone, and they usually make me giggle. Oh, and I know how to say
"drain pipe" (not really "pipe") in a number of languages, by now.

Other than that they are hard to grep, possibly harder to translate by
an automatic agent, and they are inconsistent. For desktop environments,
distribution installers, "proper" user interfaces, I totally see the
inclusiveness advantage of translated messages. Otherwise not really.

-- 
Stefano