Re: [PATCH v2 1/8] tests: Make sure LANG is set properly for iso9660_test

[Glenn Washburn]

On Fri, 27 Aug 2021 21:13:17 +0200
"Thomas Schmitt" <scdbackup@gmx.net> wrote:

> Hi,
> 
> it turns out that it might be contraproductive to set
> LANG=en_US.UTF-8.
> 
> If the LANG locale is not listed by
>   locale -a
> then nl_langinfo(3) returns "ANSI_X3.4-1968".
> I stumbled over this when testing LANG=de_DE.UTF8, which despite of
> my location is not an available locale on my machine.

My patch only sets LANG if empty, so in this case I don't think its
counterproductive, just not productive.

> So i think that setting any locale by LANG makes daring assumptions
> about its availability, unless brain is added to inspect the installed
> locals and to choose one that is based on UTF-8.
> This reaches quite far beyond the problem which the patch shall solve.
> 
> What route to go ?
> 
> - Choose the first UTF-8 locale offered by locale -a ?
>   (What to do if none is offered ?)
> 
> - Abandon the idea of setting LANG and rather use xorrisofs charset
>   options to enforce UTF-8 as origin of the conversion ?
> 
> I will for now strive for the second alternative but am ready to
> change to number one.

I think that regardless we should be using the xorrisofs charset. The
issue of LANG may or may not be an issue, I'm inclined to believe not.
I was only using this method to get the test working. The configuration
of locales and LANG should be a user problem, but ideally should not
affect how the tests are run. So in my judgement option 2 is the right
way forward. If there are other LANG related issues we may have to mess
with it in the future, but we can cross that bridge if we need to. I'm
looking forward to your patch and will make sure this one gets dropped.

Glenn