Re: (patch) regression diffs on collate.linux.utf8 test

On tis, 2011-10-18 at 15:21 -0400, Tom Lane wrote:
> Peter Eisentraut <peter_e@gmx.net> writes:
> > On tis, 2011-10-18 at 01:07 -0700, Jeff Davis wrote:
> >> If I qualify it as "tr_TR.UTF-8" it works. Perhaps I have something
> >> misconfigured on my system (Ubuntu 11.10)? I just installed:
> >> language-pack-de
> >> language-pack-tr
> >> language-pack-sv
> >> in an attempt to make the test work, and it works all except for that
> >> lc_time settng.
> 
> > I think the language-pack packages have nothing to do with it; they only
> > supply translations.
> 
> > Possibly, things are set up so that only UTF-8 locales are installed by
> > default.  Since the collate.linux.utf8 requires a UTF-8 environment, it
> > seems reasonable to use the tr_TR.UTF-8 locale for LC_TIME, instead of
> > requiring an unrelated (ISO-8859-9) locale to be installed.  So I think
> > the change you propose is reasonable.
> 
> As I said to Jeff earlier, I'd rather not embed assumptions about the
> spelling of encoding names into this test.  So I don't want to do this
> just to get rid of an unexplained failure.  I don't entirely believe
> the above theory, because it's not clear why Jeff's machine is behaving
> differently from mine.

Presumably because Jeff doesn't have that particular locale installed.
locale -a would clarify that.

glibc has always accepted variant locale name spellings such as "UTF-8"
vs "utf8", so it's not a problem.