Re: non-ascii chars in octal in sub-shell windows

[Joseph Brenner <doom@kzsu.stanford.edu>]

Peter Dyballa <Peter_Dyballa@Web.DE> writes:
> Joseph Brenner:

>> When running a program that outputs utf-8 characters such as u-umlaut,
>> in a terminal window I'll see the actual character, but in an emacs
>> sub-shell I'm seeing the octal form (which looks like: \374).
>
> No, you're not running such a programme! The LATIN SMALL LETTER U
> WITH DIAERESIS, ü, is encoded in UTF-8 as C3BC. In UTF-16 it is
> 00FC –  exactly two bytes! Obviously your programme just outputs
> some ISO  Latin dialect or such...

Correct.

If anyone's interested in the details of the screw-up, here's some
off-topic chattering about perl programming:

A typical perl test script is based on the Test::More module,
which provides features to do checks such as:

  is_deeply( $some_structure, $expected_structure,
             "Testing whether structure is as expected.");

This routine outputs different messages to STDOUT and/or STDERR
depending on whether the check passes or fails.

I was seeing octal junk in those output messages, even after adding
some commands to the *.t script like so:

  binmode STDOUT, ':encoding(utf8)';
  binmode STDERR, ':encoding(utf8)';

Normally, that would be all it would take to convince perl it needs to
output UTF-8, in the case of Test::More routines, this approach fails,
because it creates new output handles of it's own.

Unbeknownst to me, the documentation for Test::More has been
recommending doing something more like this:

  my $builder = Test::More->builder;
  binmode $builder->output,         ":encoding(utf8)";
  binmode $builder->failure_output, ":encoding(utf8)";

Note that merely doing this sort of thing has no effect:
  use utf8;
  use locale;