You could also signal encoding via BOM, see =encoding _encodingname_, "=encoding encodingname"

Hi,

you're completely right. I do use =encoding and I get the right output of perldoc MyPackage. But the method abstract of App::Cmd::Command doesn't respect it when scanning for an abstract message for showing as help text. That's why I wrote earlier that this method could scan for a pod line =encoding and try to respect the given encoding. Have a look at the source code.

Best regards
McA

I asked rjbs on IRC and he said his preferred route was to just interpret all abstracts extracted from the source as being UTF-8, and forget about scanning for the encoding.

I've offered to provide him with a patch, but if anyone beats me to it, that's just fine by me.

package Cow { use Moo; has name => (is => 'lazy', default => sub { 'Mooington' }) } say Cow->new->name

I'm not really happy with that "solution", because if a perl source is not declared beeing utf8-encoded it is latin1-encoded by default. A German umlaut in an ABSTRACT would be decoded wrong.

My proposal from above:

--- App/Cmd/Command.pm  2013-01-31 01:37:50.000000000 +0100
+++ App/Cmd/Command-New.pm      2013-03-04 14:57:08.398638804 +0100
@@ -11,6 +11,7 @@
 # ABSTRACT: a base class for App::Cmd commands

 use Carp ();
+use Encode ();


 sub prepare {
@@ -115,6 +116,7 @@

   local $/ = "\n";
   my $inpod;
+  my $enc_seen;

   while (local $_ = <$fh>) {
     # =cut toggles, it doesn't end :-/
@@ -127,10 +129,10 @@

     next unless $inpod;
     chomp;
-
+    $enc_seen = $1 if(/^=encoding\s+(\S+)/);
     next unless /^(?:$class\s-\s)(.*)/;

-    $result = $1;
+    $result = $enc_seen ? Encode::decode($enc_seen, $1) : $1;
     last;
   }
[download]

Best regards
McA

:) rt://App-Cmd