Question

The following code:

#!/usr/bin/env perl
use utf8;
use strict;
use warnings;
use 5.012; # implicitly turn on feature unicode_strings
my $test = "some string";
$test =~ m/.+\x{2013}/x;

Yields:

Use of uninitialized value $test in pattern match (m//) at test.pl line 9.

This seems to happen with any 2-byte character inside \x{}. The following regexes work fine:

/a+\x{2013}/
/.*\x{2013}/
/.+\x{20}/

Also, the error goes away with use bytes, but using that pragma is discouraged. What’s going on here?

Answer 1

It is singular that you should ask this question. I looks related to a bug that I just reported yesterday

https://rt.perl.org/rt3/Ticket/Display.html?id=114808

where this code also produces "Use of uninitialized value $_ in split ..." warnings, and causes split to unexpectedly return an empty list:

use warnings;
binmode *STDOUT, ":encoding(UTF-8)";
my $pattern = "\x{abc}\x{def}ghi";
for ( "\x{444}", "norm\x{a0}l", "\x{445}", "ab\x{ccc}de\x{fff}gh" ) {
  print "--------------------\ntext is $_, pattern is /$pattern/\n";

  # expect  split  to return  ($_) , but when $pattern and $_ both
  # have wide chars, it returns  ()
  print 'split output is [', split /$pattern/, $_;

  print "]\n";
}