Warnings#

Non-fatal diagnostics. A warning prints one line to STDERR and the program runs on. The severe (S) warnings print even without the warnings pragma; the plain W warnings print only when warnings are enabled for their category:

use warnings;                    # all categories on
no warnings 'recursion';         # silence just deep-recursion
use warnings FATAL => 'utf8';    # make utf8 warnings die instead

The category in a (W category) badge is exactly the string you pass to no warnings.

Deep recursion#

`Deep recursion on subroutine "%s"`#

(W recursion) A subroutine has called itself (directly or through a chain) more than a hundred levels deep. This is usually a runaway recursion - a base case that never fires - rather than genuinely deep work. The message names the subroutine.

use warnings;

sub countdown {
    my $n = shift;
    countdown($n + 1);            # bug: should be $n - 1, never stops
}
countdown(1);
# Deep recursion on subroutine "main::countdown" at script line 5.

Fix the base case. If the recursion really is meant to run deep, silence the category for that scope:

no warnings 'recursion';

Wide characters#

`Wide character in %s`#

(S) A string containing a character above codepoint 255 was sent to a byte-oriented operation - most often print to a filehandle with no :encoding layer. Perl emits the bytes of the internal UTF-8 form and warns, because the output stream was never told it would carry Unicode.

use warnings;
print "café\x{2603}\n";          # Wide character in print at ... line 2.

The fix is to tell the stream its encoding:

binmode STDOUT, ':encoding(UTF-8)';
print "café\x{2603}\n";          # no warning; correct UTF-8 bytes

This is a severe (S) warning: it prints even without use warnings, because silently emitting raw internal bytes is almost always a bug.

Useless constructs#

`Useless use of \E`#

(W regexp) An \E in a pattern or string closes a \U, \L, \Q case/quote span that was never opened. The \E has nothing to end. Remove it, or add the opening \U/\L/\Q it was meant to close.

Group names#

`Invalid character in group name`#

(W) A regex named group ((?<name>...)) contains a character that is not allowed in a group name. Group names follow identifier rules - letters, digits, and underscore, not starting with a digit. The pattern still compiles, treating the construct leniently, but the name is suspect.

Encoding#

`Cannot find encoding "%s"`#

(S) A :encoding(...) layer or an Encode call named an encoding that is not available. Check the spelling against the encodings Encode supports (UTF-8, Latin-1, ASCII, and the rest).

Catch-all#

`Warning: something's wrong`#

(W) A generic warning the codegen raises for a condition that has no more specific message. The surrounding context names what triggered it.