I'd look at $login and $quotedLogin. When perl encounters that warning at some point later in an elsif ladder, it'll report it as happening on the line where the if starts it all, even if it's much later.

Updated with an example:

my $def1 = 'true';
my $def2 = 'foo';
my $def3 = 'blue';
my $undef1;

if ( $def1 =~ /X/ ) {   # <-- problem reported here
    print 'not here';
}
elsif ( $def2 =~ /Y/ ) {
    print 'not here either';
}
elsif ( $undef1 =~ /$def3/ ) { # <-- problem exists here
    print 'oh noes';
}
else {
    print "happiness ensues\n";
}
[download]

<AOL>Me too!</AOL>
I think kyle nailed it. Put:

use Data::Dumper; print Dumper([$login,$quotedLogin,$password]);
[download]

on line 22 and see the results...
[]s, HTH, Massa (κς,πμ,πλ)

Dear kyle, moritz,

The line number displayed for the warning was indeed misleading me. It actually occurs because of the undefined $email argument.

By the way, Fletch remark was pertinent: the warning was related to the "eq" operator, not to the "m//" one.

Thank you for your help.
Best regards,
-- Jeremie

I'm very suspicious as to this actually being the code causing that message since there's no use of eq in it anywhere; I'd expect to see Use of uninitialized value in pattern match (m//) at ... (or Use of uninitialized value in regexp compilation at ..., depending on the side of the =~ the undef is showing up) instead.

Update: Added other possible expected error message.

Perl 5.10.0 has the nice feature that it tells you exactly which variable is actually undefined, so running your script under perl 5.10.0 might shed some light on your problem.

It might also help to give us an example password (not a real one ;-) ) that throws the "uninitialized" warning.

Moritz, thank you! You have saved my sanity.

Also warnings are reported for the first line in which an if starts, and warnings from subsequent conditions (sometimes?) wrongly report to be on that same line.

I was getting the exact same error, and searching fruitlessly for the answer hence landing here, and sure enough, it was a typo farther down the eslif's and not at the line in question. I thought I was going nuts there. :) (running 5.8.5 on solaris 9 if anyone cares)

I ran the following code against this subroutine:

foreach ('test_NO PASSWORD', 'test_PASSWORD', 'test_test', 'test', 'te
+st_abc???????def') {
    print "$_: ", getPasswordStatus(split /_/), "\n";
}
[download]

test_NO PASSWORD:     empty
test_PASSWORD:        password
test_test:            login
test:                 ok
test_abc???????def:   ok
test_A gOoD PASsword: weak
[download]

1. an empty password is OK
2. a password with 7 consecutive question marks is OK
3. all other passwords are weak or are unacceptable for other reasons.

Hey CountZero,

Thanks for giving me a so high place in your ranking :D.

Actually the password argument is the result of the cracking phase. It should have been given a better name, I admit but I'm not the author of this code.

I hope this enlightens your mind :).

Regards,
-- Jeremie

If you haven't the answer you need yet post an example showing the error. You've given us a good test and bad test case, but plugging those into the code you have provided doesn't demonstrate the error you assert you are getting.

CountZero's reply Re: Puzzling warning "Use of unitialized value" in very simple code gives a good framework for the test. Using 'test_Alcatel1', 'test_Algerie1' in the test list reports both passwords as weak, but no errors or warnings (as expected).

Trying trimming down the relevant bits of code to the point where you have just enough stand alone code using the two test cases provided to demonstrate the problem. If you've not found the problem by then come back and show us what you've got.

http://rt.perl.org/rt3/Ticket/Display.html?id=37302
http://rt.perl.org/rt3/Ticket/Display.html?id=42664

Your real error is at line 28 of course. undefined $quotedLogin