Strange behaviour (bug?) of use strict 'subs' (was: ... strict 'vars')

[Hofmator]

Just the other night I came across the following. Consider

use strict;
use warnings;

my %foo;
@foo{1..3} = ();
print keys %foo;
@foo{qw/1 2 3/} = ();
print keys %foo;

both ways to assign to a hash slice work fine and produce the same results as expected.

Now add the following line to it:die "strange .." if foo eq 'bar'; which should make perl barf a "use of bareword 'foo' " error - as we are under strict. But it doesn't!! At least for me. I tried this under Perl 5.6.0 (ActivePerl Build 623) on Windows and Perl 5.6.1 on Linux (2.2.14).

If you remove the first hash slice line (@foo{1..3} = ();) then it works (or better not works i.e. throws the bareword error) correctly.

Well, so my questions are:

• can anyone reproduce that behaviour (on different versions)?
• can anyone explain it?
• is it a (known?) bug?

-- Hofmator

2002-06-23 Edit by Corion : Changed title

[particle]

it seems to me the problem is only found after using the range operator. also notice the bareword can be any word, it does not need to be in an existing glob.

> perl -w -Mstrict -e"my%f;++@f{(1,2)};die'ack'if foo eq 'bar'"
Bareword "foo" not allowed while "strict subs" in use at -e line 1.
Execution of -e aborted due to compilation errors.

> perl -w -Mstrict -e"my%f;++@f{qw/1 2/};die'ack'if foo eq 'bar'"
Bareword "foo" not allowed while "strict subs" in use at -e line 1.
Execution of -e aborted due to compilation errors.

> perl -w -Mstrict -e"my%f;++@f{1..2};die'ack'if foo eq 'bar'"

>

(ActiveState Perl 5.6.1, Win32)

a look at the code using B::Terse might help a little...

Read more... (2 kB)

~Particle *accelerates*

[ariels]

It looks like a bug. I get the buggy behaviour on 5.6.1 (<samp>This is perl, v5.6.1 built for sun4-solaris</samp>), but not on 5.005_03 (<samp>This is perl, version 5.005_03 built for sun4-solaris</samp>).

I think we need abigail-II...

Update

I get the same results (no error message, i.e. the bug is exercised) with v5.6.1 on i686-linux, IP27-irix and alpha-dec_osf.

(Temporary) update²

particle suggests I mention, until the editors fix the node title, that it's really strange behaviour of use strict 'subs': it relates to the use of a bareword.

[Anonymous Monk]

Behavior confirmed in 5.8.0-RC2 (linux) as well. No bare word warnings anywhere if a list-context range op is used in a statement prior to the bare words.

use strict;
use warnings;
print 1..2,"\n";
die 'bug' if foo eq bar;

[Joost]

Hmm... My Perl 5.8.0 RC 2 (i686-linux-thread-multi-64all-ld) on linux (RedHat 7.3) gives a Bareword "foo" not allowed while "strict subs" in use at - line 5. on that code. It also raises an error on the orgininal post when use strict is in effect.

Can anyone test this on another 5.8.0 RC 2 ?

-- 
Joost       downtime n. The period during which a system
            is error-free and immune from user input.

[Anonymous Monk]

My mistake. It appears I wasn't calling the perl I thought I was. 5.8.0-RC2 (linux) does raise exceptions on the code just as Joost observes. My apologies for any confusion.

[little]

Mh, funny.


# condition will not be true but raises no error
if (defined foo) {
    print ref(foo);
};

# raises an error as foo is marked as sub
die "strange .." if foo() eq 'bar';

# also
die "strange .." if (foo() eq 'bar');

# and even this does not show up _any_ errors or warnings
#!/usr/bin/perl -w
use strict;
use diagnostics;
.
.
.
print ref(foo);

# So is foo something special ?
# No, no error or warning also
# when replacing above line simply with
print ref(bar);

So, I have no idea either what perl is assuming here.

Read more... (2 kB)

Have a nice day
All decision is left to your taste
Update
Changed title to ".. use strict" as it seems not clear which pragma is conflicted, could be strict 'refs' as well
Update 2
even more funny, as this is different from particle's command line call in the use of -m instead of -M.

C:\web\>perl -w -mstrict -e"my %f;++@f{qw/1 2/};die'ack'if foo eq 'bar';"
Unquoted string "foo" may clash with future reserved word at -e line 1.

Just to show why I think that it's not clear which pragma is conflicted. :-)
addendumAristotle is correct, as here the strict pragma does not import any pragmas and is disabled. So at least its a prove as he said.

[Aristotle]

No, it's not. -m equates to use MODULE qw(); (see perlrun) which means -mstrict disables all strictures. Further, it actually demonstrates that it is the subs stricture that is conflicted, because enabling it causes Perl to skip the (then obviously redundant) bareword warnings.

Makeshifts last the longest.