Front page | perl.perl5.porters |
Postings from April 2000
From: Tom Christiansen [mailto:tchrist@chthon.perl.com]
> >Could you give an example of the usage of the code that illustrates the
> >problem please Tom.
>
> Right, sorry. It was snipped from yesterday's pathedit program.
>
> #!/usr/bin/perl
>
> use warnings;
> no warnings 'redefine';
>
> use Carp;
>
> my @names = qw(/tmp/foo /tmp/bar);
> my $code = q{$a + $b};
>
> print "Trying compile with fatal warnings\n\n";
> eval {
> compile1($code);
> pathedit() for @names;
> };
> warn if $@;
>
> print "\nTrying compile with sigwarn\n\n";
> eval {
> compile2($code);
> pathedit() for @names;
> };
> warn if $@;
>
> sub compile1 {
> my $op = shift;
> *pathedit = eval qq{
> sub () {
> use warnings qw/FATAL all/; # XXX: does not work
> $op;
> }
> } || do {
> local $_ = $@;
> s/at \(eval.*//s;
> die "$0: can't compile perlexpr $op: $_\n";
> }
> }
>
> sub compile2 {
> my $op = shift;
> *pathedit = eval qq{
> sub () {
> local \$SIG{__WARN__} = sub {
> local \$_ = "\@_";
> s/at \\(eval.*//;
> confess "FATAL WARNING: \$_";
> };
> $op;
> }
> } || do {
> local $_ = $@;
> s/at \(eval.*//s;
> die "$0: can't compile perlexpr $op: $_\n";
> }
> }
>
> That produces this:
>
> Trying compile with fatal warnings
>
> Use of uninitialized value in addition (+) at (eval 1) line 4.
> Use of uninitialized value in addition (+) at (eval 1) line 4.
> Use of uninitialized value in addition (+) at (eval 1) line 4.
> Use of uninitialized value in addition (+) at (eval 1) line 4.
>
> Trying compile with sigwarn
>
> FATAL WARNING: Use of uninitialized value in addition (+)
> main::__ANON__('Use of uninitialized value in addition (+) at
(eval 2) line 8.^J') called at (eval 2) line 8
> main::__ANON__() called at /tmp/p line 21
> require 0 called at /tmp/p line 19
> ...caught at /tmp/p line 23.
>
> The question is: shouldn't the fatal warnings version also be, well,
fatal?
Indeed they should. The logic that processes "FATAL all" is broken. Expect a
patch in the next couple of days.
Paul