The problem is not in closing fd 0
Let's look at the following sample:
# dup.pl
open my $fh, '<&=',0 or die;
print fileno STDIN, ":", scalar readline STDIN;
close STDIN;
print fileno $fh, ":", scalar readline $fh;
########
bash $ { echo 1; sleep 1; echo 2; } | perl dup.pl
0:1
0:2
(During this I've investigated, that if we do echo -ne "1\n2\n" | perl dup.pl,
then <$fh> receive nothing. I think that's because of buffered read. Don't know
should be this considered a bug)
Then we add the same
open my $fh, '<&=',0 or die;
to the beginning of the example and check the fileno at the end
0 == fileno $fh or die "fileno fh nonzero";
then we will see, that fileno 0 was kept, it still assigned to stdin (this
could be checked additionally), but the behavior of this program still the
same: we got an output of PerlIO.pm
On 08.12.2011, at 21:18, Father Chrysostomos via RT wrote:
> On Thu Dec 08 03:25:08 2011, [email protected] wrote:
>> #!/usr/bin/perl
>> use strict;
>> use warnings;
>>
>> my $data = "aaa\nbbb\nccc\n";
>>
>> # Ok
>> close STDIN or die $!;
>> open( STDIN, '<', \$data ) or die $!;
>> @ARGV = ();
>> while (<STDIN>)
>> {
>> print "[[$ARGV]]***$_";
>> }
>>
>> # Bug
>> close STDIN or die $!;
>> open( STDIN, '<', \$data ) or die $!;
>> @ARGV = ();
>> while (<>)
>> {
>> # print PerlIO.pod
>> print "[[$ARGV]]***$_";
>> }
>
> By closing STDIN you are closing fd 0, which is reused by the next file
> opened.
>
> I seem to remember this came up about three years ago and the consensus
> was that it is not a bug. I’m not sure.
>
> If my memory serves me correctly, searching for ‘you passed sub
> protections’ and ‘This gotta change’ in the archives will bring it up.
>
> --
>
> Father Chrysostomos
>
--
Moscow.pm mailing list
[email protected] | http://moscow.pm.org
