Yes, that's right. I'd already planned to patch Fatal in bleadperl. That'll likely be a few days away, at least.

Seems to be a bug in context evaluation in FATAL, this code works and throws an exception when readdir returns undef:

#!/usr/bin/perl

use strict;
use warnings;

use Fatal qw(readdir);

$| = 1;

my $start_dir = '.';
opendir(my $dir, $start_dir);
while($_= readdir($dir)) {
        print "$_ ";
}
closedir $dir;
[download]

Update 2006/04/28: This apparently isn't a bug in Fatal, but simply the fact that readdir can't be correctly "fatalized" 'cause it doesn't clearly indicates failure (and only failure) returning a false value. The documentation of Fatal will be updated to reflect better when to use it.

$ perl -MFatal+qw/open close/ -e 'open( F, "fasel"); print <F>;close F
+;'
Can't open perl script "close/": No such file or directory
[download]

You should reread the error message you got there.

perl -MFatal+qw/open close/ -e 'open( F, "fasel"); print <F>;close F;'
Can't open perl script "close/": No such file or directory
[download]

should be written

perl -MFatal=open,close -e 'open( F, "fasel"); print <F>;close F;'
[download]

or on win32:

perl -MFatal=open,close -e "open( F, q(fasel)); print <F>; close F;"
[download]

which produces

Can't open(F, fasel): No such file or directory at (eval 1) line 3
        main::__ANON__('F', 'fasel') called at -e line 1
[download]

---
$world=~s/war/peace/g

On Thu, Mar 23, 2006 at 02:03:57PM +0100, Rafael Garcia-Suarez wrote:
> Tom Hukins (via RT) wrote:
> > Fatal.pm and readdir() do not play well together.  Run the followi
+ng
> > script in a directory containing several files.  It only prints ou
+t '.'.
> > Then comment out the 'use Fatal' line and run the script again.  I
+t now
> > prints out all the files in the directory.
> > 
> > #!/usr/bin/perl
> > 
> > use strict;
> > use warnings;
> > 
> > use Fatal qw(readdir);
> > 
> > my $start_dir = '.';
> > opendir(my $dir, $start_dir);
> > my [at]subdir = readdir $dir;
> > closedir $dir;
> > print "@subdir\n";
>
[download]

It's interesting that this wasn't noticed before as not only have at least two regular p5p contributors replied in this thread but the original author of the bug report can occasionally be found here. I suppose it just goes to show how uncritical we all are when we don't suspect something might be afoot.

/J\

See also Re: Package level scope callbacks.