Bug #78392 for Digest-Perl-MD5: Localize $_ to save naive modules from having it clobbered

Sun Jul 15 21:32:01 2012 WINTER [...] cpan.org - Ticket created

Subject:

Localize $_ to save naive modules from having it clobbered

Some modules (App::FatPacker) require their packages in such a way that modifying $_ without localizing it will break them. This patch just localizes $_ before reading from <DATA>. (I've also submitted a patch to App::FatPacker to have it stop doing that.)

Subject:

digest-perl-md5.patch

--- /opt/perl5/perls/perl-5.16.0/lib/site_perl/5.16.0/Digest/Perl/MD5.pm 2004-08-27 16:28:25.000000000 -0400 +++ Digest/Perl/MD5.pm 2012-07-15 21:27:31.000000000 -0400 @@ -61,6 +61,7 @@ ); my $insert = "\n"; + local $_; while(<DATA>) { chomp; next unless /^[FGHI]/;

Mon Jul 16 01:04:32 2012 $_ = 'spro^^*%*^6ut# [...] &$%*c>#!^!#&!pan.org'; y/a-z.@//cd; print - Correspondence added

On Sun Jul 15 21:32:01 2012, WINTER wrote: Show quoted text

> Some modules (App::FatPacker) require their packages in such a way that > modifying $_ without localizing it will break them. This patch just > localizes $_ before reading from <DATA>. (I've also submitted a patch to > App::FatPacker to have it stop doing that.)

Due to bugs in earlier perls, it’s best to avoid local($_), as it does not play along nicely with tied variables. You should use local *_ (but beware that @_ will no longer be accessible) or for(my $dummy) {...}

Mon Jul 16 01:04:32 2012 The RT System itself - Status changed from 'new' to 'open'

Mon Jul 16 01:20:16 2012 WINTER [...] cpan.org - Correspondence added

On Mon Jul 16 01:04:32 2012, SPROUT wrote: Show quoted text

> Due to bugs in earlier perls, it’s best to avoid local($_), as it does > not play along nicely with tied > variables. > > You should use local *_ (but beware that @_ will no longer be > accessible) or for(my $dummy) {...}

Ah, good to know! The subroutine in question (gen_code) doesn't take any arguments so local *_ should be safe here. I picked it because it was the most minimal change I could make to fix the problem. However, that said, looking again, the changes needed to use a variable there are small. I've attached an updated patch that avoids using local.

Subject:

no-local.patch

--- a/Digest/Perl/MD5.pm 2004-08-27 16:28:25.000000000 -0400 +++ b/Digest/Perl/MD5.pm 2012-07-16 01:16:42.000000000 -0400 @@ -61,10 +61,10 @@ ); my $insert = "\n"; - while(<DATA>) { - chomp; - next unless /^[FGHI]/; - my ($func,@x) = split /,/; + while(defined( my $data = <DATA> )) { + chomp $data; + next unless $data =~ /^[FGHI]/; + my ($func,@x) = split /,/, $data; my $c = $f{$func}; $c =~ s/X(\d)/$x[$1]/g; $c =~ s/(S\d{2})/$s{$1}/;

Thu Dec 12 15:48:51 2013 DOUGW [...] cpan.org - Correspondence added

On Mon Jul 16 01:20:16 2012, WINTER wrote: Show quoted text

> > I've attached an updated patch that avoids using local.

Spent some time tracking down this bug, glad to know there's already a patch. Looks good, can you please apply it? My specific case was a script with a for loop which created an object from a library which dynamically require'd Spreadsheet::ParseExcel, which used this library and wiped out $_. I usually don't use $_ in a loop that calls any sort of external functions, but sometimes I forget to be paranoid...

Sat Dec 14 08:27:05 2013 delta [...] lackas.net - Correspondence added

Resolved in 1.9.

Sat Dec 14 08:27:05 2013 delta [...] lackas.net - Status changed from 'open' to 'resolved'