develooper Front page | perl.perl5.porters | Postings from May 2015

Re: FYI - PM posted about a horrible List::Util::first bug

Thread Previous | Thread Next
From:
Dave Mitchell
Date:
May 28, 2015 10:09
Subject:
Re: FYI - PM posted about a horrible List::Util::first bug
Message ID:
20150528100909.GR15483@iabyn.com
On Wed, May 27, 2015 at 09:30:22PM +0100, Paul "LeoNerd" Evans wrote:
> What's the conclusion here? Do I have to fix any code in ListUtils.xs
> then?

Well I think the provisional conclusion is that lc() et al are at fault.
As to whether List::Util is doing anything wrong,  the two differences
between the first() implementation and perl's grep() implementation are:

1) grep temporarily increases the ref count of each SV it aliases to $_,
via:

    # define DEFSV_set(sv) \
        (SvREFCNT_dec(GvSV(PL_defgv)), GvSV(PL_defgv) = SvREFCNT_inc(sv))

2) grep makes a mortal copy of each PADTMP it encounters.

Either of those would make List::Util work with the bad lc/uc/etc in 5.20.0
and 5.22.0.

The second of those was to fix

    #78194: Referencing a PADTMP twice produces two copies

(which I don't fully understand from a cursory read of the ticket).

The first, if nothing else, avoids crashes if the thing being iterated
is freed elsewhere, e.g.

    use List::Util qw{ first };
    use Devel::Peek;
    my $a = "A";
    @a = (1);
    @b = first { shift @a; Dump $_ } @a;

which outputs:

    SV = UNKNOWN(0xff) (0x1b19038) at 0x1af8f88
      REFCNT = 0
      FLAGS = ()




-- 
"Do not dabble in paradox, Edward, it puts you in danger of fortuitous wit."
    -- Lady Croom, "Arcadia"

Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About