Bug#158539: perl: segfault while doing odd things with glob refs

[Dominic Hargreaves]

On Tue, Aug 27, 2002 at 09:38:59PM +0100, Andrew Suffield wrote:
> I have no idea how wide-ranging or serious this is, nor do I really
> understand what is going on here.
> 
> aps100 at cyclone:~$ cat foo.pl
> use IO::Tee;
> 
> open FOO, ">foo";
> my $tee = IO::Tee->new(\*STDOUT, \*FOO);
> *STDOUT = $tee;
> print $tee "foo\n";
> aps100 at cyclone:~$ perl foo.pl
> Segmentation fault (core dumped)
> aps100 at cyclone:~$ 
> 
> IO::Tee is available from CPAN. Everything is fine unless I try to set
> *STDOUT = $tee, after which writes to STDOUT or $tee will segfault.
> 
> This is also broken on woody, so it's not new in 5.8.0.
> 
> I'm not familiar with how the perl interpreter works, but I'm fairly
> sure that a stack trace which begins:
> 
> #0  0x080b4a6a in Perl_vivify_ref ()
> #1  0x080b5332 in Perl_pp_method ()
> #2  0x080ae381 in Perl_runops_standard ()
> #3  0x08066018 in Perl_nothreadhook ()
> #4  0x08062292 in Perl_call_sv ()
> #5  0x08065a08 in Perl_call_method ()
> #6  0x080af021 in Perl_pp_print ()
> #7  0x080ae381 in Perl_runops_standard ()
> #8  0x08066018 in Perl_nothreadhook ()
> #9  0x08062292 in Perl_call_sv ()
> #10 0x08065a08 in Perl_call_method ()
> 
> And then continues to repeat the same sequence from #6-#10 for
> approximately 19000 (yes, thousand) stack frames, is a bad thing. I'm
> guessing I managed to throw a self-glob-reference in there somehow.

Back in 2009 this report got forwarded upstream, and this week got some
replies; you can see the discussions at

https://rt.perl.org/rt3//Ticket/Display.html?id=70001

Opinion is divided over exactly whether it's a valid bug or not, but
it seems unlikely that it will be fixed any time soon.

It's probably a bug in IO::Tee (if it indeed still exists in IO::Tee)
and also a lesser bug in perl.

Cheers,
Dominic.

-- 
Dominic Hargreaves | http://www.larted.org.uk/~dom/
PGP key 5178E2A5 from the.earth.li (keyserver,web,email)