Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > perl.debugger > #3 > unrolled thread

Segmentation fault under -d flag

Started bykes-kes@yandex.ua (KES)
First post2015-12-06 21:15 +0200
Last post2015-12-09 13:22 +0200
Articles 4 — 2 participants

Back to article view | Back to perl.debugger


Contents

  Segmentation fault under -d flag kes-kes@yandex.ua (KES) - 2015-12-06 21:15 +0200
    Re: Segmentation fault under -d flag shlomif@shlomifish.org (Shlomi Fish) - 2015-12-07 15:59 +0200
      Re: Segmentation fault under -d flag kes-kes@yandex.ua (KES) - 2015-12-09 13:07 +0200
        Re: Segmentation fault under -d flag kes-kes@yandex.ua (KES) - 2015-12-09 13:22 +0200

#3 — Segmentation fault under -d flag

Fromkes-kes@yandex.ua (KES)
Date2015-12-06 21:15 +0200
SubjectSegmentation fault under -d flag
Message-ID<1525651449429326@web1g.yandex.ru>
Hi.

This code woks fine:
package Devel::KP;
 
 our $VERSION =  '0.01';
 
 
 package    # hide the package from the PAUSE indexer
   DB;

   sub sub {
   	goto &$DB::sub;
   }
   
   use Benchmark qw/ cmpthese /;
   sub DB {
   }
   
   1;
   
   
   But if I change DB::sub to look like:
   sub sub {
           print @_;
   	goto &$DB::sub;
   }
   
   
I get segfault

[toc] | [next] | [standalone]


#5

Fromshlomif@shlomifish.org (Shlomi Fish)
Date2015-12-07 15:59 +0200
Message-ID<20151207155948.2fd15583@telaviv1.shlomifish.org>
In reply to#3
Hi KES,

I was able to reduce the code significantly and it seems to be caused by
Time::HiRes:

https://github.com/shlomif/perl5-Time-HiRes-perl-d-segfault

I was also able to reproduce it on the latest bleadperl. I'm going to report
this bug.

Thanks!

Regards,

	Shlomi Fish

On Sun, 06 Dec 2015 21:15:26 +0200
KES <kes-kes@yandex.ua> wrote:

> Hi.
> 
> This code woks fine:
> package Devel::KP;
>  
>  our $VERSION =  '0.01';
>  
>  
>  package    # hide the package from the PAUSE indexer
>    DB;
> 
>    sub sub {
>    	goto &$DB::sub;
>    }
>    
>    use Benchmark qw/ cmpthese /;
>    sub DB {
>    }
>    
>    1;
>    
>    
>    But if I change DB::sub to look like:
>    sub sub {
>            print @_;
>    	goto &$DB::sub;
>    }
>    
>    
> I get segfault



-- 
-----------------------------------------------------------------
Shlomi Fish       http://www.shlomifish.org/
Perl Humour - http://perl-begin.org/humour/

To have bugs is human; to fix them — divine.

Please reply to list if it's a mailing list post - http://shlom.in/reply .

[toc] | [prev] | [next] | [standalone]


#6

Fromkes-kes@yandex.ua (KES)
Date2015-12-09 13:07 +0200
Message-ID<3256481449659236@web28o.yandex.ru>
In reply to#5
Here the another 'segmentation fault'

$cat t.pl
my $x;

$ cat Devel/PkgATCompileTime;
package Devel::PkgAtCompileTime;

use Devel::ImportArgs;

1;

$cat Devel/ImportArgs;
package Devel::ImportArgs;


sub import {
}



package    # hide the package from the PAUSE indexer
    DB;

use Package::Stash;
my $this =  Package::Stash->new( 'DB' );

sub sub {
        # I feel this is a crazy thing and I should not do it here. But occasionally I have did it.
         my $all =  $this->get_all_symbols;
         &$DB::sub;
}


sub DB {
}

1;


$ perl -d:PkgAtCompileTime t.pl 
Segmentation fault



07.12.2015, 16:05, "Shlomi Fish" <shlomif@shlomifish.org>:
> Hi KES,
>
> I was able to reduce the code significantly and it seems to be caused by
> Time::HiRes:
>
> https://github.com/shlomif/perl5-Time-HiRes-perl-d-segfault
>
> I was also able to reproduce it on the latest bleadperl. I'm going to report
> this bug.
>
> Thanks!
>
> Regards,
>
>         Shlomi Fish
>

[toc] | [prev] | [next] | [standalone]


#7

Fromkes-kes@yandex.ua (KES)
Date2015-12-09 13:22 +0200
Message-ID<3272991449660122@web4m.yandex.ru>
In reply to#6
That is because program run into recursion:
Package::Stash::XS::namespace
Package::Stash::XS::namespace
Package::Stash::XS::namespace
Package::Stash::XS::namespace
Segmentation fault

http://search.cpan.org/~abigail/perl-5.23.5/pod/perlvar.pod#$^P
Actually I should disable first bit:
local $^P = $^P & ~1
my $all = $this->get_all_symbols;

but the $^P does not work at runtime. Because the 'JMP' ops are created at compile time.
but the perl may check first $^P bit at run time. And if it is not set do not pass control to DB::sub/lsub/goto. And in compile time create 'JMP' not to DB::sub but to the 
DB::_sub {
     &DB::sub   if $^P & 1;
}
same code for goto and lsub. 

This seems simple to fix.

09.12.2015, 13:07, "KES" <kes-kes@yandex.ua>:
> Here the another 'segmentation fault'
>
> $cat t.pl
> my $x;
>
> $ cat Devel/PkgATCompileTime;
> package Devel::PkgAtCompileTime;
>
> use Devel::ImportArgs;
>
> 1;
>
> $cat Devel/ImportArgs;
> package Devel::ImportArgs;
>
> sub import {
> }
>
> package # hide the package from the PAUSE indexer
>     DB;
>
> use Package::Stash;
> my $this = Package::Stash->new( 'DB' );
>
> sub sub {
>         # I feel this is a crazy thing and I should not do it here. But occasionally I have did it.
>          my $all = $this->get_all_symbols;
>          &$DB::sub;
> }
>
> sub DB {
> }
>
> 1;
>
> $ perl -d:PkgAtCompileTime t.pl
> Segmentation fault
>
> 07.12.2015, 16:05, "Shlomi Fish" <shlomif@shlomifish.org>:
>>  Hi KES,
>>
>>  I was able to reduce the code significantly and it seems to be caused by
>>  Time::HiRes:
>>
>>  https://github.com/shlomif/perl5-Time-HiRes-perl-d-segfault
>>
>>  I was also able to reproduce it on the latest bleadperl. I'm going to report
>>  this bug.
>>
>>  Thanks!
>>
>>  Regards,
>>
>>          Shlomi Fish

[toc] | [prev] | [standalone]


Back to top | Article view | perl.debugger


csiph-web