Bug #8761 for Math-Pari: Does not work (crashes) under perl 5.8.6

Tue Dec 07 00:15:30 2004 Guest - Ticket created

Subject:

Does not work (crashes) under perl 5.8.6

Math::Pari fails its selftests under perl 5.8.6 on Linux (Debian Sarge, Fedora Core 3) as well as solaris 8 The selftests fail because perl crashes when (some) Math::Pari objects are gc'd. This does not happen when Math::Pari is used by an identically configured/built perl 5.8.5 instead. perl -Iblib/lib -Iblib/arch -MMath::Pari -e '$e=new Math::Pari "matrix(1,1)"' will produce the same failure. the stack moveoff stuff seems to be implicated here....

Download gdbsession
application/octet-stream 25k

Message body not shown because it is not plain text.

Tue Dec 14 05:16:19 2004 nospam-abuse [...] ilyaz.org - Correspondence added

Date:	Fri, 10 Dec 2004 21:35:40 -0800
From:	Ilya Zakharevich <nospam-abuse [...] ilyaz.org>
To:	Guest via RT <bug-Math-Pari [...] rt.cpan.org>
CC:	undisclosed-recipients: ;
Subject:	Re: [cpan #8761] Does not work (crashes) under perl 5.8.6
RT-Send-Cc:

On Tue, Dec 07, 2004 at 12:15:33AM -0500, Guest via RT wrote: Show quoted text

> > This message about Math-Pari was sent to you by guest <> via rt.cpan.org > > Full context and any attached attachments can be found at: > <URL: https://rt.cpan.org/Ticket/Display.html?id=8761 > > > Math::Pari fails its selftests under perl 5.8.6 on Linux (Debian Sarge, Fedora Core 3) as well as solaris 8 > > The selftests fail because perl crashes when (some) Math::Pari objects are gc'd. This does not happen when Math::Pari is used by an identically configured/built perl 5.8.5 instead. > > perl -Iblib/lib -Iblib/arch -MMath::Pari -e '$e=new Math::Pari "matrix(1,1)"' > will produce the same failure. > > the stack moveoff stuff seems to be implicated here....

Thanks for a detailed report. I built 5.8.6 and can reproduce your crash with your example; simplified to perl -Mblib -MMath::Pari -wle "$e=PARI 'matrix(1,1)'" No estimate on when I can look at it in more details, Ilya

Tue Dec 14 05:16:20 2004 nospam-abuse [...] ilyaz.org - Correspondence added

Date:	Sat, 11 Dec 2004 13:42:57 -0800
From:	Ilya Zakharevich <nospam-abuse [...] ilyaz.org>
To:	Guest via RT <bug-Math-Pari [...] rt.cpan.org>
Subject:	Re: [cpan #8761] Does not work (crashes) under perl 5.8.6
RT-Send-Cc:

On Fri, Dec 10, 2004 at 09:35:40PM -0800, Ilya Zakharevich wrote: Show quoted text

> On Tue, Dec 07, 2004 at 12:15:33AM -0500, Guest via RT wrote:

> > > > This message about Math-Pari was sent to you by guest <> via rt.cpan.org > > > > Full context and any attached attachments can be found at: > > <URL: https://rt.cpan.org/Ticket/Display.html?id=8761 > > > > > Math::Pari fails its selftests under perl 5.8.6 on Linux (Debian Sarge, Fedora Core 3) as well as solaris 8 > > > > The selftests fail because perl crashes when (some) Math::Pari objects are gc'd. This does not happen when Math::Pari is used by an identically configured/built perl 5.8.5 instead. > > > > perl -Iblib/lib -Iblib/arch -MMath::Pari -e '$e=new Math::Pari "matrix(1,1)"' > > will produce the same failure. > > > > the stack moveoff stuff seems to be implicated here....

I uploaded a new version to PAUSE. Should propagate to CPAN in a few hours. Actually, the analysis about moveoff() turned out to be misleading; so it is NOT a failure of Perl 5.8.6. This new version just uses a new algorithm to determine that an array is empty (for the purpose of DESTROY'ing the elements). Those who construct arrays by hand should take notice that AvALLOC == 0 is not enough for the array to be considered empty. Yours, Ilya