Bug #102110 for Compress-Raw-Zlib: Library is not 100% threadsafe

Fri Feb 13 14:11:35 2015 anall [...] andreanall.com - Ticket created

Subject:

Library is not 100% threadsafe

One-liner reproduction case: perl -e 'use Compress::Raw::Zlib; use threads; my $x = new Compress::Raw::Zlib::Deflate(); threads->new(sub {})->join()' Causes the error: *** Error in `perl': double free or corruption (!prev): 0x0000000000764df0 *** This seems to be because the Compress::Raw::Zlib::Deflate object gets cloned to the new thread with the same pointer. I am unsure how to go about fixing this. The easiest method would be adding "sub CLONE_SKIP { 1; }" to all packages. -- This will not actually *solve* the problem, but would at least prevent the crash at the cost of the cloned objects being undef. Thank you.

Sun Feb 15 05:08:01 2015 pmqs [...] cpan.org - Correspondence added

On Fri Feb 13 14:11:35 2015, anall wrote: Show quoted text

> One-liner reproduction case: > > perl -e 'use Compress::Raw::Zlib; use threads; my $x = new > Compress::Raw::Zlib::Deflate(); threads->new(sub {})->join()' > > Causes the error: *** Error in `perl': double free or corruption > (!prev): 0x0000000000764df0 *** > > > This seems to be because the Compress::Raw::Zlib::Deflate object gets > cloned to the new thread with the same pointer. > > I am unsure how to go about fixing this. The easiest method would be > adding "sub CLONE_SKIP { 1; }" to all packages. -- This will not > actually *solve* the problem, but would at least prevent the crash at > the cost of the cloned objects being undef. > > Thank you.

Thanks for the bug report. The trick with this one is to decide which thread "owns" the Compress::Raw::Zlib::Deflate object (and is the one that ends up eventually destroying it). Think some other modules have a way of explicitly setting the ownership. Will investigate.. cheers Paul

Sun Feb 15 05:08:02 2015 The RT System itself - Status changed from 'new' to 'open'