On Tue, Dec 30, 2003 at 04:51:48PM -0800, Jan Dubois wrote: > On Wed, 10 Jul 2002 23:32:03 -0700, Jan Dubois <JanD@ActiveState.com> wrote: > > >On Thu, 11 Jul 2002 08:13:59 +0200, Arthur Bergman <arthur@contiller.se> > >wrote: > > > >>The myperl that holds the function pointer gets freed by perl_free and > >>then PerlMemShared_free is left on it its own, on unix > >>PerlMemShared_free is just a simple safefree so it doesn't need the > >>my_perl, nor does it happen on join since then the aTHX is the callee. > >> > >>Please try the following patch, > > > >Seems to solve the problem! I ran the test script 6 times without any > >access violations. Pass it on to the "Thanks, applied!" crowd! :) > > I've just spent 2 hours hunting down a threading problem and ended up in the > exact same corner again. I then checked both perl-latest and perl-5.8.x, and > neither seem to include your patch. Any reason it didn't get committed? It > looks like it may almost be too late to get it into Perl 5.8.3 now either. I guess that it was overlooked. I can't find any trace of this patch on p5p - is the Cc: to p5p new, and the rest of this thread was a private discussion? I'm curious as to what the original problem was. (and what the patch looks like before it got quoted multiple times) Nicholas ClarkThread Previous | Thread Next