develooper Front page | perl.wxperl.users | Postings from December 2010

Re: Crashy bug in either Wx or Perl 5.12

Thread Previous | Thread Next
Eric J. Roode
December 2, 2010 06:43
Re: Crashy bug in either Wx or Perl 5.12
Message ID:
    Very interesting, Mark.

    Long-term, I wonder if perhaps a change can be made to perl itself, to
manage conflicting uses of PERL_MAGIC_ext.  So when an extension calls the
macro or function to store its data in the variable's magic, perl tags it
with something that identifies the extension, and files it as part of the
data.  Then on fetching the data, perl looks up and returns only that
extension's data.

    Medium-term, perhaps perl could be modified to croak gracefully when
someone stores data in PERL_MAGIC_ext when something's already there.  Ah,
but then how could perl tell whether the extension is updating its own data
or overwriting someone else's data?  You're left with tagging again.

    Short-term, I think you're right -- wxperl should put a marker on its
PERL_MAGIC_ext data, and croak gracefully when it reads data that it didn't
put there.
Eric J. Roode

On Wed, Dec 1, 2010 at 11:15 PM, Mark Dootson <> wrote:

> Hi,
> The code actually dies at '$self->InsertColumn(0, 'Zero');'
> wxPerl uses PERL_MAGIC_ext to store a pointer to the underlying wxWidgets C
> object data. This bit of 'magic' is supposed to be for the use of
> extensions. (like wxPerl).
> Hash::Util::FieldHash also uses the PERL_MAGIC_ext belonging to the wxPerl
> extensions to store some data.
> When '$self->InsertColumn(0, 'Zero');', the 'magic' that wxPerl retrieves
> no longer points to the wxObject - but to the Hash::Util::FieldHash data -
> so bad things happen.
> This particular issue could probably be fixed. It is possible to have
> multiple bits of magic of the same type - and wxPerl could add an identifier
> for its own magic and only retrieve the PERL_MAGIC_ext that has that
> identifier.
> However, reading the Hash::Util::FieldHash pod, that talks of calling the
> DESTROY method of referenced objects, amongst other things. I would expect
> there to be more work needed to accommodate Hash::Util::FieldHash than just
> fixing the 'magic' issue.
> Hope this is useful info.

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About