> -----Original Message-----
> From: Doug MacEachern [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, August 09, 2000 11:26 PM
> To: Geoffrey Young
> Cc: [EMAIL PROTECTED]
> Subject: Re: Apache::Symbol::undef_functions;
>
>
> On Wed, 9 Aug 2000, Geoffrey Young wrote:
>
> > Hi Doug...
> >
> > sorry to bother you with this... I'm looking into it, but
> I thought just in
> > case you had any ideas off the top of your head :)
>
> no worries, this is good stuff for [EMAIL PROTECTED], more
> heads with
> stuff on the top ;)
cool - I thought this was for 2.0 devel stuff only.
>
> > I'm trying to replicate StatINC in Dispatch.pm and ran
> across something
> > interesting.
> >
> > a call to
> > $class->Apache::Symbol::undef_functions;
> >
> > makes my OO behave oddly - instead of returning the rc of
> the method, it
> > returns the method name.
> >
> > I know - "What?"
> >
> >
> > $method = "foo;
> > $rc = $class->$method(); # $rc=1
> >
> > #now, undef_functions, delete and require, then
> > $rc = $class->$method(); # $rc="foo"
> >
> > #just delete and require, no undef_functions
> > $rc = $class->$method(); # $rc=1, but warns "Subroutine
> foo redefined"
> >
> > weird, eh?
>
> yeah, do you have a drop in test case?
well, not really :)
I think I managed to work around it - I was use()ing Apache::Symbol at the
top of the script - changing to require() like StatINC made everything work
ok...
>
> > I'm trying to trace through Apache::Symbol, but my C isn't
> very strong,
> > so...
>
> Apache::Symbol::undef is only different from the built-in
> undef, in the
> way it avoids 'prototype mismatch' warnings.
> if you have a
> -DDEBUGGING
> libperl or libperld.a + PERL_DEBUG=1 (SUPPORT has hints),
> local $^D = 8|16; might shed some light.
well, it took me two days and a handfull of box restarts to get 5.6 on my
test box - I should have compiled with debugging from the start :(
anyway, I did use Devel::Peek which showed some really strange results where
$rc went from IV to PV to PVNV with IOK to PVNV with POK on successive
calls. It was only returning OK that gave me problems and inconsistent
results - returning any other constant or 0 went fine... stranger than
fiction.
>
> > anyway, it's really kinda FYI, but just in case you were
> working feverishly
> > on a mod_perl-2.0 counterpart ;)
>
> i don't think Apache::Symbol will be in 2.0, because it was
> written just
> to work around things that are fixed in Perl 5.004+. but it
> might have to
> stick around for backwards compat.
does that mean that there's a way to avoid the 'subroutine redefined' stuff
without Apache::Symbol in 5.6?
anyway, no need to worry about this any longer - I'm sure it's an isolated
incident. thanks for your help...
--Geoff
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
