develooper Front page | perl.perl5.porters | Postings from May 2016

Re: revert MG consting (Coro breakage) for 5.24?

From:
Craig A. Berry
Date:
May 4, 2016 16:39
Subject:
Re: revert MG consting (Coro breakage) for 5.24?
Message ID:
CA+vYcVxku0c3jz=Qvz-RwVgmqnBjaM2UjEp1uKtgXGjfj0f9PQ@mail.gmail.com
On Wed, May 4, 2016 at 9:12 AM, Ricardo Signes
<perl.p5p@rjbs.manxome.org> wrote:
> * Nicholas Clark <nick@ccl4.org> [2016-05-04T09:47:40]
>> On Wed, May 04, 2016 at 09:37:17AM -0400, David Golden wrote:
>>
>> > I strongly urge that any core changes be deferred until 5.25 and backported
>> > later if successful and possible (which I think is what Nicholas is
>> > suggesting).
>>
>> I certainly think that that's what I think that I'm suggesting.
>> (In that coffee is no substitute for sleep, even at the strength that the
>> office coffee machine is calibrated to)
>
> Isn't the further complication that v5.24.1 can't unconst something consted in
> v5.24.0?

That's true, but as far as I can tell, Leon's patch as amended by
Nicholas makes the consting orthogonal to fixing Coro.  It changes
logic inside Perl_magic_getsig and Perl_magic_setsig without changing
any APIs or global data that would break binary compatibility. It
isn't the sort of thing that we *usually* backport, but I can't see
any absolute reason that we couldn't. It would of course require a new
version of Coro that doesn't (since it no longer needs to) mess with
the vtable.

I don't have a strong preference about reverting the consting patch
but if we plan to reapply it, I would find a it little weird to be
flip-flopping back and forth with 5.20 not const, 5.22 const, 5.24 not
const, 5.26 const.



nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About