Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:113477
Content-Type: text/plain;
	charset=utf-8
Content-Transfer-Encoding: quoted-printable
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.120.23.2.4\))
Date: Thu, 11 Mar 2021 10:05:00 -0600
References: <D376DEFE-26A4-4F11-8548-D9454087E9FD@trowski.com>
 <8812EE70-4720-4639-926E-1EA206624BC1@trowski.com>
To: php internals <internals@lists.php.net>
In-Reply-To: <8812EE70-4720-4639-926E-1EA206624BC1@trowski.com>
Message-ID: <18FC107D-C446-4F5F-AD12-5B08905E7463@trowski.com>
Subject: Re: [PHP-DEV] [VOTE] Fibers
From: aaron@trowski.com (Aaron Piotrowski)


> Hi all!
>=20
> A concern was raised off list that due to the complexity of the way =
this feature interacts with the engine, it may be best to mark the =
feature as experimental. This would allow some changes to be made to =
certain edge-case behaviors and, while I don't think it would be =
necessary, the public API.
>=20
> We (Niklas and I) agree with this and propose that if this feature is =
accepted, it is marked as experimental through the 8.x release cycle or =
until we're comfortable removing that label, whichever comes first.
>=20
> Experimental in this context would mean fibers would be compiled and =
available in all releases, but the documentation would denote that =
behavioral and API changes may be made in future minor releases =E2=80=93 =
use at your own risk.
>=20
> As this feature is targeted at library authors and not average PHP =
users, we believe there would be little effect to the community to take =
this step.
>=20
> Cheers,
> Aaron Piotrowski
> --
> PHP Internals - PHP Runtime Development Mailing List
> To unsubscribe, visit: https://www.php.net/unsub.php
>=20

Hi everyone!

I=E2=80=99m concerned my email above may have had the opposite effect =
from which I had hoped.

I think the Fiber implementation is sound and the API, mirroring that of =
Ruby, is proven and unlikely to require change.

There are certain edge case behaviors that I=E2=80=99m concerned may =
need to changed, such as the error reporting level that is active when a =
fiber is switched. If the @ operator is active, should it remain active =
after a fiber switch? Right now, it does not, the @ operator is on a =
per-fiber basis, restoring error reporting when switching fibers. =
Perhaps though this was the wrong decision. What I was hoping to =
accomplish with marking the feature as experimental was to get feedback =
on such behaviors and make sure we make the right choices.

Recent changes to the JIT also make fibers as an extension incompatible =
due to defining a user opcode handler. This is unfortunate, as cli apps =
including amphp and ReactPHP greatly benefited from the JIT: =
https://www.php.net/releases/8.0/en.php (see Relative JIT contribution =
to PHP 8 performance). Integrating fibers in core will allow the opcode =
handler to be predefined.

Hopefully that provides some clarity to our intensions. Cheers!

Aaron Piotrowski