Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:115915
MIME-Version: 1.0
References: <CADJ56BRPi_GHE7R70k-NdgQ6ea2ogLSe6Z6Eg-YGKpeYSmvhbQ@mail.gmail.com>
 <CAJvSBVP2pCT6wDNrpVDp=1_JQtrG=H2Y2fy8L-vcvgawhWstCw@mail.gmail.com>
In-Reply-To: <CAJvSBVP2pCT6wDNrpVDp=1_JQtrG=H2Y2fy8L-vcvgawhWstCw@mail.gmail.com>
Date: Thu, 2 Sep 2021 12:15:33 +0100
Message-ID: <CADJ56BSHDTZ86oRqMHfsB860WXsO3ORv7JVGArJx+Q7FN4nD5g@mail.gmail.com>
Cc: PHP internals <internals@lists.php.net>
Content-Type: text/plain; charset="UTF-8"
Subject: Re: [PHP-DEV] Adding a way to disable the stat cache
From: kevin@lyda.ie (Kevin Lyda)

Just as an aside, Lynn and Peter are two more people who have had this
issue - in addition to the folks mentioned in the various bugs about this.
Really would like clear guidance on what's needed to move the PR on
once I resolve the current merge conflict.

On Thu, Sep 2, 2021 at 9:49 AM Peter Bowyer <phpmailinglists@gmail.com> wrote:
> On Thu, 2 Sept 2021 at 09:27, Kevin Lyda <kevin@ie.suberic.net> wrote:
> > The change I've made will allow people to disable the
> > cache so that it won't cause errors and it leaves the current broken
> > behaviour in place.
>
> Is there any good reason not to remove it completely?

Removing it completely would be ideal, however a number of people objected
in the linked bug. And while it's not needed in modern Unix operating
systems, it's not clear if Windows might benefit from this.

Personally I'd argue that due to its buggy behaviour it should be
removed. There's no way to make this work correctly in userspace so it's
not a good place for it. OS and filesystem developers implemented caching
on those levels because they have the state to properly maintain a cache.

However if you use the filesystem a certain way and you pretend other
processes can't modify a file out from under you, the current caching
system could provide a performance benefit on OSs and filesystems
that don't offer this sort of caching. And maybe someone depends on
this. That's the best argument I can make for retaining it.

> I've been bitten by the stat cache before and would be pleased to see it
> gone for good.

As have I. And I had to write some goofy sed scripts to insert
clearstatcache() in order to stop rare bugs from popping up.  Which means
I had to experience it at least twice.

Kevin