[VOTE] fsync function

On Thu, Feb 25, 2021 at 1:01 PM Christoph M. Becker cmbecker69@gmx.de
wrote:

Hey David,
http://ocramius.github.com/

On Wed, Feb 24, 2021 at 6:15 PM David Gebler davidgebler@gmail.com
wrote:

Voting is now open for 2 weeks on
https://wiki.php.net/rfc/fsync_function

Regards,
David Gebler

On Tue, Feb 23, 2021 at 5:15 PM David Gebler davidgebler@gmail.com
wrote:

Hi internals,
As there appear to be no objections or concerns, I intend to open
voting
on https://wiki.php.net/rfc/fsync_function tomorrow and voting will
remain open for two weeks.

The RFC and its implementation

• Adds functions fsync() and fdatasync() for plain file streams on Unix
  systems.
• Adds function fsync() on Windows with fdatasync() available as an
  alias

PR ref https://github.com/php/php-src/pull/6650

Just clarifying on my "NO" vote: I'm opposed to having more E_WARNING
raised by the language.

We have better approaches for that:

• exceptions
• union types
• Maybe/Either types

I don't see a reason to introduce a warning here, and it makes the
function
unusable unless some poor soul implements a library that gets rid of the
warning.

Note that this warning is never supposed to occur in production,
because it is a programming error to pass a wrong resource type.
Rasing a warning for wrong resource types is standard behavior of
zend_fetch_resource() (unless no resource type name is passed). I don't
see why these new functions should use a non standard mechanism.

Of course, in the long run resources should go away, which would resolve
this issue for these functions as well.

A wrong resource type will throw a TypeError as usual -- the warning is
thrown if the resource is of the correct type (stream), but does not
support sync operations. For example, if you tried to do an fsync on an
http:// stream.

Ah, thanks for the correction! (I'm doing too much PHP 7, still.)

Throwing a warning in such a case is consistent with how functions for
other optional stream operations work.

ACK.

--
Christoph

Het Nikita,

Throwing a warning in such a case is consistent with how functions for
other optional stream operations work.

I understand that, and I'm OK with being against it.

Marco Pivetta

http://twitter.com/Ocramius

http://ocramius.github.com/