PHP 8.6 is scheduled for general availability on November 19, 2026, following a release candidate cycle that ran to RC7. The feature list is long, but it splits cleanly into two piles: a handful of changes that will show up in everyday Laravel code, and a larger set of runtime plumbing that most application teams will never touch directly. Here is how we would sort it, and what we would do about the deprecations.
What's new in PHP 8.6
Laravel News has a thorough rundown, and the items fall into three groups.
Language features you will actually type:
- Partial function application. Calling a function with a
?placeholder returns a closure that waits for the missing argument, and...stands in for all the remaining ones. Something likearray_map(strtolower(?), $titles)replaces a throwaway arrow function. clamp(). A built-in that holds a value between a minimum and a maximum, and throws aValueErrorif the minimum is larger than the maximum.Time\Duration. A final readonly class for lengths of time with nanosecond precision, with factory methods per unit, arithmetic, and comparison.- Readonly property defaults. A readonly property can now declare a default value, which used to be a compile-time error.
- Smaller additions. Doc comments directly on function parameters (readable through
ReflectionParameter::getDocComment()), a globalSortDirectionenum, and__debugInfo()on enums.
Runtime plumbing: a unified polling API under Io\Poll (wrapping epoll, kqueue, event ports, and WSAPoll), a stream error model that can throw exceptions instead of emitting warnings, and builder classes for the URI extension that arrived in PHP 8.5.
Configuration: three php.ini session defaults change for new installations. session.use_strict_mode and session.cookie_httponly both move from 0 to 1, and session.cookie_samesite goes from unset to Lax.
On the deprecation side, the PHP 8.6 deprecations RFC was voted on item by item, with removal targeted for PHP 9 at the earliest. Accepted items include returning from a finally block, the alias functions is_double(), is_integer(), is_long(), and doubleval(), strcoll(), spl_object_hash(), mysqli_get_charset(), and passing session_set_save_handler() a handler object that lacks create_sid() and validateId() methods. A proposal to deprecate the list() construct tied 23 to 23 and did not pass.
Our take
This is a good release that asks for very little urgency. That is a compliment.
Partial application is the headline, and it needs a team convention. It makes collection pipelines and callbacks tidier, and we expect it to become idiomatic quickly. But a ? buried in the middle of a call is easy to miss in code review, and one detail matters: the arguments you do supply are evaluated when the partial is created, not when it runs. That is a subtle difference from an arrow function, and a likely source of confusing bugs. We would agree as a team where partials are welcome (short, single-placeholder callbacks) and where a named closure is still clearer.
Duration is the sleeper. Most codebases have a timeout or TTL somewhere that is an integer with an ambiguous unit. Seconds? Milliseconds? A real type for "a length of time" fixes a category of bug rather than a single one. The honest trade-off is that the payoff depends on frameworks and libraries accepting it in their own APIs, and that will take time. Until then it is most useful inside your own domain code.
The session defaults are what clients will ask about, and mostly a non-event. The change applies to php.ini defaults on new installations, and as Laravel News points out, Laravel manages its own session cookies, so most Laravel apps are unaffected. The apps worth a second look are the older ones that call PHP's native session functions directly and quietly lean on the old defaults. A well-run site should already have these settings, so if tightening them breaks something, that is useful information about the app.
The plumbing matters, just not to your sprint. The polling RFC itself says its main motivation is internal, with userland async frameworks as the secondary audience. Application teams will feel that work indirectly, through the runtime and the libraries built on it.
Do not upgrade production on a .0. After October's emergency PHP patches, our appetite for being early on a runtime version is low. The sensible order is: get 8.6 into CI, let a patch release or two land, confirm your framework and dependencies support it, then move. Nothing in this release is worth skipping that sequence for.
Deprecations are cheap now and expensive later. None of the accepted items break anything in 8.6. They emit notices. But every notice you clear this winter is one you will not be untangling under deadline when PHP 9 removes the feature.
Practical recommendations
- Add PHP 8.6 to your CI matrix now, with deprecations reported, and treat it as allowed-to-fail until your dependencies catch up. A test suite you trust is what makes this cheap, and Pest 5 is a good place to start.
- Grep for the easy deprecations. The alias functions (
is_double(),is_integer(),is_long(),doubleval()) are mechanical replacements. Areturninside afinallyblock deserves a human read, because it can silently discard an exception. - Audit custom session handlers. If any project passes its own handler object to
session_set_save_handler(), check that it definescreate_sid()andvalidateId(). - Review session settings on legacy apps before provisioning new servers, so the new defaults confirm your configuration instead of surprising it.
- Write down a team rule for partial application before it shows up in a pull request.
- Hold production upgrades until a patch release is out and your framework version officially supports 8.6.
PHP 8.6 rewards teams that upgrade steadily and punishes no one for waiting a few weeks. Use the quiet to clear deprecations.
Originally referenced: What's New in PHP 8.6 on Laravel News.
If you want help planning a PHP 8.6 upgrade path for a Laravel, WordPress, or Drupal site, or an audit of what the new deprecations touch in your codebase, get in touch.



