Stdlib: nextafter() missing — IEEE next representable float (ext/standard/math.c) #9241

Category

stdlib · php-src-strict

Problem

nextafter(float $num, float $next) is part of ext/standard math in php-src (PHP 8.4+). This compiler has no builtin — function_exists('nextafter') is false and calls fatal as undefined.

php-src reference

• php/php-src ext/standard/math.c — PHP_FUNCTION(nextafter) wrapping php_nextafter()
• php/php-src ext/standard/basic_functions.stub.php — signature

Repro

<?php
declare(strict_types=1);
var_dump(function_exists('nextafter'));
var_export(nextafter(1.0, 2.0));
echo "\n";
var_export(nextafter(1.0, 0.0));
echo "\n";

./script/docker-exec.sh -- bash -lc 'php repro_nextafter.php
php bin/vm.php repro_nextafter.php'

Scope (PHP-in-PHP)

Implement with platform nextafter(3) only behind thin FFI if needed; prefer PHP VmMath wrapper calling libc via existing native helper patterns (VmMathNative).

Done when (php-src-strict)

• Repro prints true + same floats as Zend for toward +inf and toward 0.0
• Non-float operands raise same TypeError as Zend
• JIT/AOT agree with VM (or follow-up JIT issue filed with link)
• docs/capabilities.md regenerated via php script/capability-matrix.php
• ./script/ci-fast.sh --filter nextafter green

Related

• Stdlib: acosh()/asinh()/atanh() missing (ext/standard/math.c parity) #9220 acosh/asinh/atanh missing hyperbolic trio · Stdlib: fdiv() — var_export(true) prints INF.0 not INF (ext/standard/math.c) #5471 fdiv INF formatting