Stdlib: sscanf() %e/%g float conversions — ValueError unsupported (ext/standard/formatted_io.c)

[PurHur]

Summary

sscanf() must support standard float scan conversions %e, %E, %f, %F, and %g per php-src formatted_io.c.

This compiler rejects %e / %g with:

ValueError: Bad scan conversion character "e"

Repro

php test/repro/maintainer_gap_sscanf_float_conversions.php               # Zend: exit 0
./script/docker-exec.sh -- php bin/vm.php test/repro/maintainer_gap_sscanf_float_conversions.php

Zend: sscanf('1.5e2', '%e', $f) → $f === 150.0; sscanf('3.14', '%g', $g) → $g ≈ 3.14.

This compiler: fatal ValueError on first %e pattern.

php-src reference

• ext/standard/formatted_io.c — php_sscanf_internal() float conversions
• ext/standard/sscanf.c

Implementation (PHP-in-PHP)

• ext/standard/VmSscanf.php — add %e/%g/%f parsing (scientific + fixed); share float parsing with %F if present.
• JIT: route through VmSscanf per php-in-php: JIT sscanf/vfscanf — route through VmSscanf PHP not __compiler_sscanf libc LLVM (#1492) #9134 (no new libc sscanf table in runtime/).

Done when (php-src-strict)

• php bin/vm.php test/repro/maintainer_gap_sscanf_float_conversions.php exits 0.
• %d overflow/saturation behavior remains covered by Stdlib: sscanf() %d overflow must saturate to PHP_INT_MAX not TypeError (ext/standard/formatted_io.c) #9536.

Related: #4512, #9325, #9560, #9536, #9134.

Refs #1492