Regression: mb_substr/mb_strpos/mb_str_split/… invalid encoding LogicException — Zend ValueError (re-#13377, ext/mbstring)

[PurHur]

Category

Regression · php-src-strict · ext/mbstring invalid ``

Problem

After #13377 fixed mb_strtoupper/mb_strtolower (and mb_strlen/mb_convert_case already match), several mbstring APIs still throw a compiler-private LogicException for an unknown encoding instead of Zend’s ValueError.

Call (Zend 8.2.32 / HEAD bf5802236, 2026-08-06)	Zend	VM / PROFILE=8.4
mb_substr('a',0,1,'nope')	ValueError: … must be a valid encoding, \"nope\" given	LogicException: mb_substr requires mbstring for encoding nope…
mb_strpos('a','a',0,'nope')	ValueError (arg #4)	LogicException: mbstring search requires…
mb_str_split('a',1,'nope')	ValueError (arg #3)	LogicException
mb_strcut / mb_strimwidth / mb_strwidth	ValueError	LogicException
mb_stripos / mb_strrpos / mb_strripos	ValueError	LogicException
mb_strstr / mb_stristr / mb_strrchr / mb_strrichr	ValueError	LogicException
mb_str_pad(..., 'nope') (PROFILE≥8.4)	ValueError	LogicException

Already green (do not regress): mb_strlen, mb_strtoupper, mb_strtolower, mb_convert_case.

php-src reference

• php/php-src ext/mbstring/mbstring.c — encoding resolve → zend_argument_value_error / ValueError
• Prior fix shape: Stdlib: mb_strtoupper()/mb_strtolower() invalid encoding — ValueError not LogicException (ext/mbstring/mbstring.c) #13377 (mb_strtoupper/mb_strtolower)

PHP implementation target

• Shared encoding-validation helper used by the listed builtins in ext/mbstring/ (and JIT lowering if separate) — throw Zend-identical ValueError, not LogicException
• PHP-in-PHP only — no new runtime/*.c

Repro

./script/docker-exec.sh -- bash -lc 'cat > /tmp/mb_enc_valueerror.php <<'"'"'PHP'"'"'
<?php
foreach ([
  fn() => mb_substr(\"a\", 0, 1, \"nope\"),
  fn() => mb_strpos(\"a\", \"a\", 0, \"nope\"),
  fn() => mb_str_split(\"a\", 1, \"nope\"),
] as $i => $fn) {
  try { $fn(); echo \"$i ok\n\"; }
  catch (Throwable $e) { echo $i, \" \", get_class($e), \": \", $e->getMessage(), \"\n\"; }
}
PHP
php /tmp/mb_enc_valueerror.php
php bin/vm.php /tmp/mb_enc_valueerror.php'

Done when

• Listed APIs throw Zend-identical ValueError (message + arg index) on VM (and JIT when in scope)
• No LogicException / "in this compiler build" for invalid encodings
• Compliance or test/repro/ guard covering at least mb_substr + mb_strpos + mb_str_split
• php-src-strict; no php-compiler-strict shortcut

Trust snapshot (this maintainer run)

• bootstrap-inventory --check: OK 7075/7075
• make north-star5-verify-fast: OK
• make bootstrap-selfhost-helloworld: OK (Hello World)

[PurHur]

claim: PHP Compiler Ext — I/O & Crypto — starting this run (released #27995 — not a Zend gap; this is the real mbstring ValueError parity).

[PurHur]

Merged in #28019

Verification: PHP_COMPILER_PROFILE=8.4 php bin/vm.php test/repro/issue_27945_mb_encoding_valueerror.php — all listed APIs throw Zend-identical ValueError; utf8 alias works; SJIS remains LogicException (valid, unsupported). Messages matched php:8.4-cli.

Not covered: full VMTest provider (broken unrelated DOM PHPT on master); JIT/AOT.

[PurHur]

Verified on master @ b239e6dd3 (PR #28019): mb_substr(..., "not-an-encoding") throws Zend-identical ValueError. Closing.