Stdlib: html_entity_decode() / htmlspecialchars_decode() — ENT_HTML5 named entities

[PurHur]

Category

stdlib

Problem

html_entity_decode() and htmlspecialchars_decode() ignore ENT_HTML5 (constant 48): HTML5 named entities such as ',  , ½, and © are left unchanged. Zend decodes them when ENT_HTML5 is set in the quote-style flags.

php-src reference

• php/php-src ext/standard/html.c — php_html_entities() / ENT_HTML5 entity tables (ent_html5 / ent_html5_apos)
• php/php-src ext/standard/html_tables.h — HTML5 named entity map

Repro (failure today)

./script/docker-exec.sh -- bash -lc 'source script/php-env.sh
cat > /tmp/ent_html5.php <<'"'"'PHP'"'"'
<?php
$flags = ENT_QUOTES | ENT_HTML5;
echo html_entity_decode("&apos;", $flags), "\n";
echo html_entity_decode("&nbsp;", ENT_HTML5), "\n";
echo html_entity_decode("&frac12;", ENT_HTML5), "\n";
PHP
php /tmp/ent_html5.php
php bin/vm.php /tmp/ent_html5.php
'

Input	Zend (ENT_HTML5)	This compiler
' + ENT_QUOTES|ENT_HTML5	'	'
 	U+00A0 (non-breaking space)	 
½	½	½

Scope (this repo)

• ext/standard/VmString.php — extend htmlspecialchars_decode() / entity table lookup for HTML5 names
• ext/standard/html_entity_decode.php, htmlspecialchars_decode.php — pass ENT_HTML5 flag through
• JIT/AOT paths registered alongside VM handlers
• test/compliance/cases/stdlib/html_entity_decode_ent_html5.phpt

Related: #2472 (ENT_COMPAT subset); this is the HTML5 named-entity gap.

Done when

• VM + JIT + AOT: repro script output matches Zend byte-for-byte (or codepoint-normalized for  )
• ENT_HTML5 without other flags follows php-src rules for which entities decode
• Compliance .phpt under test/compliance/cases/stdlib/ compares Zend vs VM on the repro script

[PurHur]

claim: worker-lane-c — starting this run

[PurHur]

Merged via PR #8038.

Verification

./script/docker-exec.sh -- bash -lc 'source script/php-env.sh
php bin/vm.php test/repro/maintainer-ent-html5.php
vendor/bin/phpunit test/compliance/HtmlentitiesVMTest.php
vendor/bin/phpunit test/compliance/HtmlentitiesJITTest.php
vendor/bin/phpunit --filter html_entity_decode_ent_html5 test/aot/AotTest.php
'

All green: VM/JIT/AOT output matches Zend for ' (ENT_QUOTES|ENT_HTML5),  , ½, © with ENT_HTML5.

PHP implementation: ext/standard/VmString.php, ext/standard/Html5NamedEntities.php (2125 names from php-src html_tables.h). No C runtime growth.