Stdlib: str_getcsv()/fgetcsv() — escape=enclosure " parses quoted fields wrong (ext/standard/file.c)

[PurHur]

Category

stdlib · php-src-strict

Problem

When the CSV escape character equals the enclosure (", PHP 8.4+ default for str_getcsv() / fgetcsv()), this compiler stops parsing after the first field. Zend/php-src returns all fields with doubled-quote unescaping inside enclosures.

VM and JIT share ext/standard/VmCsv::parseLine() — one PHP fix covers all backends.

php-src reference

• php/php-src ext/standard/file.c — php_fgetcsv() / php_str_getcsv() state machine when escape char equals enclosure
• php/php-src ext/standard/file.c — fgetcsv() shares parser

Repro

Run inside ./script/docker-exec.sh:

cat > /tmp/repro_csv_escape_quote.php <<'PHP'
<?php
$line = 'a,"b""c",d';
var_dump(str_getcsv($line, ',', '"', '"'));
$f = fopen('php://memory', 'r+');
fwrite($f, $line . "\n");
rewind($f);
var_dump(fgetcsv($f, 0, ',', '"', '"'));
fclose($f);
PHP
php /tmp/repro_csv_escape_quote.php
php bin/vm.php /tmp/repro_csv_escape_quote.php

Engine	str_getcsv(..., '"', '"')
Zend PHP 8.2+	['a', 'b"c', 'd']
VM (today)	['a']

Verified 2026-06-17.

Scope (PHP-in-PHP)

Layer	Path
Parser SSOT	ext/standard/VmCsv.php — fix doubled-enclosure branch when $esc === $enc
VM	ext/standard/str_getcsv.php, ext/standard/fgetcsv.php (delegate to VmCsv)
JIT/AOT	lib/JIT/Builtin/StringStrGetcsvJit.php, StringFgetcsvJit.php (reuse VmCsv)
Tests	test/compliance/cases/stdlib/str_getcsv_escape_enclosure.phpt

No new C runtime — parser logic stays in PHP (VmCsv).

Done when (php-src-strict)

• Repro returns ['a', 'b"c', 'd'] on VM and JIT/AOT
• Backslash escape ('\\') path unchanged
• Compliance PHPT green in Docker
• Capability matrix unchanged (functions already registered)

Related

• Stdlib: str_getcsv() — scalar-to-string coercion for argument #1 (ext/standard/file.c parity) #4318 scalar coercion · Stdlib: str_getcsv() — enum case string operands must TypeError (ext/standard/string.c, php-src-strict) #8853 enum operands · Stdlib: fgetcsv() — enum case separator/enclosure/escape must TypeError (ext/standard/file.c, php-src-strict) #8975 fgetcsv enum guards · Epic: Self-host critical path — compiler compiles itself (M3→M5) #1492

[PurHur]

claim: PHP Compiler STD Lib — starting this run

[PurHur]

Merged via #15698.

Fix: VmCsv::parseLine() — when escape equals enclosure ("), skip the escape+next-byte branch; only doubled-enclosure unescaping applies (php-src ext/standard/file.c).

Verification:

php test/repro/maintainer_gap_str_getcsv_escape_enclosure.php   # ok
php bin/vm.php test/repro/maintainer_gap_str_getcsv_escape_enclosure.php   # ok
php bin/compile.php -l test/repro/maintainer_gap_str_getcsv_escape_enclosure.php
vendor/bin/phpunit --filter testEscapeEqualsEnclosureDoubledQuotes   # OK