为什么带有PCRE_UTF8的PHP preg_match_all在CLI和Apache/mod_php上会得到不同的结果？

Question

以下代码在通过CLI和Apache/mod_php运行时会产生不同的结果：

<pre>
<?php
error_reporting(E_ALL);
ini_set('display_errors', '1');

echo setlocale(LC_ALL, 0)."\n";
// echo setlocale(LC_ALL, "en_GB.UTF-8")."\n";

$terms = array
(
    //Always matches:
    "Label Generation",
    //Doesn't match when using u (PCRE_UTF8) modifier:
    "Receipt of Prescription and Validation of Patient Information",

);

$text       = "Some terms to match: ".implode(", ",$terms);
$pattern    = "/(".implode(")|(", $terms).")/is";
$regexps    = array
(
   "Unicode"     => $pattern."u", //Add u (PCRE_UTF8) modifier
   "Non-unicode" => $pattern
);

echo "Text:\n'$text'\n";

foreach($regexps as $type=>$regexp)
{
    $matches    = array();
    $total      = preg_match_all($regexp,$text,$matches);

    echo "\n\n";
    echo "$type regex:\n'$regexp'\n\n";
    echo "Total $type matches: ";
    var_dump($total);
    echo "\n$type matches: ";
    var_dump($matches[0]);
}
?>
</pre>

CLI输出(正确)：

<pre>
/en_GB.UTF-8/C/C/C/C/C
Text:
'Some terms to match: Label Generation, Receipt of Prescription and Validation of Patient Information'


Unicode regex:
'/(Label Generation)|(Receipt of Prescription and Validation of Patient Information)/isu'

Total Unicode matches: int(2)

Unicode matches: array(2) {
  [0]=>
  string(16) "Label Generation"
  [1]=>
  string(61) "Receipt of Prescription and Validation of Patient Information"
}


Non-unicode regex:
'/(Label Generation)|(Receipt of Prescription and Validation of Patient Information)/is'

Total Non-unicode matches: int(2)

Non-unicode matches: array(2) {
  [0]=>
  string(16) "Label Generation"
  [1]=>
  string(61) "Receipt of Prescription and Validation of Patient Information"
}
</pre>

Apache/mod_php webserver结果(不正确-不使用/u修饰符时仅匹配字符串)：

/en_GB.ISO8859-1/C/C/C/C/C
Text:
'Some terms to match: Label Generation, Receipt of Prescription and Validation of Patient Information'


Unicode regex:
'/(Label Generation)|(Receipt of Prescription and Validation of Patient Information)/isu'

Total Unicode matches: int(1)

Unicode matches: array(1) {
  [0]=>
  string(16) "Label Generation"
}


Non-unicode regex:
'/(Label Generation)|(Receipt of Prescription and Validation of Patient Information)/is'

Total Non-unicode matches: int(2)

Non-unicode matches: array(2) {
  [0]=>
  string(16) "Label Generation"
  [1]=>
  string(61) "Receipt of Prescription and Validation of Patient Information"
}

使用/u (PCRE_UTF8)选项时，web服务器无法同时匹配两个字符串。我已经尝试过setlocale(LC_ALL, "en_GB.UTF-8");将web服务器的区域设置与CLI区域设置进行匹配，它成功地做到了这一点，但它对输出没有影响。我怀疑PCRE库有问题，但我不明白它在CLI和web服务器之间有什么不同- PHP在两个环境中报告相同的库版本:PHP5.4.14PCRE (Perl兼容正则表达式)支持启用=>的PCRE库版本=> 8.32 2012-11-30

pcretest报告不支持UTF-8，但尽管如此，CLI版本仍会产生正确的结果

$> pcretest -C
PCRE version 8.32 2012-11-30
Compiled with
  8-bit support
  No UTF-8 support
  No Unicode properties support
  No just-in-time compiler support
  Newline sequence is LF
  \R matches all Unicode newlines
  Internal link size = 2
  POSIX malloc threshold = 10
  Default match limit = 10000000
  Default recursion depth limit = 10000000
  Match recursion uses stack

Answer 1

这个PHP设置对我很有帮助：

pcre.jit=0 

Answer 2

阿拉斯泰尔，揭示了这个古老的问题，因为它涉及到一个永恒的问题的程序员感兴趣的各个时代。

正如Dino所说，在同一机器上有多个版本的PCRE是很常见的。我总是惊讶于一个普通的cPanel版本上安装了多少个版本的PCRE。这可能不是您的情况，但您似乎也有多个版本。

要查看安装了哪些PCRE，请在unix shell中键入：

find / -name libpcre.*

如果你想得到一些有意义的信息，你会想要使用你一直在做的pcretest，所以你可以find / -name pcretest然后somepath/pcretest -C

如果您使用的是cPanel、according to cPanel staff，则EasyApache安装的PCRE版本为opt/文件夹中的版本。您可以通过运行以下命令获取版本

/opt/pcre/bin/pcretest -C

它是一团糟，但这使我们保持警惕。:)

Answer 3

一些Linux发行版(例如Ubuntu)将它们的PHP打包成单独的php.ini文件，用于CLI和Apache。如果这是您的情况，那么您可能需要查看/etc/php5并调查其中的差异。