Consolidate environmental initialization

Summary:
We have a bunch of code duplication now between __init_script__.php and webroot/index.php. Consoldiate these methods and move them into PhabricatorEnv.

Merge PhabricatorRequestOverseer into PhabricatorStartup.

Test Plan: Loaded page, ran script. Wiped PHABRICATOR_ENV; loaded page, ran script; got errors.

Reviewers: btrahan, vrana

Reviewed By: btrahan

CC: aran

Maniphest Tasks: T2223

Differential Revision: https://secure.phabricator.com/D4283

scripts/__init_script__.php

@@ -1,5 +1,6 @@
 <?php
 
+function init_phabricator_script() {
   error_reporting(E_ALL | E_STRICT);
   ini_set('display_errors', 1);
 
@@ -14,59 +15,10 @@ if (!@constant('__LIBPHUTIL__')) {
     exit(1);
   }
 
+  phutil_load_library('arcanist/src');
   phutil_load_library(dirname(__FILE__).'/../src/');
 
-// NOTE: This is dangerous in general, but we know we're in a script context and
+  PhabricatorEnv::initializeScriptEnvironment();
-// are not vulnerable to CSRF.
-AphrontWriteGuard::allowDangerousUnguardedWrites(true);
-
-require_once dirname(dirname(__FILE__)).'/conf/__init_conf__.php';
-
-$env = isset($_SERVER['PHABRICATOR_ENV'])
-  ? $_SERVER['PHABRICATOR_ENV']
-  : getenv('PHABRICATOR_ENV');
-if (!$env) {
-  echo phutil_console_wrap(
-    phutil_console_format(
-      "**ERROR**: PHABRICATOR_ENV Not Set\n\n".
-      "Define the __PHABRICATOR_ENV__ environment variable before running ".
-      "this script. You can do it on the command line like this:\n\n".
-      "  $ PHABRICATOR_ENV=__custom/myconfig__ %s ...\n\n".
-      "Replace __custom/myconfig__ with the path to your configuration file. ".
-      "For more information, see the 'Configuration Guide' in the ".
-      "Phabricator documentation.\n\n",
-      $argv[0]));
-  exit(1);
 }
 
-$conf = phabricator_read_config_file($env);
+init_phabricator_script();
-$conf['phabricator.env'] = $env;
-
-PhabricatorEnv::setEnvConfig($conf);
-
-phutil_load_library('arcanist/src');
-
-foreach (PhabricatorEnv::getEnvConfig('load-libraries') as $library) {
-  phutil_load_library($library);
-}
-
-PhutilErrorHandler::initialize();
-PhabricatorEventEngine::initialize();
-
-$tz = PhabricatorEnv::getEnvConfig('phabricator.timezone');
-if ($tz) {
-  date_default_timezone_set($tz);
-}
-
-$translation = PhabricatorEnv::newObjectFromConfig('translation.provider');
-PhutilTranslator::getInstance()
-  ->setLanguage($translation->getLanguage())
-  ->addTranslations($translation->getTranslations());
-
-// Append any paths to $PATH if we need to.
-$paths = PhabricatorEnv::getEnvConfig('environment.append-paths');
-if (!empty($paths)) {
-  $current_env_path = getenv('PATH');
-  $new_env_paths = implode(PATH_SEPARATOR, $paths);
-  putenv('PATH='.$current_env_path.PATH_SEPARATOR.$new_env_paths);
-}

src/__phutil_library_map__.php

@@ -1080,7 +1080,6 @@ phutil_register_library_map(array(
     'PhabricatorRepositorySymbol' => 'applications/repository/storage/PhabricatorRepositorySymbol.php',
     'PhabricatorRepositoryTestCase' => 'applications/repository/storage/__tests__/PhabricatorRepositoryTestCase.php',
     'PhabricatorRepositoryType' => 'applications/repository/constants/PhabricatorRepositoryType.php',
-    'PhabricatorRequestOverseer' => 'infrastructure/PhabricatorRequestOverseer.php',
     'PhabricatorS3FileStorageEngine' => 'applications/files/engine/PhabricatorS3FileStorageEngine.php',
     'PhabricatorSQLPatchList' => 'infrastructure/storage/patch/PhabricatorSQLPatchList.php',
     'PhabricatorSSHWorkflow' => 'infrastructure/ssh/PhabricatorSSHWorkflow.php',

src/infrastructure/PhabricatorEnv.php

@@ -53,6 +53,102 @@ final class PhabricatorEnv {
   private static $env;
   private static $stack = array();
 
+  /**
+   * @phutil-external-symbol class PhabricatorStartup
+   */
+  public static function initializeWebEnvironment() {
+    $env = self::getSelectedEnvironmentName();
+    if (!$env) {
+      PhabricatorStartup::didFatal(
+        "The 'PHABRICATOR_ENV' environmental variable is not defined. Modify ".
+        "your httpd.conf to include 'SetEnv PHABRICATOR_ENV <env>', where ".
+        "'<env>' is one of 'development', 'production', or a custom ".
+        "environment.");
+    }
+
+    self::initializeCommonEnvironment();
+  }
+
+  public static function initializeScriptEnvironment() {
+    $env = self::getSelectedEnvironmentName();
+    if (!$env) {
+      echo phutil_console_wrap(
+        phutil_console_format(
+          "**ERROR**: PHABRICATOR_ENV Not Set\n\n".
+          "Define the __PHABRICATOR_ENV__ environment variable before ".
+          "running this script. You can do it on the command line like ".
+          "this:\n\n".
+          "  $ PHABRICATOR_ENV=__custom/myconfig__ %s ...\n\n".
+          "Replace __custom/myconfig__ with the path to your configuration ".
+          "file. For more information, see the 'Configuration Guide' in the ".
+          "Phabricator documentation.\n\n",
+          $GLOBALS['argv'][0]));
+      exit(1);
+    }
+
+    self::initializeCommonEnvironment();
+
+    // NOTE: This is dangerous in general, but we know we're in a script context
+    // and are not vulnerable to CSRF.
+    AphrontWriteGuard::allowDangerousUnguardedWrites(true);
+  }
+
+  /**
+   * @phutil-external-symbol function phabricator_read_config_file
+   */
+  private static function initializeCommonEnvironment() {
+    $env = self::getSelectedEnvironmentName();
+
+    $root = dirname(phutil_get_library_root('phabricator'));
+    require_once $root.'/conf/__init_conf__.php';
+    $conf = phabricator_read_config_file($env);
+    $conf['phabricator.env'] = $env;
+
+    PhabricatorEnv::setEnvConfig($conf);
+
+    PhutilErrorHandler::initialize();
+
+    $tz = PhabricatorEnv::getEnvConfig('phabricator.timezone');
+    if ($tz) {
+      date_default_timezone_set($tz);
+    }
+
+    // Append any paths to $PATH if we need to.
+    $paths = PhabricatorEnv::getEnvConfig('environment.append-paths');
+    if (!empty($paths)) {
+      $current_env_path = getenv('PATH');
+      $new_env_paths = implode(PATH_SEPARATOR, $paths);
+      putenv('PATH='.$current_env_path.PATH_SEPARATOR.$new_env_paths);
+    }
+
+    foreach (PhabricatorEnv::getEnvConfig('load-libraries') as $library) {
+      phutil_load_library($library);
+    }
+
+    PhabricatorEventEngine::initialize();
+
+    $translation = PhabricatorEnv::newObjectFromConfig('translation.provider');
+    PhutilTranslator::getInstance()
+      ->setLanguage($translation->getLanguage())
+      ->addTranslations($translation->getTranslations());
+  }
+
+  public static function getSelectedEnvironmentName() {
+    $env_var = 'PHABRICATOR_ENV';
+
+    $env = idx($_SERVER, $env_var);
+
+    if (!$env) {
+      $env = getenv($env_var);
+    }
+
+    if (!$env) {
+      $env = idx($_ENV, $env_var);
+    }
+
+    return $env;
+  }
+
 
 /* -(  Reading Configuration  )---------------------------------------------- */
 

src/infrastructure/PhabricatorRequestOverseer.php

@@ -1,98 +0,0 @@
-<?php
-
-final class PhabricatorRequestOverseer {
-
-  public function didStartup() {
-    $this->detectPostMaxSizeTriggered();
-  }
-
-  /**
-   * Detect if this request has had its POST data stripped by exceeding the
-   * 'post_max_size' PHP configuration limit.
-   *
-   * PHP has a setting called 'post_max_size'. If a POST request arrives with
-   * a body larger than the limit, PHP doesn't generate $_POST but processes
-   * the request anyway, and provides no formal way to detect that this
-   * happened.
-   *
-   * We can still read the entire body out of `php://input`. However according
-   * to the documentation the stream isn't available for "multipart/form-data"
-   * (on nginx + php-fpm it appears that it is available, though, at least) so
-   * any attempt to generate $_POST would be fragile.
-   *
-   * @phutil-external-symbol class PhabricatorStartup
-   */
-  private function detectPostMaxSizeTriggered() {
-    // If this wasn't a POST, we're fine.
-    if ($_SERVER['REQUEST_METHOD'] != 'POST') {
-      return;
-    }
-
-    // If there's POST data, clearly we're in good shape.
-    if ($_POST) {
-      return;
-    }
-
-    // For HTML5 drag-and-drop file uploads, Safari submits the data as
-    // "application/x-www-form-urlencoded". For most files this generates
-    // something in POST because most files decode to some nonempty (albeit
-    // meaningless) value. However, some files (particularly small images)
-    // don't decode to anything. If we know this is a drag-and-drop upload,
-    // we can skip this check.
-    if (isset($_REQUEST['__upload__'])) {
-      return;
-    }
-
-    // PHP generates $_POST only for two content types. This routing happens
-    // in `main/php_content_types.c` in PHP. Normally, all forms use one of
-    // these content types, but some requests may not -- for example, Firefox
-    // submits files sent over HTML5 XMLHTTPRequest APIs with the Content-Type
-    // of the file itself. If we don't have a recognized content type, we
-    // don't need $_POST.
-    //
-    // NOTE: We use strncmp() because the actual content type may be something
-    // like "multipart/form-data; boundary=...".
-    //
-    // NOTE: Chrome sometimes omits this header, see some discussion in T1762
-    // and http://code.google.com/p/chromium/issues/detail?id=6800
-    $content_type = idx($_SERVER, 'CONTENT_TYPE', '');
-
-    $parsed_types = array(
-      'application/x-www-form-urlencoded',
-      'multipart/form-data',
-    );
-
-    $is_parsed_type = false;
-    foreach ($parsed_types as $parsed_type) {
-      if (strncmp($content_type, $parsed_type, strlen($parsed_type)) === 0) {
-        $is_parsed_type = true;
-        break;
-      }
-    }
-
-    if (!$is_parsed_type) {
-      return;
-    }
-
-    // Check for 'Content-Length'. If there's no data, we don't expect $_POST
-    // to exist.
-    $length = (int)$_SERVER['CONTENT_LENGTH'];
-    if (!$length) {
-      return;
-    }
-
-    // Time to fatal: we know this was a POST with data that should have been
-    // populated into $_POST, but it wasn't.
-
-    $config = ini_get('post_max_size');
-    PhabricatorStartup::didFatal(
-      "As received by the server, this request had a nonzero content length ".
-      "but no POST data.\n\n".
-      "Normally, this indicates that it exceeds the 'post_max_size' setting ".
-      "in the PHP configuration on the server. Increase the 'post_max_size' ".
-      "setting or reduce the size of the request.\n\n".
-      "Request size according to 'Content-Length' was '{$length}', ".
-      "'post_max_size' is set to '{$config}'.");
-  }
-
-}

support/PhabricatorStartup.php

@@ -63,11 +63,6 @@ final class PhabricatorStartup {
     self::$startTime = microtime(true);
     self::$globals = array();
 
-    self::setupPHP();
-    self::verifyPHP();
-
-    self::verifyRewriteRules();
-
     static $registered;
     if (!$registered) {
       // NOTE: This protects us against multiple calls to didStartup() in the
@@ -76,6 +71,13 @@ final class PhabricatorStartup {
       register_shutdown_function(array(__CLASS__, 'didShutdown'));
       $registered = true;
     }
+
+    self::setupPHP();
+    self::verifyPHP();
+
+    self::verifyRewriteRules();
+
+    self::detectPostMaxSizeTriggered();
   }
 
 
@@ -118,6 +120,35 @@ final class PhabricatorStartup {
     self::didFatal($msg);
   }
 
+  public static function loadCoreLibraries() {
+    $phabricator_root = dirname(dirname(__FILE__));
+    $libraries_root = dirname($phabricator_root);
+
+    $root = null;
+    if (!empty($_SERVER['PHUTIL_LIBRARY_ROOT'])) {
+      $root = $_SERVER['PHUTIL_LIBRARY_ROOT'];
+    }
+
+    ini_set(
+      'include_path',
+      $libraries_root.PATH_SEPARATOR.ini_get('include_path'));
+
+    @include_once $root.'libphutil/src/__phutil_library_init__.php';
+    if (!@constant('__LIBPHUTIL__')) {
+      self::didFatal(
+        "Unable to load libphutil. Put libphutil/ next to phabricator/, or ".
+        "update your PHP 'include_path' to include the parent directory of ".
+        "libphutil/.");
+    }
+
+    phutil_load_library('arcanist/src');
+
+    // Load Phabricator itself using the absolute path, so we never end up doing
+    // anything surprising (loading index.php and libraries from different
+    // directories).
+    phutil_load_library($phabricator_root.'/src');
+  }
+
 
 /* -(  In Case of Apocalypse  )---------------------------------------------- */
 
@@ -219,4 +250,96 @@ final class PhabricatorStartup {
     }
   }
 
+
+  /**
+   * Detect if this request has had its POST data stripped by exceeding the
+   * 'post_max_size' PHP configuration limit.
+   *
+   * PHP has a setting called 'post_max_size'. If a POST request arrives with
+   * a body larger than the limit, PHP doesn't generate $_POST but processes
+   * the request anyway, and provides no formal way to detect that this
+   * happened.
+   *
+   * We can still read the entire body out of `php://input`. However according
+   * to the documentation the stream isn't available for "multipart/form-data"
+   * (on nginx + php-fpm it appears that it is available, though, at least) so
+   * any attempt to generate $_POST would be fragile.
+   *
+   * @task validation
+   */
+  private static function detectPostMaxSizeTriggered() {
+    // If this wasn't a POST, we're fine.
+    if ($_SERVER['REQUEST_METHOD'] != 'POST') {
+      return;
+    }
+
+    // If there's POST data, clearly we're in good shape.
+    if ($_POST) {
+      return;
+    }
+
+    // For HTML5 drag-and-drop file uploads, Safari submits the data as
+    // "application/x-www-form-urlencoded". For most files this generates
+    // something in POST because most files decode to some nonempty (albeit
+    // meaningless) value. However, some files (particularly small images)
+    // don't decode to anything. If we know this is a drag-and-drop upload,
+    // we can skip this check.
+    if (isset($_REQUEST['__upload__'])) {
+      return;
+    }
+
+    // PHP generates $_POST only for two content types. This routing happens
+    // in `main/php_content_types.c` in PHP. Normally, all forms use one of
+    // these content types, but some requests may not -- for example, Firefox
+    // submits files sent over HTML5 XMLHTTPRequest APIs with the Content-Type
+    // of the file itself. If we don't have a recognized content type, we
+    // don't need $_POST.
+    //
+    // NOTE: We use strncmp() because the actual content type may be something
+    // like "multipart/form-data; boundary=...".
+    //
+    // NOTE: Chrome sometimes omits this header, see some discussion in T1762
+    // and http://code.google.com/p/chromium/issues/detail?id=6800
+    $content_type = isset($_SERVER['CONTENT_TYPE'])
+      ? $_SERVER['CONTENT_TYPE']
+      : '';
+
+    $parsed_types = array(
+      'application/x-www-form-urlencoded',
+      'multipart/form-data',
+    );
+
+    $is_parsed_type = false;
+    foreach ($parsed_types as $parsed_type) {
+      if (strncmp($content_type, $parsed_type, strlen($parsed_type)) === 0) {
+        $is_parsed_type = true;
+        break;
+      }
+    }
+
+    if (!$is_parsed_type) {
+      return;
+    }
+
+    // Check for 'Content-Length'. If there's no data, we don't expect $_POST
+    // to exist.
+    $length = (int)$_SERVER['CONTENT_LENGTH'];
+    if (!$length) {
+      return;
+    }
+
+    // Time to fatal: we know this was a POST with data that should have been
+    // populated into $_POST, but it wasn't.
+
+    $config = ini_get('post_max_size');
+    PhabricatorStartup::didFatal(
+      "As received by the server, this request had a nonzero content length ".
+      "but no POST data.\n\n".
+      "Normally, this indicates that it exceeds the 'post_max_size' setting ".
+      "in the PHP configuration on the server. Increase the 'post_max_size' ".
+      "setting or reduce the size of the request.\n\n".
+      "Request size according to 'Content-Length' was '{$length}', ".
+      "'post_max_size' is set to '{$config}'.");
+  }
+
 }

webroot/index.php

@@ -3,50 +3,10 @@
 require_once dirname(dirname(__FILE__)).'/support/PhabricatorStartup.php';
 PhabricatorStartup::didStartup();
 
-$access_log = null;
-
-$env = getenv('PHABRICATOR_ENV'); // Apache
-if (!$env) {
-  if (isset($_ENV['PHABRICATOR_ENV'])) {
-    $env = $_ENV['PHABRICATOR_ENV'];
-  }
-}
-
-if (!$env) {
-  PhabricatorStartup::didFatal(
-    "The 'PHABRICATOR_ENV' environmental variable is not defined. Modify ".
-    "your httpd.conf to include 'SetEnv PHABRICATOR_ENV <env>', where '<env>' ".
-    "is one of 'development', 'production', or a custom environment.");
-}
-
-
-require_once dirname(dirname(__FILE__)).'/conf/__init_conf__.php';
-
 try {
-  setup_aphront_basics();
+  PhabricatorStartup::loadCoreLibraries();
 
-  $overseer = new PhabricatorRequestOverseer();
+  PhabricatorEnv::initializeWebEnvironment();
-  $overseer->didStartup();
-
-  $conf = phabricator_read_config_file($env);
-  $conf['phabricator.env'] = $env;
-
-  PhabricatorEnv::setEnvConfig($conf);
-
-  // This needs to be done before we create the log, because
-  // PhabricatorAccessLog::getLog() calls date()
-  $tz = PhabricatorEnv::getEnvConfig('phabricator.timezone');
-  if ($tz) {
-    date_default_timezone_set($tz);
-  }
-
-  // Append any paths to $PATH if we need to.
-  $paths = PhabricatorEnv::getEnvConfig('environment.append-paths');
-  if (!empty($paths)) {
-    $current_env_path = getenv('PATH');
-    $new_env_paths = implode(':', $paths);
-    putenv('PATH='.$current_env_path.':'.$new_env_paths);
-  }
 
   // This is the earliest we can get away with this, we need env config first.
   PhabricatorAccessLog::init();
@@ -63,15 +23,9 @@ try {
 
   DarkConsoleXHProfPluginAPI::hookProfiler();
 
-  PhutilErrorHandler::initialize();
-
   PhutilErrorHandler::setErrorListener(
     array('DarkConsoleErrorLogPluginAPI', 'handleErrors'));
 
-  foreach (PhabricatorEnv::getEnvConfig('load-libraries') as $library) {
-    phutil_load_library($library);
-  }
-
   if (PhabricatorEnv::getEnvConfig('phabricator.setup')) {
     try {
       PhabricatorSetup::runSetup();
@@ -84,11 +38,6 @@ try {
 
   phabricator_detect_bad_base_uri();
 
-  $translation = PhabricatorEnv::newObjectFromConfig('translation.provider');
-  PhutilTranslator::getInstance()
-    ->setLanguage($translation->getLanguage())
-    ->addTranslations($translation->getTranslations());
-
   $host = $_SERVER['HTTP_HOST'];
   $path = $_REQUEST['__path__'];
 
@@ -216,38 +165,6 @@ try {
   PhabricatorStartup::didFatal("[Exception] ".$ex->getMessage());
 }
 
-
-/**
- * @group aphront
- */
-function setup_aphront_basics() {
-  $aphront_root   = dirname(dirname(__FILE__));
-  $libraries_root = dirname($aphront_root);
-
-  $root = null;
-  if (!empty($_SERVER['PHUTIL_LIBRARY_ROOT'])) {
-    $root = $_SERVER['PHUTIL_LIBRARY_ROOT'];
-  }
-
-  ini_set(
-    'include_path',
-    $libraries_root.PATH_SEPARATOR.ini_get('include_path'));
-  @include_once $root.'libphutil/src/__phutil_library_init__.php';
-  if (!@constant('__LIBPHUTIL__')) {
-    echo "ERROR: Unable to load libphutil. Put libphutil/ next to ".
-         "phabricator/, or update your PHP 'include_path' to include ".
-         "the parent directory of libphutil/.\n";
-    exit(1);
-  }
-
-  // Load Phabricator itself using the absolute path, so we never end up doing
-  // anything surprising (loading index.php and libraries from different
-  // directories).
-  phutil_load_library($aphront_root.'/src');
-  phutil_load_library('arcanist/src');
-
-}
-
 function phabricator_detect_bad_base_uri() {
   $conf = PhabricatorEnv::getEnvConfig('phabricator.base-uri');
   $uri = new PhutilURI($conf);