* @param $method string HTTP method. Usually GET/POST
* @param $url string Full URL to act on
* @param $options options to pass to HttpRequest object
+ * Possible keys for the array:
+ * timeout Timeout length in seconds
+ * postData An array of key-value pairs or a url-encoded form data
+ * proxy The proxy to use. Will use $wgHTTPProxy (if set) otherwise.
+ * noProxy Override $wgHTTPProxy (if set) and don't use any proxy at all.
+ * sslVerifyHost (curl only) Verify the SSL certificate
+ * caInfo (curl only) Provide CA information
+ * maxRedirects Maximum number of redirects to follow (defaults to 5)
+ * followRedirects Whether to follow redirects (defaults to true)
* @returns mixed (bool)false on failure or a string on success
*/
public static function request( $method, $url, $options = array() ) {
protected $url;
protected $parsedUrl;
protected $callback;
+ protected $maxRedirects = 5;
+ protected $followRedirects = true;
+
+ protected $cookieJar;
+
+ protected $headerList = array();
+ protected $respVersion = "0.9";
+ protected $respStatus = "200 Ok";
+ protected $respHeaders = array();
+
public $status;
/**
* @param $url string url to use
- * @param $options array (optional) extra params to pass
- * Possible keys for the array:
- * method
- * timeout
- * targetFilePath
- * requestKey
- * postData
- * proxy
- * noProxy
- * sslVerifyHost
- * caInfo
+ * @param $options array (optional) extra params to pass (see Http::request())
*/
function __construct( $url, $options = array() ) {
global $wgHTTPTimeout;
$this->parsedUrl = parse_url( $url );
if ( !Http::isValidURI( $this->url ) ) {
- $this->status = Status::newFromFatal('http-invalid-url');
+ $this->status = Status::newFatal('http-invalid-url');
} else {
$this->status = Status::newGood( 100 ); // continue
}
$this->timeout = $wgHTTPTimeout;
}
- $members = array( "targetFilePath", "requestKey", "postData",
- "proxy", "noProxy", "sslVerifyHost", "caInfo", "method" );
+ $members = array( "postData", "proxy", "noProxy", "sslVerifyHost", "caInfo",
+ "method", "followRedirects", "maxRedirects" );
foreach ( $members as $o ) {
if ( isset($options[$o]) ) {
$this->$o = $options[$o];
* Generate a new request object
* @see HttpRequest::__construct
*/
- public static function factory( $url, $options ) {
+ public static function factory( $url, $options = null ) {
if ( !Http::$httpEngine ) {
Http::$httpEngine = function_exists( 'curl_init' ) ? 'curl' : 'php';
} elseif ( Http::$httpEngine == 'curl' && !function_exists( 'curl_init' ) ) {
- throw new MWException( __METHOD__.': curl (http://php.net/curl) is not installed, but Http::$httpEngine is set to "curl"' );
+ throw new MWException( __METHOD__.': curl (http://php.net/curl) is not installed, but'.
+ ' Http::$httpEngine is set to "curl"' );
}
switch( Http::$httpEngine ) {
return new CurlHttpRequest( $url, $options );
case 'php':
if ( !wfIniGetBool( 'allow_url_fopen' ) ) {
- throw new MWException( __METHOD__.': allow_url_fopen needs to be enabled for pure PHP http requests to work. '.
- 'If possible, curl should be used instead. See http://php.net/curl.' );
+ throw new MWException( __METHOD__.': allow_url_fopen needs to be enabled for pure PHP'.
+ ' http requests to work. If possible, curl should be used instead. See http://php.net/curl.' );
}
return new PhpHttpRequest( $url, $options );
default:
public function proxySetup() {
global $wgHTTPProxy;
-
if ( $this->proxy ) {
return;
}
$this->proxy = 'http://localhost:80/';
} elseif ( $wgHTTPProxy ) {
$this->proxy = $wgHTTPProxy ;
+ } elseif ( getenv( "http_proxy" ) ) {
+ $this->proxy = getenv( "http_proxy" );
}
}
public function getHeaderList() {
$list = array();
+ if( $this->cookieJar ) {
+ $this->reqHeaders['Cookie'] =
+ $this->cookieJar->serializeToHttpRequest($this->parsedUrl['path'],
+ $this->parsedUrl['host']);
+ }
foreach($this->reqHeaders as $name => $value) {
$list[] = "$name: $value";
}
}
/**
- * A generic callback to read in the response from a remote server
+ * A generic callback to read the body of the response from a remote
+ * server.
* @param $fh handle
* @param $content string
*/
$this->setUserAgent(Http::userAgent());
}
}
+
+ /**
+ * Parses the headers, including the HTTP status code and any
+ * Set-Cookie headers. This function expectes the headers to be
+ * found in an array in the member variable headerList.
+ * @returns nothing
+ */
+ protected function parseHeader() {
+ $lastname = "";
+ foreach( $this->headerList as $header ) {
+ if( preg_match( "#^HTTP/([0-9.]+) (.*)#", $header, $match ) ) {
+ $this->respVersion = $match[1];
+ $this->respStatus = $match[2];
+ } elseif( preg_match( "#^[ \t]#", $header ) ) {
+ $last = count($this->respHeaders[$lastname]) - 1;
+ $this->respHeaders[$lastname][$last] .= "\r\n$header";
+ } elseif( preg_match( "#^([^:]*):[\t ]*(.*)#", $header, $match ) ) {
+ $this->respHeaders[strtolower( $match[1] )][] = $match[2];
+ $lastname = strtolower( $match[1] );
+ }
+ }
+
+ $this->parseCookies();
+ }
+
+ /**
+ * Sets the member variable status to a fatal status if the HTTP
+ * status code was not 200.
+ * @returns nothing
+ */
+ protected function setStatus() {
+ if( !$this->respHeaders ) {
+ $this->parseHeader();
+ }
+
+ if((int)$this->respStatus !== 200) {
+ list( $code, $message ) = explode(" ", $this->respStatus, 2);
+ $this->status->fatal("http-bad-status", $code, $message );
+ }
+ }
+
+
+ /**
+ * Returns true if the last status code was a redirect.
+ * @return bool
+ */
+ public function isRedirect() {
+ if( !$this->respHeaders ) {
+ $this->parseHeader();
+ }
+
+ $status = (int)$this->respStatus;
+ if ( $status >= 300 && $status < 400 ) {
+ return true;
+ }
+ return false;
+ }
+
+ /**
+ * Returns an associative array of response headers after the
+ * request has been executed. Because some headers
+ * (e.g. Set-Cookie) can appear more than once the, each value of
+ * the associative array is an array of the values given.
+ * @return array
+ */
+ public function getResponseHeaders() {
+ if( !$this->respHeaders ) {
+ $this->parseHeader();
+ }
+ return $this->respHeaders;
+ }
+
+ /**
+ * Returns the value of the given response header.
+ * @param $header string
+ * @return string
+ */
+ public function getResponseHeader($header) {
+ if( !$this->respHeaders ) {
+ $this->parseHeader();
+ }
+ if ( isset( $this->respHeaders[strtolower ( $header ) ] ) ) {
+ $v = $this->respHeaders[strtolower ( $header ) ];
+ return $v[count( $v ) - 1];
+ }
+ return null;
+ }
+
+ /**
+ * Tells the HttpRequest object to use this pre-loaded CookieJar.
+ * @param $jar CookieJar
+ */
+ public function setCookieJar( $jar ) {
+ $this->cookieJar = $jar;
+ }
+
+ /**
+ * Returns the cookie jar in use.
+ * @returns CookieJar
+ */
+ public function getCookieJar() {
+ if( !$this->respHeaders ) {
+ $this->parseHeader();
+ }
+ return $this->cookieJar;
+ }
+
+ /**
+ * Sets a cookie. Used before a request to set up any individual
+ * cookies. Used internally after a request to parse the
+ * Set-Cookie headers.
+ * @see Cookie::set
+ */
+ public function setCookie( $name, $value = null, $attr = null) {
+ if( !$this->cookieJar ) {
+ $this->cookieJar = new CookieJar;
+ }
+ $this->cookieJar->setCookie($name, $value, $attr);
+ }
+
+ /**
+ * Parse the cookies in the response headers and store them in the cookie jar.
+ */
+ protected function parseCookies() {
+ if( !$this->cookieJar ) {
+ $this->cookieJar = new CookieJar;
+ }
+ if( isset( $this->respHeaders['set-cookie'] ) ) {
+ $url = parse_url( $this->getFinalUrl() );
+ foreach( $this->respHeaders['set-cookie'] as $cookie ) {
+ $this->cookieJar->parseCookieResponseHeader( $cookie, $url['host'] );
+ }
+ }
+ }
+
+ /**
+ * Returns the final URL after all redirections.
+ * @returns string
+ */
+ public function getFinalUrl() {
+ $location = $this->getResponseHeader("Location");
+ if ( $location ) {
+ return $location;
+ }
+
+ return $this->url;
+ }
}
+
+class Cookie {
+ protected $name;
+ protected $value;
+ protected $expires;
+ protected $path;
+ protected $domain;
+ protected $isSessionKey = true;
+ // TO IMPLEMENT protected $secure
+ // TO IMPLEMENT? protected $maxAge (add onto expires)
+ // TO IMPLEMENT? protected $version
+ // TO IMPLEMENT? protected $comment
+
+ function __construct( $name, $value, $attr ) {
+ $this->name = $name;
+ $this->set( $value, $attr );
+ }
+
+ /**
+ * Sets a cookie. Used before a request to set up any individual
+ * cookies. Used internally after a request to parse the
+ * Set-Cookie headers.
+ * @param $name string the name of the cookie
+ * @param $value string the value of the cookie
+ * @param $attr array possible key/values:
+ * expires A date string
+ * path The path this cookie is used on
+ * domain Domain this cookie is used on
+ */
+ public function set( $value, $attr ) {
+ $this->value = $value;
+ if( isset( $attr['expires'] ) ) {
+ $this->isSessionKey = false;
+ $this->expires = strtotime( $attr['expires'] );
+ }
+ if( isset( $attr['path'] ) ) {
+ $this->path = $attr['path'];
+ } else {
+ $this->path = "/";
+ }
+ if( isset( $attr['domain'] ) ) {
+ if( self::validateCookieDomain( $attr['domain'] ) ) {
+ $this->domain = $attr['domain'];
+ }
+ } else {
+ throw new MWException("You must specify a domain.");
+ }
+ }
+
+ /**
+ * Return the true if the cookie is valid is valid. Otherwise,
+ * false. The uses a method similar to IE cookie security
+ * described here:
+ * http://kuza55.blogspot.com/2008/02/understanding-cookie-security.html
+ * A better method might be to use a blacklist like
+ * http://publicsuffix.org/
+ *
+ * @param $domain string the domain to validate
+ * @param $originDomain string (optional) the domain the cookie originates from
+ * @return bool
+ */
+ public static function validateCookieDomain( $domain, $originDomain = null) {
+ // Don't allow a trailing dot
+ if( substr( $domain, -1 ) == "." ) return false;
+
+ $dc = explode(".", $domain);
+
+ // Don't allow cookies for "localhost", "ls" or other dot-less hosts
+ if( count($dc) < 2 ) return false;
+
+ // Only allow full, valid IP addresses
+ if( preg_match( '/^[0-9.]+$/', $domain ) ) {
+ if( count( $dc ) != 4 ) return false;
+
+ if( ip2long( $domain ) === false ) return false;
+
+ if( $originDomain == null || $originDomain == $domain ) return true;
+
+ }
+
+ // Don't allow cookies for "co.uk" or "gov.uk", etc, but allow "supermarket.uk"
+ if( strrpos( $domain, "." ) - strlen( $domain ) == -3 ) {
+ if( (count($dc) == 2 && strlen( $dc[0] ) <= 2 )
+ || (count($dc) == 3 && strlen( $dc[0] ) == "" && strlen( $dc[1] ) <= 2 ) ) {
+ return false;
+ }
+ if( (count($dc) == 2 || (count($dc) == 3 && $dc[0] == "") )
+ && preg_match( '/(com|net|org|gov|edu)\...$/', $domain) ) {
+ return false;
+ }
+ }
+
+ if( $originDomain != null ) {
+ if( substr( $domain, 0, 1 ) != "." && $domain != $originDomain ) {
+ return false;
+ }
+ if( substr( $domain, 0, 1 ) == "."
+ && substr_compare( $originDomain, $domain, -strlen( $domain ),
+ strlen( $domain ), TRUE ) != 0 ) {
+ return false;
+ }
+ }
+
+ return true;
+ }
+
+ /**
+ * Serialize the cookie jar into a format useful for HTTP Request headers.
+ * @param $path string the path that will be used. Required.
+ * @param $domain string the domain that will be used. Required.
+ * @return string
+ */
+ public function serializeToHttpRequest( $path, $domain ) {
+ $ret = "";
+
+ if( $this->canServeDomain( $domain )
+ && $this->canServePath( $path )
+ && $this->isUnExpired() ) {
+ $ret = $this->name ."=". $this->value;
+ }
+
+ return $ret;
+ }
+
+ protected function canServeDomain( $domain ) {
+ if( $domain == $this->domain
+ || ( strlen( $domain) > strlen( $this->domain )
+ && substr( $this->domain, 0, 1) == "."
+ && substr_compare( $domain, $this->domain, -strlen( $this->domain ),
+ strlen( $this->domain ), TRUE ) == 0 ) ) {
+ return true;
+ }
+ return false;
+ }
+
+ protected function canServePath( $path ) {
+ if( $this->path && substr_compare( $this->path, $path, 0, strlen( $this->path ) ) == 0 ) {
+ return true;
+ }
+ return false;
+ }
+
+ protected function isUnExpired() {
+ if( $this->isSessionKey || $this->expires > time() ) {
+ return true;
+ }
+ return false;
+ }
+
+}
+
+class CookieJar {
+ private $cookie = array();
+
+ /**
+ * Set a cookie in the cookie jar. Make sure only one cookie per-name exists.
+ * @see Cookie::set()
+ */
+ public function setCookie ($name, $value, $attr) {
+ /* cookies: case insensitive, so this should work.
+ * We'll still send the cookies back in the same case we got them, though.
+ */
+ $index = strtoupper($name);
+ if( isset( $this->cookie[$index] ) ) {
+ $this->cookie[$index]->set( $value, $attr );
+ } else {
+ $this->cookie[$index] = new Cookie( $name, $value, $attr );
+ }
+ }
+
+ /**
+ * @see Cookie::serializeToHttpRequest
+ */
+ public function serializeToHttpRequest( $path, $domain ) {
+ $cookies = array();
+
+ foreach( $this->cookie as $c ) {
+ $serialized = $c->serializeToHttpRequest( $path, $domain );
+ if ( $serialized ) $cookies[] = $serialized;
+ }
+
+ return implode("; ", $cookies);
+ }
+
+ /**
+ * Parse the content of an Set-Cookie HTTP Response header.
+ * @param $cookie string
+ */
+ public function parseCookieResponseHeader ( $cookie, $domain ) {
+ $len = strlen( "Set-Cookie:" );
+ if ( substr_compare( "Set-Cookie:", $cookie, 0, $len, TRUE ) === 0 ) {
+ $cookie = substr( $cookie, $len );
+ }
+
+ $bit = array_map( 'trim', explode( ";", $cookie ) );
+ if ( count($bit) >= 1 ) {
+ list($name, $value) = explode( "=", array_shift( $bit ), 2 );
+ $attr = array();
+ foreach( $bit as $piece ) {
+ $parts = explode( "=", $piece );
+ if( count( $parts ) > 1 ) {
+ $attr[strtolower( $parts[0] )] = $parts[1];
+ } else {
+ $attr[strtolower( $parts[0] )] = true;
+ }
+ }
+
+ if( !isset( $attr['domain'] ) ) {
+ $attr['domain'] = $domain;
+ } elseif ( !Cookie::validateCookieDomain( $attr['domain'], $domain ) ) {
+ return null;
+ }
+
+ $this->setCookie( $name, $value, $attr );
+ }
+ }
+}
+
+
/**
* HttpRequest implemented using internal curl compiled into PHP
*/
class CurlHttpRequest extends HttpRequest {
+ static $curlMessageMap = array(
+ 6 => 'http-host-unreachable',
+ 28 => 'http-timed-out'
+ );
+
protected $curlOptions = array();
+ protected $headerText = "";
+
+ protected function readHeader( $fh, $content ) {
+ $this->headerText .= $content;
+ return strlen( $content );
+ }
public function execute() {
parent::execute();
if ( !$this->status->isOK() ) {
return $this->status;
}
-
- // A lot of the action up front should probably be in
- // set* methods, but we'll leave that for another time.
-
$this->curlOptions[CURLOPT_PROXY] = $this->proxy;
$this->curlOptions[CURLOPT_TIMEOUT] = $this->timeout;
$this->curlOptions[CURLOPT_HTTP_VERSION] = CURL_HTTP_VERSION_1_0;
$this->curlOptions[CURLOPT_WRITEFUNCTION] = $this->callback;
+ $this->curlOptions[CURLOPT_HEADERFUNCTION] = array($this, "readHeader");
+ $this->curlOptions[CURLOPT_FOLLOWLOCATION] = $this->followRedirects;
+ $this->curlOptions[CURLOPT_MAXREDIRS] = $this->maxRedirects;
/* not sure these two are actually necessary */
if(isset($this->reqHeaders['Referer'])) {
curl_setopt_array( $curlHandle, $this->curlOptions );
if ( false === curl_exec( $curlHandle ) ) {
- // re-using already translated error messages
- $this->status->fatal( 'upload-curl-error'.curl_errno( $curlHandle ).'-text' );
+ $code = curl_error( $curlHandle );
+
+ if ( isset( self::$curlMessageMap[$code] ) ) {
+ $this->status->fatal( self::$curlMessageMap[$code] );
+ } else {
+ $this->status->fatal( 'http-curl-error', curl_error( $curlHandle ) );
+ }
+ } else {
+ $this->headerList = explode("\r\n", $this->headerText);
}
curl_close( $curlHandle );
+ $this->parseHeader();
+ $this->setStatus();
return $this->status;
}
}
class PhpHttpRequest extends HttpRequest {
- private $fh;
+ protected $manuallyRedirect = false;
protected function urlToTcp( $url ) {
$parsedUrl = parse_url( $url );
}
public function execute() {
- if ( $this->parsedUrl['scheme'] != 'http' ) {
- $this->status->fatal( 'http-invalid-scheme', $this->parsedURL['scheme'] );
- }
-
parent::execute();
- if ( !$this->status->isOK() ) {
- return $this->status;
+
+ // At least on Centos 4.8 with PHP 5.1.6, using max_redirects to follow redirects
+ // causes a segfault
+ if ( version_compare( '5.1.7', phpversion(), '>' ) ) {
+ $this->manuallyRedirect = true;
}
- // A lot of the action up front should probably be in
- // set* methods, but we'll leave that for another time.
+ if ( $this->parsedUrl['scheme'] != 'http' ) {
+ $this->status->fatal( 'http-invalid-scheme', $this->parsedUrl['scheme'] );
+ }
$this->reqHeaders['Accept'] = "*/*";
if ( $this->method == 'POST' ) {
$options['request_fulluri'] = true;
}
+ if ( !$this->followRedirects || $this->manuallyRedirect ) {
+ $options['max_redirects'] = 0;
+ } else {
+ $options['max_redirects'] = $this->maxRedirects;
+ }
+
$options['method'] = $this->method;
- $options['timeout'] = $this->timeout;
$options['header'] = implode("\r\n", $this->getHeaderList());
- // FOR NOW: Force everyone to HTTP 1.0
- /* if ( version_compare( "5.3.0", phpversion(), ">" ) ) { */
- $options['protocol_version'] = "1.0";
- /* } else { */
- /* $options['protocol_version'] = "1.1"; */
- /* } */
+ // Note that at some future point we may want to support
+ // HTTP/1.1, but we'd have to write support for chunking
+ // in version of PHP < 5.3.1
+ $options['protocol_version'] = "1.0";
+
+ // This is how we tell PHP we want to deal with 404s (for example) ourselves.
+ // Only works on 5.2.10+
+ $options['ignore_errors'] = true;
if ( $this->postData ) {
$options['content'] = $this->postData;
}
+ $oldTimeout = false;
+ if ( version_compare( '5.2.1', phpversion(), '>' ) ) {
+ $oldTimeout = ini_set('default_socket_timeout', $this->timeout);
+ } else {
+ $options['timeout'] = $this->timeout;
+ }
+
$context = stream_context_create( array( 'http' => $options ) );
- try {
- $this->fh = fopen( $this->url, "r", false, $context );
- } catch ( Exception $e ) {
- $this->status->fatal( $e->getMessage() ); /* need some l10n help */
+
+ $this->headerList = array();
+ $reqCount = 0;
+ $url = $this->url;
+ do {
+ $again = false;
+ $reqCount++;
+ wfSuppressWarnings();
+ $fh = fopen( $url, "r", false, $context );
+ wfRestoreWarnings();
+ if ( $fh ) {
+ $result = stream_get_meta_data( $fh );
+ $this->headerList = $result['wrapper_data'];
+ $this->parseHeader();
+ $url = $this->getResponseHeader("Location");
+ $again = $this->manuallyRedirect && $this->followRedirects && $url
+ && $this->isRedirect() && $this->maxRedirects > $reqCount;
+ }
+ } while ( $again );
+
+ if ( $oldTimeout !== false ) {
+ ini_set('default_socket_timeout', $oldTimeout);
+ }
+ $this->setStatus();
+
+ if ( $fh === false ) {
+ $this->status->fatal( 'http-request-error' );
return $this->status;
}
- $result = stream_get_meta_data( $this->fh );
if ( $result['timed_out'] ) {
$this->status->fatal( 'http-timed-out', $this->url );
return $this->status;
}
- $this->headers = $result['wrapper_data'];
-
- $end = false;
- while ( !$end ) {
- $contents = fread( $this->fh, 8192 );
- $size = 0;
- if ( $contents ) {
- $size = call_user_func_array( $this->callback, array( $this->fh, $contents ) );
+ if($this->status->isOK()) {
+ while ( !feof( $fh ) ) {
+ $buf = fread( $fh, 8192 );
+ if ( $buf === false ) {
+ $this->status->fatal( 'http-read-error' );
+ break;
+ }
+ if ( strlen( $buf ) ) {
+ call_user_func( $this->callback, $fh, $buf );
+ }
}
- $end = ( $size == 0 ) || feof( $this->fh );
}
- fclose( $this->fh );
+ fclose( $fh );
return $this->status;
}