+global\MimeMagic::doGuessMimeType()

Complexity

Conditions	35
Paths	317

Size

Total Lines	187
Code Lines	104

Duplication

Lines	5
Ratio	2.67 %

<?php
	private function doGuessMimeType( $file, $ext ) {
		// Read a chunk of the file
		MediaWiki\suppressWarnings();
		$f = fopen( $file, 'rb' );
		MediaWiki\restoreWarnings();

		if ( !$f ) {
			return 'unknown/unknown';
		}

		$fsize = filesize( $file );
		if ( $fsize === false ) {
			return 'unknown/unknown';
		}

		$head = fread( $f, 1024 );
		$tailLength = min( 65558, $fsize ); // 65558 = maximum size of a zip EOCDR
		if ( fseek( $f, -1 * $tailLength, SEEK_END ) === -1 ) {
			throw new MWException(
				"Seeking $tailLength bytes from EOF failed in " . __METHOD__ );
		}
		$tail = $tailLength ? fread( $f, $tailLength ) : '';
		fclose( $f );

		wfDebug( __METHOD__ . ": analyzing head and tail of $file for magic numbers.\n" );

		// Hardcode a few magic number checks...
		$headers = [
			// Multimedia...
			'MThd'             => 'audio/midi',
			'OggS'             => 'application/ogg',

			// Image formats...
			// Note that WMF may have a bare header, no magic number.
			"\x01\x00\x09\x00" => 'application/x-msmetafile', // Possibly prone to false positives?
			"\xd7\xcd\xc6\x9a" => 'application/x-msmetafile',
			'%PDF'             => 'application/pdf',
			'gimp xcf'         => 'image/x-xcf',

			// Some forbidden fruit...
			'MZ'               => 'application/octet-stream', // DOS/Windows executable
			"\xca\xfe\xba\xbe" => 'application/octet-stream', // Mach-O binary
			"\x7fELF"          => 'application/octet-stream', // ELF binary
		];

		foreach ( $headers as $magic => $candidate ) {
			if ( strncmp( $head, $magic, strlen( $magic ) ) == 0 ) {
				wfDebug( __METHOD__ . ": magic header in $file recognized as $candidate\n" );
				return $candidate;
			}
		}

		/* Look for WebM and Matroska files */
		if ( strncmp( $head, pack( "C4", 0x1a, 0x45, 0xdf, 0xa3 ), 4 ) == 0 ) {
			$doctype = strpos( $head, "\x42\x82" );
			if ( $doctype ) {
				// Next byte is datasize, then data (sizes larger than 1 byte are very stupid muxers)
				$data = substr( $head, $doctype + 3, 8 );
				if ( strncmp( $data, "matroska", 8 ) == 0 ) {
					wfDebug( __METHOD__ . ": recognized file as video/x-matroska\n" );
					return "video/x-matroska";
				} elseif ( strncmp( $data, "webm", 4 ) == 0 ) {
					wfDebug( __METHOD__ . ": recognized file as video/webm\n" );
					return "video/webm";
				}
			}
			wfDebug( __METHOD__ . ": unknown EBML file\n" );
			return "unknown/unknown";
		}

		/* Look for WebP */
		if ( strncmp( $head, "RIFF", 4 ) == 0 && strncmp( substr( $head, 8, 7 ), "WEBPVP8", 7 ) == 0 ) {
			wfDebug( __METHOD__ . ": recognized file as image/webp\n" );
			return "image/webp";
		}

		/**
		 * Look for PHP.  Check for this before HTML/XML...  Warning: this is a
		 * heuristic, and won't match a file with a lot of non-PHP before.  It
		 * will also match text files which could be PHP. :)
		 *
		 * @todo FIXME: For this reason, the check is probably useless -- an attacker
		 * could almost certainly just pad the file with a lot of nonsense to
		 * circumvent the check in any case where it would be a security
		 * problem.  On the other hand, it causes harmful false positives (bug
		 * 16583).  The heuristic has been cut down to exclude three-character
		 * strings like "<? ", but should it be axed completely?
		 */
		if ( ( strpos( $head, '<?php' ) !== false ) ||
			( strpos( $head, "<\x00?\x00p\x00h\x00p" ) !== false ) ||
			( strpos( $head, "<\x00?\x00 " ) !== false ) ||
			( strpos( $head, "<\x00?\x00\n" ) !== false ) ||
			( strpos( $head, "<\x00?\x00\t" ) !== false ) ||
			( strpos( $head, "<\x00?\x00=" ) !== false ) ) {

			wfDebug( __METHOD__ . ": recognized $file as application/x-php\n" );
			return 'application/x-php';
		}

		/**
		 * look for XML formats (XHTML and SVG)
		 */
		$xml = new XmlTypeCheck( $file );
		if ( $xml->wellFormed ) {
			$xmlMimeTypes = $this->mConfig->get( 'XMLMimeTypes' );
			if ( isset( $xmlMimeTypes[$xml->getRootElement()] ) ) {
				return $xmlMimeTypes[$xml->getRootElement()];
			} else {
				return 'application/xml';
			}
		}

		/**
		 * look for shell scripts
		 */
		$script_type = null;

		# detect by shebang
		if ( substr( $head, 0, 2 ) == "#!" ) {
			$script_type = "ASCII";
		} elseif ( substr( $head, 0, 5 ) == "\xef\xbb\xbf#!" ) {
			$script_type = "UTF-8";
		} elseif ( substr( $head, 0, 7 ) == "\xfe\xff\x00#\x00!" ) {
			$script_type = "UTF-16BE";
		} elseif ( substr( $head, 0, 7 ) == "\xff\xfe#\x00!" ) {
			$script_type = "UTF-16LE";
		}

		if ( $script_type ) {

The expression $script_type of type string|null is loosely compared to true; this is ambiguous if the string can be empty. You might want to explicitly use !== null instead.

			if ( $script_type !== "UTF-8" && $script_type !== "ASCII" ) {
				// Quick and dirty fold down to ASCII!
				$pack = [ 'UTF-16BE' => 'n*', 'UTF-16LE' => 'v*' ];
				$chars = unpack( $pack[$script_type], substr( $head, 2 ) );
				$head = '';
				foreach ( $chars as $codepoint ) {
					if ( $codepoint < 128 ) {
						$head .= chr( $codepoint );
					} else {
						$head .= '?';
					}
				}
			}

			$match = [];

			if ( preg_match( '%/?([^\s]+/)(\w+)%', $head, $match ) ) {
				$mime = "application/x-{$match[2]}";
				wfDebug( __METHOD__ . ": shell script recognized as $mime\n" );
				return $mime;
			}
		}

		// Check for ZIP variants (before getimagesize)
		if ( strpos( $tail, "PK\x05\x06" ) !== false ) {
			wfDebug( __METHOD__ . ": ZIP header present in $file\n" );
			return $this->detectZipType( $head, $tail, $ext );
		}

		MediaWiki\suppressWarnings();
		$gis = getimagesize( $file );
		MediaWiki\restoreWarnings();

		if ( $gis && isset( $gis['mime'] ) ) {

The expression $gis of type array is implicitly converted to a boolean; are you sure this is intended? If so, consider using ! empty($expr) instead to make it clear that you intend to check for an array without elements.

			$mime = $gis['mime'];
			wfDebug( __METHOD__ . ": getimagesize detected $file as $mime\n" );
			return $mime;
		}

		// Also test DjVu
		$deja = new DjVuImage( $file );
		if ( $deja->isValid() ) {
			wfDebug( __METHOD__ . ": detected $file as image/vnd.djvu\n" );
			return 'image/vnd.djvu';
		}

		# Media handling extensions can guess the MIME by content
		# It's intentionally here so that if core is wrong about a type (false positive),
		# people will hopefully nag and submit patches :)
		$mime = false;
		# Some strings by reference for performance - assuming well-behaved hooks
		Hooks::run(
			'MimeMagicGuessFromContent',
			[ $this, &$head, &$tail, $file, &$mime ]
		);

		return $mime;
	}