Don't fail on libxml errors if the RSD URL can still be found

Various things, such as duplicate element IDs or repeated
attributes, break DOMDocument::loadHTML() and so this turns off
the direct reporting (E_ERROR) of these and istead adds them
to the RsdException if the RSD URL really can't be determined
from the HTML. In most cases, the URL can be found correctly
and the errors can be disregarded.

A test is added for this as well, although it does *not* test
for the case when the RSD URL can't be found and there are
libxml errors (because we need to serve up a broken HTML file,
and the mediawiki-api-base test system can only interact with
MediaWiki via the API, which makes it hard to produce broken
HTML that doesn't also have the correct LINK element for the
RSD URL).

A new TestEnvironment::savePage method is added, for easier
creation of test wiki pages.

Bug: https://phabricator.wikimedia.org/T163527

src/MediawikiApi.php

@@ -79,15 +79,30 @@ public static function newFromApiEndpoint( $apiEndpoint ) {
 	 * @throws RsdException If the RSD URL could not be found in the page's HTML.
 	 */
 	public static function newFromPage( $url ) {
+		// Set up HTTP client and HTML document.
 		$tempClient = new Client( [ 'headers' => [ 'User-Agent' => 'addwiki-mediawiki-client' ] ] );
-
-		// Get the page HTML and extract the RSD link.
 		$pageHtml = $tempClient->get( $url )->getBody();
 		$pageDoc = new DOMDocument();
+
+		// Try to load the HTML (turn off errors temporarily; most don't matter, and if they do get
+		// in the way of finding the API URL, will be reported in the RsdException below).
+		$internalErrors = libxml_use_internal_errors( true );
 		$pageDoc->loadHTML( $pageHtml );
-		$link = ( new DOMXpath( $pageDoc ) )->query( 'head/link[@type="application/rsd+xml"][@href]' );
+		$libXmlErrors = libxml_get_errors();
+		libxml_use_internal_errors( $internalErrors );
+
+		// Extract the RSD link.
+		$xpath = 'head/link[@type="application/rsd+xml"][@href]';
+		$link = ( new DOMXpath( $pageDoc ) )->query( $xpath );
 		if ( $link->length === 0 ) {
-			throw new RsdException( "Unable to find RSD URL in page: $url" );
+			// Format libxml errors for display.
+			$libXmlErrorStr = array_reduce( $libXmlErrors, function( $prevErr, $err ) {
+				return $prevErr . ', ' . $err->message . ' (line '.$err->line . ')';
+			} );
+			if ( $libXmlErrorStr ) {
+				$libXmlErrorStr = "In addition, libxml had the following errors: $libXmlErrorStr";
+			}
+			throw new RsdException( "Unable to find RSD URL in page: $url $libXmlErrorStr" );
 		}
 		$rsdUrl = $link->item( 0 )->attributes->getnamedItem( 'href' )->nodeValue;
 

tests/Integration/MediawikiApiTest.php

@@ -30,6 +30,28 @@ public function testNewFromPageInvalidHtml() {
 		MediawikiApi::newFromPage( $nonWikiPage );
 	}
 
+	/**
+	 * Duplicate element IDs break DOMDocument::loadHTML
+	 * @see https://phabricator.wikimedia.org/T163527#3219833
+	 * @covers Mediawiki\Api\MediawikiApi::newFromPage
+	 */
+	public function testNewFromPageWithDuplicateId() {
+		$testPageName = __METHOD__;
+		$testEnv = TestEnvironment::newInstance();
+		$wikiPageUrl = str_replace( 'api.php', "index.php?title=$testPageName", $testEnv->getApiUrl() );
+
+		// Test with no duplicate IDs.
+		$testEnv->savePage( $testPageName, '<p id="unique-id"></p>' );
+		$api1 = MediawikiApi::newFromPage( $wikiPageUrl );
+		$this->assertInstanceOf( MediawikiApi::class, $api1 );
+
+		// Test with duplicate ID.
+		$wikiText = '<p id="duplicated-id"></p><div id="duplicated-id"></div>';
+		$testEnv->savePage( $testPageName, $wikiText );
+		$api2 = MediawikiApi::newFromPage( $wikiPageUrl );
+		$this->assertInstanceOf( MediawikiApi::class, $api2 );
+	}
+
 	/**
 	 * @covers Mediawiki\Api\MediawikiApi::getRequest
 	 * @covers Mediawiki\Api\MediawikiApi::getClientRequestOptions

tests/Integration/TestEnvironment.php

@@ -4,6 +4,7 @@
 
 use Exception;
 use Mediawiki\Api\MediawikiApi;
+use Mediawiki\Api\SimpleRequest;
 
 /**
  * @author Addshore
@@ -68,4 +69,19 @@ public function getApi() {
 		return $this->api;
 	}
 
+	/**
+	 * Save a wiki page.
+	 * @param string $title
+	 * @param string $content
+	 */
+	public function savePage( $title, $content ) {
+
+		$params = [
+			'title' => $title,
+			'text' => $content,
+			'md5' => md5( $content ),
+			'token' => $this->api->getToken(),
+		];
+		$this->api->postRequest( new SimpleRequest( 'edit', $params ) );
+	}
 }