diff --git a/src/ConfluencePageContentDownloader.php b/src/ConfluencePageContentDownloader.php
index 6abe720..5911771 100755
--- a/src/ConfluencePageContentDownloader.php
+++ b/src/ConfluencePageContentDownloader.php
@@ -68,15 +68,33 @@ public function downloadPageContent(ConfluencePage $page, bool $withAttachments
}
}
+ /**
+ * Normalises the page content by letting DOMDocument repair malformed markup and then
+ * stripping the document wrapper that loadHTML() adds, keeping only the body inner HTML.
+ *
+ * We deliberately do not call DOMDocument::validate(): it would validate the content
+ * against the HTML 4.0 Transitional DTD referenced in the doctype, which forces an
+ * outbound HTTP request to w3.org on every page. In locked-down environments without
+ * outbound internet access that request fails and PHP emits a warning. Validation adds
+ * no value here anyway (Confluence storage format is never valid HTML 4.0), so removing
+ * it also drops an unnecessary network dependency.
+ */
private function repairPageContent(ConfluencePage $page): ConfluencePage
{
+ $content = $page->getContent();
+ if ($content === '') {
+ return $page;
+ }
+
$previousLibxmlState = libxml_use_internal_errors(true);
$domDocument = new DOMDocument();
- $domDocument->loadHTML($page->getContent());
- if (!$domDocument->validate()) {
+ $domDocument->loadHTML($content, LIBXML_NONET);
+
+ $body = $domDocument->getElementsByTagName('body')->item(0);
+ if ($body !== null) {
$pageContent = '';
- foreach ($domDocument->getElementsByTagName('body')->item(0)->childNodes ?? [] as $child) {
+ foreach ($body->childNodes as $child) {
$pageContent .= $domDocument->saveHTML($child);
}
diff --git a/tests/ConfluencePageContentDownloaderTest.php b/tests/ConfluencePageContentDownloaderTest.php
index 95f669b..2e4d045 100644
--- a/tests/ConfluencePageContentDownloaderTest.php
+++ b/tests/ConfluencePageContentDownloaderTest.php
@@ -43,6 +43,39 @@ public function testAppliesReplacersAndDownloadsContentAndAttachments(): void
$downloader->downloadPageContent($this->page('123'));
}
+ /**
+ * Regression test for #33547: page content arrives wrapped in a full HTML document whose
+ * doctype references the HTML 4.0 loose DTD. repairPageContent() must strip the wrapper down
+ * to the body inner HTML without ever validating against (and thus fetching) that external
+ * DTD, which previously triggered a "Failed to open stream: HTTP request failed!" warning in
+ * environments without outbound internet access.
+ */
+ public function testStripsHtmlDocumentWrapperWithoutFetchingExternalDtd(): void
+ {
+ $wrapped = ''
+ . 'xoriginal
';
+
+ $savedContent = null;
+ $download = $this->createMock(Download::class);
+ $download->expects($this->once())->method('downloadPageContent')
+ ->with($this->callback(static function (ConfluencePage $page) use (&$savedContent): bool {
+ $savedContent = $page->getContent();
+
+ return true;
+ }), 'content.html');
+
+ $logger = $this->createMock(LoggerInterface::class);
+ $logger->expects($this->never())->method('error');
+
+ $page = new ConfluencePage(['id' => '123', 'body' => ['storage' => ['value' => $wrapped]]]);
+
+ $downloader = new ConfluencePageContentDownloader(self::createStub(Content::class), $download, [], $logger);
+ $downloader->downloadPageContent($page, false);
+
+ self::assertSame('original
', $savedContent);
+ }
+
public function testDoesNotDownloadAttachmentsWhenDisabled(): void
{
$content = $this->createMock(Content::class);