diff --git a/CHANGELOG.md b/CHANGELOG.md index f0f2e3d..1f97d6a 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -14,6 +14,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/). ### Fixed - Fix PDF export crash on tickets with tables containing nested-quote styles +- Fix base64 images (e.g. in ticket tasks) corrupted or missing in PDF export ## [4.2.0] - 2026-10-06 diff --git a/inc/simplepdf.class.php b/inc/simplepdf.class.php index e68d757..4f0ab4e 100644 --- a/inc/simplepdf.class.php +++ b/inc/simplepdf.class.php @@ -458,75 +458,32 @@ public function displayText($name, $content = '', $minline = 3, $maxline = 100) $content = html_entity_decode($content, ENT_QUOTES, 'UTF-8'); $content = RichText::getSafeHtml($content); - preg_match_all( - '/]*>\s*(]*src=["\']([^"\']*docid=(\d+)[^"\']*)["\'][^>]*>)\s*<\/a>/i', + // The sanitizer encodes some characters in attributes (e.g. "=" => "=", "+" => "+"), which TCPDF + // does not decode: document URLs are no longer recognized and base64 data URIs get corrupted. + // Images are therefore resolved here from their decoded "src" and passed to TCPDF as validated data streams. + // Resolved images are replaced by placeholders until the end of the processing: their (large) base64 data + // would otherwise make the table regex hit "pcre.backtrack_limit" and the whole content would be lost. + $images = []; + $placeholder_token = 'pdfimg' . bin2hex(random_bytes(8)) . '_'; + $content = preg_replace_callback( + '/(]*>\s*)?(]*>)(\s*<\/a>)?/i', + function (array $match) use (&$images, $placeholder_token): string { + $img_tag = $this->buildPdfImageTag($match[2]); + if ($img_tag === null) { + return $match[0]; + } + $placeholder = $placeholder_token . count($images); + $images[$placeholder] = $img_tag; + if ($match[1] !== '' && ($match[3] ?? '') !== '') { + // Remove the link wrapping the image + return $placeholder; + } + + return $match[1] . $placeholder . ($match[3] ?? ''); + }, $content, - $matches, - PREG_SET_ORDER, ); - if ($matches !== []) { - $document = new Document(); - foreach ($matches as $match) { - $full_a_tag = $match[0]; - $full_img_tag = $match[1]; - $docid = (int) $match[3]; - - if ( - $document->getFromDB($docid) - && $document->canViewFile() - && isset($document->fields['filepath']) - ) { - $file_path = GLPI_DOC_DIR . '/' . $document->fields['filepath']; - if (file_exists($file_path)) { - // Get actual image dimensions from file - $img_info_actual = @getimagesize($file_path); - $actual_width = 450; // Default - $actual_height = 300; // Default - - if ($img_info_actual !== false) { - $actual_width = $img_info_actual[0]; - $actual_height = $img_info_actual[1]; - } - - // Extract width and height from the HTML img tag - $html_width = null; - $html_height = null; - - if (preg_match('/width=["\']?(\d+)["\']?/i', $full_img_tag, $w_match)) { - $html_width = (int) $w_match[1]; - } - if (preg_match('/height=["\']?(\d+)["\']?/i', $full_img_tag, $h_match)) { - $html_height = (int) $h_match[1]; - } - - // Calculate display dimensions preserving aspect ratio - $display_width = $html_width ?? $actual_width; - $display_height = $html_height ?? $actual_height; - - // If only one dimension is specified, calculate the other to preserve ratio - if ($html_width && !$html_height) { - $display_height = (int) ($html_width * $actual_height / $actual_width); - } elseif ($html_height && !$html_width) { - $display_width = (int) ($html_height * $actual_width / $actual_height); - } - - // Apply maximum width constraint (200 pixels max) - $max_width = 200; - if ($display_width > $max_width) { - $ratio = $max_width / $display_width; - $display_width = $max_width; - $display_height = (int) ($display_height * $ratio); - } - - // Replace the original image by a properly resized one + remove the a tag - $display_image_tag = ''; - $content = str_replace($full_a_tag, $display_image_tag, $content); - } - } - } - } - // Split content by tables, keeping tables in the result $segments = preg_split('/(]*>.*?<\/table>)/is', $content, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY); @@ -546,12 +503,99 @@ public function displayText($name, $content = '', $minline = 3, $maxline = 100) } } - $this->displayInternal(240, 0.5, self::LEFT, $minline * 5, [$formatted_content]); + // Put back the resolved images + $formatted_content = strtr($formatted_content, $images); + + // TCPDF also matches tables with a lazy regex and silently drops the whole content when it fails: + // raise "pcre.backtrack_limit" to fit the content size (large embedded images) while rendering it. + $backtrack_limit = ini_get('pcre.backtrack_limit'); + ini_set('pcre.backtrack_limit', (string) max((int) $backtrack_limit, 2 * strlen($formatted_content))); + try { + $this->displayInternal(240, 0.5, self::LEFT, $minline * 5, [$formatted_content]); + } finally { + ini_set('pcre.backtrack_limit', $backtrack_limit); + } /* Restore */ [$this->cols, $this->colsx, $this->colsw, $this->align, ] = $save; } + /** + * Build an tag usable by TCPDF from a sanitized tag. + * + * Supported sources are GLPI documents (document.send.php?docid=X) and base64 data URIs. + * Image data is validated with GD before being embedded, to prevent TCPDF from failing on corrupted images. + * + * @param string $img_tag sanitized tag + * + * @return string|null new tag, empty string if the image must be dropped, null if the tag must be kept as is + */ + private function buildPdfImageTag(string $img_tag): ?string + { + if (!preg_match('/\bsrc\s*=\s*(["\'])(.*?)\1/is', $img_tag, $src_match)) { + return null; + } + $src = html_entity_decode($src_match[2], ENT_QUOTES | ENT_HTML5, 'UTF-8'); + + $data = null; + if (preg_match('/^data:image\/[a-z0-9.+-]+;base64,(.*)$/is', $src, $data_match)) { + $data = base64_decode(preg_replace('/\s+/', '', $data_match[1]), true); + } elseif (str_contains(parse_url($src, PHP_URL_PATH) ?? '', 'document.send.php')) { + parse_str(parse_url($src, PHP_URL_QUERY) ?? '', $query); + $document = new Document(); + if ( + isset($query['docid']) + && $document->getFromDB((int) $query['docid']) + && $document->canViewFile() + && !empty($document->fields['filepath']) + ) { + $file_path = GLPI_DOC_DIR . '/' . $document->fields['filepath']; + $data = is_readable($file_path) ? file_get_contents($file_path) : false; + } else { + $data = false; + } + } else { + return null; + } + + // Drop unreadable or corrupted images instead of letting TCPDF fail + if ( + !is_string($data) + || ($img_info = @getimagesizefromstring($data)) === false + || @imagecreatefromstring($data) === false + ) { + return ''; + } + $actual_width = $img_info[0]; + $actual_height = $img_info[1]; + + // Extract width and height from the HTML img tag + $html_width = preg_match('/\bwidth=["\']?(\d+)/i', $img_tag, $w_match) ? (int) $w_match[1] : null; + $html_height = preg_match('/\bheight=["\']?(\d+)/i', $img_tag, $h_match) ? (int) $h_match[1] : null; + + // Calculate display dimensions preserving aspect ratio + $display_width = $html_width ?: $actual_width; + $display_height = $html_height ?: $actual_height; + + // If only one dimension is specified, calculate the other to preserve ratio + if ($html_width && !$html_height) { + $display_height = (int) ($html_width * $actual_height / $actual_width); + } elseif ($html_height && !$html_width) { + $display_width = (int) ($html_height * $actual_width / $actual_height); + } + + // Apply maximum width constraint (200 pixels max) + $max_width = 200; + if ($display_width > $max_width) { + $ratio = $max_width / $display_width; + $display_width = $max_width; + $display_height = (int) ($display_height * $ratio); + } + + // "@" prefix: TCPDF data stream, no local file or remote URL access needed + return ''; + } + /** * Display space between row * diff --git a/tests/SimplePDFTest.php b/tests/SimplePDFTest.php index 92ef792..e5c3bd0 100644 --- a/tests/SimplePDFTest.php +++ b/tests/SimplePDFTest.php @@ -31,9 +31,9 @@ * -------------------------------------------------------------------------- */ -use Glpi\Tests\GLPITestCase; +use Glpi\Tests\DbTestCase; -class SimplePDFTest extends GLPITestCase +class SimplePDFTest extends DbTestCase { private function getStringWidth(PluginPdfSimplePDF $pdf, string $string): float { @@ -131,4 +131,89 @@ public function testCleanTableHtmlWithNestedQuoteUrlStyle(): void $this->assertStringNotContainsStringIgnoringCase('height: 15.75pt', $cleaned); $this->assertStringContainsString("url('https://example.com/pics/a.jpg')", $cleaned); } + + /** + * Test a base64 image in a rich text (e.g. a ticket task) + * `+` and `=` were encoded as `+` and `=` by the sanitizer, so the image data was corrupted and TCPDF crashed + */ + public function testDisplayTextEmbedsBase64Image(): void + { + // 1x1 semi-transparent PNG, its base64 contains both `+` and `=` + $base64 = 'iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAACXBIWXMAAA7EAAAOxAGVKw4bAAAADUlEQVQImWP4z8BQBwAEfwF+fk6erwAAAABJRU5ErkJggg=='; + + $pdf = new PluginPdfSimplePDF(); + $pdf->newPage(); + $pdf->displayText('Description', '

before after

'); + + $this->assertStringContainsString('/Subtype /Image', $pdf->output()); + } + + /** + * Test a corrupted base64 image in a rich text + */ + public function testDisplayTextDropsCorruptedBase64Image(): void + { + $base64 = base64_encode($this->getCorruptedPngContent()); + + $pdf = new PluginPdfSimplePDF(); + $pdf->newPage(); + $pdf->displayText('Description', '

before after

'); + + $output = $pdf->output(); + $this->assertStringStartsWith('%PDF-', $output); + $this->assertStringNotContainsString('/Subtype /Image', $output); + } + + /** + * Test a large document image inside a table + * Its base64 data made the table regex hit `pcre.backtrack_limit`, and the whole content vanished from the PDF + */ + public function testDisplayTextEmbedsLargeDocumentImageInTable(): void + { + $this->login(); + + // ~1 MB PNG: random pixels are not compressible + $image = imagecreatetruecolor(600, 600); + for ($x = 0; $x < 600; $x++) { + for ($y = 0; $y < 600; $y++) { + imagesetpixel($image, $x, $y, random_int(0, 0xFFFFFF)); + } + } + $filename = uniqid('glpitest_', true) . '.png'; + $this->assertTrue(imagepng($image, GLPI_TMP_DIR . '/' . $filename, 0)); + $this->assertGreaterThan(1_000_000, filesize(GLPI_TMP_DIR . '/' . $filename)); + + $document = $this->createItem( + Document::class, + [ + 'filename' => $filename, + 'entities_id' => getItemByTypeName('Entity', '_test_root_entity', true), + '_filename' => [$filename], + ], + ['filename', '_filename'], + ); + + $pdf = new PluginPdfSimplePDF(); + $pdf->newPage(); + $pdf->displayText( + 'Description', + '
' + . '
', + ); + + $this->assertStringContainsString('/Subtype /Image', $pdf->output()); + } + + /** + * Build a PNG file with a valid RGBA header but undecodable image data + */ + private function getCorruptedPngContent(): string + { + $chunk = (static fn(string $type, string $data): string => pack('N', strlen($data)) . $type . $data . pack('N', crc32($type . $data))); + + return "\x89PNG\r\n\x1a\n" + . $chunk('IHDR', pack('NNCCCCC', 2, 2, 8, 6, 0, 0, 0)) + . $chunk('IDAT', 'not zlib compressed data') + . $chunk('IEND', ''); + } }