Allow for CRC23 checksum gzuncompress() (#622)

* Allow for CRC23 checksum gzuncompress()

A zlib compressed stream may have a CRC32 checksum instead of Adler-32 which the PHP gzuncompress() function expects. Add a second zlib decompression attempt if the first one fails. See: https://www.php.net/manual/en/function.gzuncompress.php#79042
Partially resolves #592.

* Simplify decodeFilterFlateDecode() error-handling

Instead of setting an error handler to catch the E_WARNING's that gzuncompress() emits, suppress it with an @ so we can do away with the try/catch. Make a note of this in the comments.
Switch from using tempnam() to tmpfile() because tempnam() can emit E_NOTICE's and would have to be suppressed as well. tmpfile() just returns a handle or false.
Limit file_get_contents() by the $decodeMemoryLimit. Unlike gzuncompress() for which a limit value of zero (0) means "no limit", file_get_contents() takes null to mean "no limit".

* Update FilterHelper.php

Fix for PHP < 8.0 that doesn't like a length limit of null for file_get_contents().

src/Smalot/PdfParser/RawData/FilterHelper.php

@@ -233,32 +233,32 @@ protected function decodeFilterASCII85Decode(string $data): string
      */
     protected function decodeFilterFlateDecode(string $data, int $decodeMemoryLimit): ?string
     {
-        /*
-         * gzuncompress may throw a not catchable E_WARNING in case of an error (like $data is empty)
-         * the following set_error_handler changes an E_WARNING to an E_ERROR, which is catchable.
-         */
-        set_error_handler(function ($errNo, $errStr) {
-            if (\E_WARNING === $errNo) {
-                throw new \Exception($errStr);
-            } else {
-                // fallback to default php error handler
-                return false;
-            }
-        });
+        // Uncatchable E_WARNING for "data error" is @ suppressed
+        // so execution may proceed with an alternate decompression
+        // method.
+        $decoded = @gzuncompress($data, $decodeMemoryLimit);
 
-        $decoded = null;
-
-        // initialize string to return
-        try {
-            $decoded = gzuncompress($data, $decodeMemoryLimit);
-            if (false === $decoded) {
-                throw new \Exception('decodeFilterFlateDecode: invalid code');
+        if (false === $decoded) {
+            // If gzuncompress() failed, try again using the compress.zlib://
+            // wrapper to decode it in a file-based context.
+            // See: https://www.php.net/manual/en/function.gzuncompress.php#79042
+            // Issue: https://github.com/smalot/pdfparser/issues/592
+            $ztmp = tmpfile();
+            if (false != $ztmp) {
+                fwrite($ztmp, "\x1f\x8b\x08\x00\x00\x00\x00\x00".$data);
+                $file = stream_get_meta_data($ztmp)['uri'];
+                if (0 === $decodeMemoryLimit) {
+                    $decoded = file_get_contents('compress.zlib://'.$file);
+                } else {
+                    $decoded = file_get_contents('compress.zlib://'.$file, false, null, 0, $decodeMemoryLimit);
+                }
+                fclose($ztmp);
             }
-        } catch (\Exception $e) {
-            throw $e;
-        } finally {
-            // Restore old handler just in case it was customized outside of PDFParser.
-            restore_error_handler();
+        }
+
+        if (false === \is_string($decoded) || '' === $decoded) {
+            // If the decoded string is empty, that means decoding failed.
+            throw new \Exception('decodeFilterFlateDecode: invalid data');
         }
 
         return $decoded;

tests/PHPUnit/Integration/RawData/FilterHelperTest.php

@@ -36,6 +36,7 @@
 namespace PHPUnitTests\Integration\RawData;
 
 use PHPUnitTests\TestCase;
+use Smalot\PdfParser\Parser;
 use Smalot\PdfParser\RawData\FilterHelper;
 
 class FilterHelperTest extends TestCase
@@ -113,7 +114,7 @@ public function testDecodeFilterFlateDecode(): void
     public function testDecodeFilterFlateDecodeEmptyString(): void
     {
         $this->expectException(\Exception::class);
-        $this->expectExceptionMessage('gzuncompress(): data error');
+        $this->expectExceptionMessage('decodeFilterFlateDecode: invalid data');
 
         $this->fixture->decodeFilter('FlateDecode', '');
     }
@@ -124,13 +125,24 @@ public function testDecodeFilterFlateDecodeEmptyString(): void
     public function testDecodeFilterFlateDecodeUncompressedString(): void
     {
         $this->expectException(\Exception::class);
-        $this->expectExceptionMessage('gzuncompress(): data error');
+        $this->expectExceptionMessage('decodeFilterFlateDecode: invalid data');
 
         $this->fixture->decodeFilter('FlateDecode', 'something');
     }
 
     /**
-     * How does function behave if an uncompressed string was given.
+     * How does function behave if compression checksum is CRC32 instead of Adler-32.
+     * See: https://github.com/smalot/pdfparser/issues/592
+     */
+    public function testDecodeFilterFlateDecodeCRC32Checksum(): void
+    {
+        $document = (new Parser())->parseFile($this->rootDir.'/samples/bugs/Issue592.pdf');
+
+        self::assertStringContainsString('Two Westbrook Corporate Center Suite 500', $document->getText());
+    }
+
+    /**
+     * How does function behave if an unknown filter name was given.
      */
     public function testDecodeFilterUnknownFilter(): void
     {