Found the reason why the naive search sometimes fails and fixed it.

Now the (in a zipfile potentially very slow) "proper" search should really never be necessary, but keep it, just in case.
Maybe there will be even more of those alterations of the real name.

Author: horstle

lib/inputstreamhelper/widevine/arm_chromeos.py

@@ -73,34 +73,54 @@ def chromeos_offset(self):
 
         return offset
 
-    def _find_file_naive(self, filename):
+    def _find_file_in_chunk(self, bfname, chunk):
+        """
+        Checks if the filename is found in the given chunk.
+        Then makes some plausibility checks, to see if it is in fact a proper dentry.
+
+        returns the dentry if found, else False.
+        """
+
+        if bfname in chunk:
+            i_index_pos = chunk.index(bfname) - 8  # the filename is the last element of the dentry, the elements before are 8 bytes total
+            file_entry = self.dir_entry(chunk[i_index_pos:i_index_pos + len(bfname) + 8])  # 8 because see above
+            if file_entry['inode'] < self.sb_dict['s_inodes_count'] and file_entry['name_len'] == len(bfname):
+                return file_entry
+
+            log(0, 'Found filename, but checks did not pass:')
+            log(0, 'inode number: {inode} < {count}, name_len: {name_len} == {len_fname}'.format(inode=file_entry['inode'],
+                                                                                                 count=self.sb_dict['s_inodes_count'],
+                                                                                                 name_len=file_entry['name_len'],
+                                                                                                 len_fname=len(bfname)))
+        return False
+
+    def _find_file_naive(self, fname):
         """
         Finds a file by basically searching for the filename as bytes in the bytestream.
         Searches through the whole image only once, making it fast, but may be unreliable at times.
 
         Returns a directory entry.
         """
 
-        bin_filename = filename.encode('ascii')
+        fname_alt = fname + '#new'  # Sometimes the filename has "#new" at the end
+        bfname = fname.encode('ascii')
+        bfname_alt = fname_alt.encode('ascii')
         chunksize = 4 * 1024**2
         chunk1 = self.read_stream(chunksize)
         while True:
             chunk2 = self.read_stream(chunksize)
             if not chunk2:
-                raise ChromeOSError('File {filename} not found in the ChromeOS image'.format(filename=filename))
+                raise ChromeOSError('File {fname} not found in the ChromeOS image'.format(fname=fname))
 
             chunk = chunk1 + chunk2
-            if bin_filename in chunk:
-                i_index_pos = chunk.index(bin_filename) - 8
-                file_entry = self.dir_entry(chunk[i_index_pos:i_index_pos + len(filename) + 8])
-                if file_entry['inode'] < self.sb_dict['s_inodes_count'] and file_entry['name_len'] == len(filename):
-                    break
-
-                log(0, 'Found filename, but checks did not pass:')
-                log(0, 'inode number: {inode} < {count}, name_len: {name_len} == {len_filename}'.format(inode=file_entry['inode'],
-                                                                                                       count=self.sb_dict['s_inodes_count'],
-                                                                                                       name_len=file_entry['name_len'],
-                                                                                                       len_filename=len(filename)))
+
+            file_entry = self._find_file_in_chunk(bfname, chunk)
+            if file_entry:
+                break
+
+            file_entry = self._find_file_in_chunk(bfname_alt, chunk)
+            if file_entry:
+                break
 
             chunk1 = chunk2