Don't fail on paths that can't be processed for some reason (#1)

README.md

@@ -26,13 +26,15 @@ Warning
 --------------------
 
 As illustrated in
-[#1](https://github.com/mk-fg/image-deduplication-tool/issues/1) and the "Known
-Issues" section, libpHash and/or CImg (which it's based on) can do quite strange
-and unexpected things when presented with non-images up to the point of
-executing malicious code from the filename in the shell.
+[#1](https://github.com/mk-fg/image-deduplication-tool/issues/1) and
+[CImg#49](https://sourceforge.net/p/cimg/bugs/49/), libpHash/CImg will fall back
+to using "sh -c" commands for non-image file formats and might not get
+filename-escaping correct there (especially with CImg versions up to 1.5.3).
 
 Simple safeguard for that would be only to run the tool on image paths, not
-paths that contain mixed-type files.
+paths that contain mixed-type files, or at least make sure there's no funky
+stuff in the filenames - checking that in the script is wrong, as proper image
+files work with any filename.
 
 One other precaution is that with --feh option, script will run "feh" program,
 and --feh-args parameter may contain options (e.g. --info) that will be executed

image_matcher.py

@@ -4,7 +4,7 @@
 
 
 import itertools as it, operator as op, functools as ft
-import os, sys, ctypes, pickle, select, signal, struct
+import os, sys, ctypes, errno, pickle, select, signal, struct
 
 
 class pHash(object):
@@ -16,7 +16,8 @@ def dct_imagehash_async(self, path):
 		r, w = os.pipe()
 		pid = os.fork()
 		if not pid:
-			pickle.dump(self.dct_imagehash(path), os.fdopen(w, 'wb'))
+			with os.fdopen(w, 'wb') as dst:
+				pickle.dump(self.dct_imagehash(path), dst)
 			os._exit(0) # so "finally" clauses won't get triggered
 		else:
 			os.close(w)
@@ -26,7 +27,10 @@ def dct_imagehash(self, path):
 		phash = ctypes.c_uint64()
 		if self._lib.ph_dct_imagehash(path, ctypes.pointer(phash)):
 			errno_ = ctypes.get_errno()
-			raise OSError(errno_, os.strerror(errno_))
+			print( 'Failed to get image hash ({}): {}'\
+					.format(errno.errorcode[errno_], os.strerror(errno_)),
+				file=sys.stderr )
+			return None
 		return phash.value
 
 	def hamming_distance(self, hash1, hash2):
@@ -71,9 +75,9 @@ def sort_by_similarity(dcts):
 	log.debug('Calculating/sorting Hamming distances')
 	for img1, img2 in it.combinations(dcts.viewitems(), 2):
 		for path, h in img1, img2:
-			if h == 0:
+			if h == 0 or h is None:
 				if path not in paths_skipped:
-					log.debug('Skipping 0-hash path: {}'.format(path))
+					log.debug('Skipping no-hash path: {}'.format(path))
 					paths_skipped.add(path)
 				break
 		else: