tumblr_backup: Stop if API responses stop making forward progress

Sometimes, at least when backing up likes, the API can get stuck
endlessly returning the same set of posts instead of returning an empty
list. Inspect _links and stop if the offset/before fails to change.

Fixes bbolli#217

tumblr_backup.py

@@ -585,6 +585,7 @@ def _backup(posts):
             # Get the JSON entries from the API, which we can only do for MAX_POSTS posts at once.
             # Posts "arrive" in reverse chronological order. Post #0 is the most recent one.
             i = options.skip
+            last_next_offset = None
             while True:
                 # find the upper bound
                 log(account, "Getting posts %d to %d (of %d expected)\r" % (i, i + MAX_POSTS - 1, count_estimate))
@@ -596,9 +597,20 @@ def _backup(posts):
                     continue
 
                 posts = _get_content(soup)
-                # `_backup(posts)` can be empty even when `posts` is not if we don't backup reblogged posts
-                if not posts or not _backup(posts):
-                    log(account, "Backing up posts found empty set of posts, finishing\r")
+                if not posts:
+                    log(account, "Found empty set of posts, finishing\r")
+                    break
+
+                next_params = soup['response']['_links']['next']['query_params']
+                next_offset = next_params.get('offset') or next_params.get('before')
+                if next_offset is not None:
+                    if next_offset == last_next_offset:
+                        log(account, "Found same API response twice, finishing\r")
+                        break
+                    last_next_offset = next_offset
+
+                if not _backup(posts):
+                    log(account, "Found last requested post, finishing\r")
                     break
 
                 i += MAX_POSTS