Merge pull request #6420 from matrix-org/erikj/fix_find_next_generated_user_id_localpart

Fix guest registration

changelog.d/6420.bugfix

@@ -0,0 +1 @@
+Fix broken guest registration when there are existing blocks of numeric user IDs.

synapse/storage/data_stores/main/registration.py

@@ -19,7 +19,6 @@ import logging
 import re
 
 from six import iterkeys
-from six.moves import range
 
 from twisted.internet import defer
 from twisted.internet.defer import Deferred
@@ -482,12 +481,8 @@ class RegistrationWorkerStore(SQLBaseStore):
         """
         Gets the localpart of the next generated user ID.
 
-        Generated user IDs are integers, and we aim for them to be as small as
-        we can. Unfortunately, it's possible some of them are already taken by
-        existing users, and there may be gaps in the already taken range. This
-        function returns the start of the first allocatable gap. This is to
-        avoid the case of ID 1000 being pre-allocated and starting at 1001 while
-        0-999 are available.
+        Generated user IDs are integers, so we find the largest integer user ID
+        already taken and return that plus one.
         """
 
         def _find_next_generated_user_id(txn):
@@ -497,15 +492,14 @@ class RegistrationWorkerStore(SQLBaseStore):
 
             regex = re.compile(r"^@(\d+):")
 
-            found = set()
+            max_found = 0
 
             for (user_id,) in txn:
                 match = regex.search(user_id)
                 if match:
-                    found.add(int(match.group(1)))
-            for i in range(len(found) + 1):
-                if i not in found:
-                    return i
+                    max_found = max(int(match.group(1)), max_found)
+
+            return max_found + 1
 
         return (
             (