11# unreleased
22
3+ New features since 1.8.5
4+
5+ Bug fixes
6+
7+
8+ # 1.8.5
9+
310New features since 1.8.4
411
512* The 'connect_timeout' parameter now applies to the whole process of
@@ -9,20 +16,22 @@ New features since 1.8.4
916 This is useful because sometimes socket.connect() may succeed even
1017 though the server is in fact hanging.
1118
12- * All logging messages are now at DEBUG level. There used to be one
13- INFO message "Established connection to..." but it was emitted too
14- early, before login. Logging everything as DEBUG seems more consistent.
19+ * All logging messages are now at DEBUG level. There used to be one INFO
20+ message "Established connection to..." but it was emitted too early,
21+ before login. Logging everything as DEBUG seems more consistent.
1522
1623Bug fixes
1724
18- * The decision whether a result set needs to be closed or not was accidentally
19- based on the size of the previous result set, not the current. This could
20- cause unclosed result sets to pile up until the connection was closed.
25+ * The decision whether a result set needs to be closed or not was
26+ accidentally based on the size of the previous result set, not the
27+ current. This could cause unclosed result sets to pile up until the
28+ connection was closed.
2129
22- * When scanning Unix Domain sockets, only OSErrors and 'no such database' errors
23- are now suppressed. The others may be interesting and are passed through so
24- they are no longer masked by subsequent uninteresting errors such as
25- 'connection refused'.
30+ * When scanning Unix Domain sockets, errors are to be expected. Until
31+ now, all exceptions were intercepted and if no connection could be
32+ made, the last one was rethrown. This caused important errors such as
33+ 'invalid credentials' to be masked by later less interesting errors.
34+ This has been fixed by only posponing OSErrors and 'no such database'.
2635
2736
2837# 1.8.4
0 commit comments