ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent-MP/Changes
(Generate patch)

Comparing AnyEvent-MP/Changes (file contents):
Revision 1.60 by root, Tue Sep 8 13:46:25 2009 UTC vs.
Revision 1.113 by root, Fri May 7 18:14:21 2010 UTC

1Revision history for AnyEvent::MP 1Revision history for AnyEvent::MP
2 2
3TODO: global protocol fix?
4TODO: use seednodes only to broadcast?
5TODO: testsuite 3TODO: testsuite
6TODO: intro: maybe simple job pool example? 4TODO: intro: maybe simple job pool example?
7TODO: intro: mention watchdog... 5TODO: intro: mention watchdog...
8TODO: group monitoring
9 6
10TODO: cal, cal_func 7faq: can't se anything
11TODO: mon_nodes should init-call for all up nodes 8faq: all is asynch
12TODO: better autoseeding, timout*(nconnect+1) 9faq: how to interface to non-perl nodes?
13TODO: fix the "trying to contatc node under wrogn address" problem. 10
142009-09-08 03:58:42 <9> trying connect to seed node 10.0.0.18:4040. 11TODO: maybe disbale warnings by default?
152009-09-08 03:58:42 <9> 10.0.0.18:4040 frank already connected, not connecting again. 12TODO: lese sadrak mails
13
141.29 Fri May 7 20:13:39 CEST 2010
15 - codename "Sadrak".
16 - fix error in callback at AnyEvent/MP/Global.pm line 339
17 (found by Sadrak).
18 - listener-less nodes had trouble sending keepalive
19 messages on write timeouts (found by Sadrak).
20 - the monitor guard could cause a memleak due
21 to autovivification if a mon was cleared after
22 the port was gone (analysed by Sadrak).
23 - do not overwrite the config file if we couldn't read it
24 for some reason (lesson demonstrated by Sadrak).
25
261.28 Thu Apr 1 21:23:54 CEST 2010
27 - accepted connections didn't correctly set up the SRCNODE
28 in some cases, leading to intra-node messages (such as monitoring)
29 to fail silently (great testcase by Sadrak).
30
311.27 Sat Mar 20 21:11:16 CET 2010
32 - avoid exceptions in the transport code if the connection
33 fails during TLS negotiation.
34 - delay new address broadcasts in seednodes, to reduce the
35 likelyhood of nodes connecting at the same time, leading to
36 message loss.
37 - improve general message passing efficiency by 8%, and special
38 case json framing for a 50% speed increase.
39 - add hints how to interface to aemp nodes from other programming
40 languages using the aemp protocol in AnyEvent::MP::Transport
41 (it can really be quite simple).
42 - handle unsupported remote framings more gracefully.
43
441.26 Wed Dec 30 16:48:37 CET 2009
45 - INCOMPATIBLE CHANGE: the high-level node protocol changed,
46 all nodes need to upgrade.
47 - remote kil's did not work (reported by Eugene Grigoriev).
48 - documented monitor_timeout, connect_interval, auth_offer,
49 auth_accept, framing_format, autocork and nodelay options
50 and make them configurable via bin/aemp.
51 - some doc updates.
52 - remote kil's were not synchronous to message sends which affected
53 nobody, as remote kils were broken.
54 - remote mon/unmonitor requests were not synchronous to message
55 sends either sometimes, although it likely didnt matter.
56
571.24 Wed Dec 9 14:59:52 CET 2009
58 - codename "the foul egg".
59 - increased "unmonitored local port died" message level to 2, so
60 it is displayed by default now.
61 - generate random seed at configure time, not module loading time,
62 so fork&configure works as expected.
63 - decreased "seed"-related message priority in Global from 0 (d'oh)
64 to 9, as they are not signifcant.
65 - new experimental module AnyEvent::MP::DataConn to establish
66 additional data connections between nodes.
67 - bin/aemp now loads AnyEvent::Watchdog::Util before trying
68 to call restart.
69 - bin/aemp setcert didn't properly handle the filename argument.
70 - removed empty and unused timeout= parameter from protocol greeting.
71
721.23 Fri Nov 6 18:46:26 CET 2009
73 - listener-less nodes were misinformed by their masters about
74 the location of new nodes, and therefore didn't connect
75 properly in all cases.
76 - messages send during node-up processing could sometimes get lost,
77 which would hamper global's ability to mesh the network.
78 - fixed AnyEvent::MP::Global::grp_reg to return a guard also
79 in scalar context.
80 - fixed AnyEvent::MP::Kernel::mon_nodes to return a guard also
81 in scalar context.
82 - try to improve error reporting when automatically loading
83 a function - being unable to load a module will now
84 stop the process and report the error.
85
861.22 Sat Oct 17 03:41:47 CEST 2009
87 - fix two rcv-bugs: after adding a tagged rcv, the default
88 callback would not be used or changed anymore (reported
89 by Felix Antonius Wilhelm Ostmann, with apologies for
90 taking so long to releae the fix).
91 - added Coro::MP introduction to AnyEvent::MP::Intro.
92 - add peval function.
93
941.21 Mon Oct 5 21:58:36 CEST 2009
95 - always delay kil messages.
96 - remote monitor requests were never cleaned up (found by
97 Sadrak and analysed by elmex).
98 - move $VERSION to AnyEvent::MP and AnyEvent::MP only.
99
1001.2 Thu Oct 1 22:59:35 CEST 2009
101 - fix a rather big bug where message processing could be
102 inhibited completely.
103 - grp_mon could not return a guard due to a typo.
104 - implement primitive aemp shell.
105 - do not use bound address as listen length (reported by
106 Felix Antonius Wilhelm Ostman).
107 - no longer export an undefined "reg" function from AE::MP.
108 - the mon_guard return value no longer keeps an additional
109 reference to the passed refs.
110
1111.1 Fri Sep 11 04:34:03 CEST 2009
112 - bumped the transport protocol to version 1 - sorry, but there
113 were too many bugfixes/changes.
114 - new function AnyEvent::MP::cal for simple rpc.
16 - renamed AnyEvent::MP::Global functions to grp_reg|get|mon. 115 - renamed AnyEvent::MP::Global functions to grp_reg|get|mon.
116 - implemented monitoring for Global groups.
17 - removed all userspace time-outs from the transport, instead 117 - removed all userspace time-outs from the transport, instead
18 rely on tcp retransmit timeouts and tcp keepalive. 118 rely on tcp retransmit timeouts and tcp keepalive.
19 - spawn now delays spawns on the local node artificially, so 119 - spawn now delays spawns on the local node artificially, so
20 it can return before invoking the init function. 120 it can return before invoking the init function.
21 - break endless recursion between ping-pong ports on the local 121 - break endless recursion between ping-pong ports on the local
… …
28 - new service: AnyEvent::MP::LogCatcher. 128 - new service: AnyEvent::MP::LogCatcher.
29 - rely on perl srand'ing better than we could. 129 - rely on perl srand'ing better than we could.
30 - try to backport to 5.8.2, unfortunately, recursive dependencies 130 - try to backport to 5.8.2, unfortunately, recursive dependencies
31 rely on Module::Build, which fails at 5.8.7, so YMMV. 131 rely on Module::Build, which fails at 5.8.7, so YMMV.
32 - use a vastly different reseeding strategy. 132 - use a vastly different reseeding strategy.
133 - use seed nodes to notify nodes about new nodes.
134 - in the case where a ndoe learns new addresses after it tried
135 connecting, prefer the new addrseses over the old ones.
33 136
341.0 Wed Sep 2 20:40:43 CEST 2009 1371.0 Wed Sep 2 20:40:43 CEST 2009
35 - randomise more connection times, but reduce delay to 0.2s (seeds) 138 - randomise more connection times, but reduce delay to 0.2s (seeds)
36 and 0.05s (normal connects). 139 and 0.05s (normal connects).
37 - upgrade to and require AnyEvent::Watchdog 1.0. 140 - upgrade to and require AnyEvent::Watchdog 1.0.

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines