2 * Correctly balance stack in callback. Mattew Wild <mwild1@gmail.com>
4 * Fix to allow 0 timeout values. Matthew Wild <mwild1@gmail.com>
5 * Collect traceback if an error occurs in a callback. Kim Alvefur <zash@zash.se>
7 * Updated license to MIT license
8 * Add basic continuous integration tests via Travis-CI
9 * event_callback.c: Don't pop too much. Kim Alvefur <zash@zash.se>
10 * Updates to reduce global pollution
12 * Use busted test framework for buffer tests
13 * Fix crashes due to event being deleted while in callback
14 * Fix 0 timeout handling appropriately (it is a valid value)
16 * Fixes unchecked stack increase resulting in potential overflow
18 * Adds error catching / re-raise
20 * Fixes event_callback handling to handle the no-return-value case better
21 * Cleans up timertest formatting
24 * Fixes buffer event tests to work with current bufferevent behavior
25 * Fixed buffer event and related tests to be compatible with Lua 5.2
26 * Makes core segment compatible with Lua 5.2
27 * Remove use of 'module' in sources
28 * Remove assumption that luaevent will set globals
29 - Only 'core' will modify global table (creating luaevent.core)
30 * Fixes bug in merge for receivePartial
33 * Fixed recorded version in Lua
35 * Merged all of luaevent-prosody (0.1.1) changes (missed updates)
37 * Fixed libevent 2.0 compilation errors
38 * Merged all of luaevent-prosody (0.1.0) changes
39 * Removes low-level read access to watermark & timeout values that break API
40 * Switches to watermark write accessor function to avoid API break
42 luaevent-prosody 0.1.1 - 2010-02-25
43 * Fix overflow with high event timeouts
45 luaevent-prosody 0.1.0 - 2010-02-15
46 * Fixed stack slot leak in event callbacks
47 * Fixed stack slot leak when error occurs in callback
48 * Various compilation fixes for Windows/ANSI C
49 * Silence some compiler warnings
50 + Add base:loopexit() method, with timeout support
51 + Add base:method() to discover backend in use
52 + Add core.libevent_version() to detect libevent version
53 + Add _NAME and _VERSION fields to module table
54 + base:addevent() now accepts integer fd to watch
55 + Switched from premake to standard Makefile for building
58 + Reorganized project to better fit GIT
59 + Refactored and cleaned sources
60 * Broke event_callback into its own source file
61 + Added timer/timeout capabilities + minor test
62 + Added event_buffer object
63 * Can 'add' a sequence of strings/event_buffers
64 * Can 'get_data', 'length','drain','close' and create new instances
65 * Can 'readline', 'read' from file des, 'write' to file des
66 * Added lunit tests for basic functions (read/write from/to FD)
69 + Setup system to use new coro management as described in COROUTINE_MANAGEMENT
70 The callbacks are called from the event_loop 'thread' rather than that which they are
71 created in. This will prevent the self-resume problem as well as dead-thread problems.
72 - Recognized issues to fix in next release:
73 * Socket/event closing needs to be cleaned
74 * luaevent.lua needs refactoring
75 * luaevent.[ch] need to be cleaned up
78 + Fixed event-handling code to cancel events on nothing being returned
79 + Added socket/object cleanup.
80 + Filed bug to libevent about the strange valgrind-released errors
81 - Recognized following issues:
83 Need to handle events setup from inside a coroutine... need to get a global Lua state from a thread...
85 0.1.0 - 2007-06-10 22:00 EST
86 Completed mostly working version
87 * Moved to a mode where addevent calls a callback rather than it being instantiated within.
88 If the callback returns -1, then no event is ever setup,
89 Otherwise the integer value is used to setup the event.
90 This allows for using coroutine.wrap rather than a cooked-up wrapper
91 * Tests work, although there are a few remaining issues:
92 * Need to figure a good way of preserving the event object,
93 not sure if current method is good enough, since the socket
94 is the only anchor, and it is only held inside the coro..
95 circular reference, something that Lua 'handles' well.
96 * Doing more than the maximum sockets the process is allows
97 causes strangeness to occur in libevent.. somehow
98 it is getting around to epoll_add which is causing valgrind
100 * Added cheap protection code for failures in callback handlers
102 0.0.0 - 2007-06-10 12:00 EST
103 Initial public version, was broken due to self-resume coroutines