1### 1.0.0 (unreleased)
2
3**BREAKING CHANGES**:
4
5* Bulk and multi-bulk lengths less than -1 or greater than `LLONG_MAX` are now
6  protocol errors. This is consistent with the RESP specification. On 32-bit
7  platforms, the upper bound is lowered to `SIZE_MAX`.
8
9* Change `redisReply.len` to `size_t`, as it denotes the the size of a string
10
11  User code should compare this to `size_t` values as well.  If it was used to
12  compare to other values, casting might be necessary or can be removed, if
13  casting was applied before.
14
15### 0.x.x (unreleased)
16**BREAKING CHANGES**:
17
18* Change `redisReply.len` to `size_t`, as it denotes the the size of a string
19
20User code should compare this to `size_t` values as well.
21If it was used to compare to other values, casting might be necessary or can be removed, if casting was applied before.
22
23* `redisReplyObjectFunctions.createArray` now takes `size_t` for its length parameter.
24
25### 0.14.0 (2018-09-25)
26
27* Make string2ll static to fix conflict with Redis (Tom Lee [c3188b])
28* Use -dynamiclib instead of -shared for OSX (Ryan Schmidt [a65537])
29* Use string2ll from Redis w/added tests (Michael Grunder [7bef04, 60f622])
30* Makefile - OSX compilation fixes (Ryan Schmidt [881fcb, 0e9af8])
31* Remove redundant NULL checks (Justin Brewer [54acc8, 58e6b8])
32* Fix bulk and multi-bulk length truncation (Justin Brewer [109197])
33* Fix SIGSEGV in OpenBSD by checking for NULL before calling freeaddrinfo (Justin Brewer [546d94])
34* Several POSIX compatibility fixes (Justin Brewer [bbeab8, 49bbaa, d1c1b6])
35* Makefile - Compatibility fixes (Dimitri Vorobiev [3238cf, 12a9d1])
36* Makefile - Fix make install on FreeBSD (Zach Shipko [a2ef2b])
37* Makefile - don't assume $(INSTALL) is cp (Igor Gnatenko [725a96])
38* Separate side-effect causing function from assert and small cleanup (amallia [b46413, 3c3234])
39* Don't send negative values to `__redisAsyncCommand` (Frederik Deweerdt [706129])
40* Fix leak if setsockopt fails (Frederik Deweerdt [e21c9c])
41* Fix libevent leak (zfz [515228])
42* Clean up GCC warning (Ichito Nagata [2ec774])
43* Keep track of errno in `__redisSetErrorFromErrno()` as snprintf may use it (Jin Qing [25cd88])
44* Solaris compilation fix (Donald Whyte [41b07d])
45* Reorder linker arguments when building examples (Tustfarm-heart [06eedd])
46* Keep track of subscriptions in case of rapid subscribe/unsubscribe (Hyungjin Kim [073dc8, be76c5, d46999])
47* libuv use after free fix (Paul Scott [cbb956])
48* Properly close socket fd on reconnect attempt (WSL [64d1ec])
49* Skip valgrind in OSX tests (Jan-Erik Rediger [9deb78])
50* Various updates for Travis testing OSX (Ted Nyman [fa3774, 16a459, bc0ea5])
51* Update libevent (Chris Xin [386802])
52* Change sds.h for building in C++ projects (Ali Volkan ATLI [f5b32e])
53* Use proper format specifier in redisFormatSdsCommandArgv (Paulino Huerta, Jan-Erik Rediger [360a06, 8655a6])
54* Better handling of NULL reply in example code (Jan-Erik Rediger [1b8ed3])
55* Prevent overflow when formatting an error (Jan-Erik Rediger [0335cb])
56* Compatibility fix for strerror_r (Tom Lee [bb1747])
57* Properly detect integer parse/overflow errors (Justin Brewer [93421f])
58* Adds CI for Windows and cygwin fixes (owent, [6c53d6, 6c3e40])
59* Catch a buffer overflow when formatting the error message
60* Import latest upstream sds. This breaks applications that are linked against the old hiredis v0.13
61* Fix warnings, when compiled with -Wshadow
62* Make hiredis compile in Cygwin on Windows, now CI-tested
63* Bulk and multi-bulk lengths less than -1 or greater than `LLONG_MAX` are now
64  protocol errors. This is consistent with the RESP specification. On 32-bit
65  platforms, the upper bound is lowered to `SIZE_MAX`.
66
67* Remove backwards compatibility macro's
68
69This removes the following old function aliases, use the new name now:
70
71| Old                         | New                    |
72| --------------------------- | ---------------------- |
73| redisReplyReaderCreate      | redisReaderCreate      |
74| redisReplyReaderCreate      | redisReaderCreate      |
75| redisReplyReaderFree        | redisReaderFree        |
76| redisReplyReaderFeed        | redisReaderFeed        |
77| redisReplyReaderGetReply    | redisReaderGetReply    |
78| redisReplyReaderSetPrivdata | redisReaderSetPrivdata |
79| redisReplyReaderGetObject   | redisReaderGetObject   |
80| redisReplyReaderGetError    | redisReaderGetError    |
81
82* The `DEBUG` variable in the Makefile was renamed to `DEBUG_FLAGS`
83
84Previously it broke some builds for people that had `DEBUG` set to some arbitrary value,
85due to debugging other software.
86By renaming we avoid unintentional name clashes.
87
88Simply rename `DEBUG` to `DEBUG_FLAGS` in your environment to make it working again.
89
90### 0.13.3 (2015-09-16)
91
92* Revert "Clear `REDIS_CONNECTED` flag when connection is closed".
93* Make tests pass on FreeBSD (Thanks, Giacomo Olgeni)
94
95
96If the `REDIS_CONNECTED` flag is cleared,
97the async onDisconnect callback function will never be called.
98This causes problems as the disconnect is never reported back to the user.
99
100### 0.13.2 (2015-08-25)
101
102* Prevent crash on pending replies in async code (Thanks, @switch-st)
103* Clear `REDIS_CONNECTED` flag when connection is closed (Thanks, Jerry Jacobs)
104* Add MacOS X addapter (Thanks, @dizzus)
105* Add Qt adapter (Thanks, Pietro Cerutti)
106* Add Ivykis adapter (Thanks, Gergely Nagy)
107
108All adapters are provided as is and are only tested where possible.
109
110### 0.13.1 (2015-05-03)
111
112This is a bug fix release.
113The new `reconnect` method introduced new struct members, which clashed with pre-defined names in pre-C99 code.
114Another commit forced C99 compilation just to make it work, but of course this is not desirable for outside projects.
115Other non-C99 code can now use hiredis as usual again.
116Sorry for the inconvenience.
117
118* Fix memory leak in async reply handling (Salvatore Sanfilippo)
119* Rename struct member to avoid name clash with pre-c99 code (Alex Balashov, ncopa)
120
121### 0.13.0 (2015-04-16)
122
123This release adds a minimal Windows compatibility layer.
124The parser, standalone since v0.12.0, can now be compiled on Windows
125(and thus used in other client libraries as well)
126
127* Windows compatibility layer for parser code (tzickel)
128* Properly escape data printed to PKGCONF file (Dan Skorupski)
129* Fix tests when assert() undefined (Keith Bennett, Matt Stancliff)
130* Implement a reconnect method for the client context, this changes the structure of `redisContext` (Aaron Bedra)
131
132### 0.12.1 (2015-01-26)
133
134* Fix `make install`: DESTDIR support, install all required files, install PKGCONF in proper location
135* Fix `make test` as 32 bit build on 64 bit platform
136
137### 0.12.0 (2015-01-22)
138
139* Add optional KeepAlive support
140
141* Try again on EINTR errors
142
143* Add libuv adapter
144
145* Add IPv6 support
146
147* Remove possibility of multiple close on same fd
148
149* Add ability to bind source address on connect
150
151* Add redisConnectFd() and redisFreeKeepFd()
152
153* Fix getaddrinfo() memory leak
154
155* Free string if it is unused (fixes memory leak)
156
157* Improve redisAppendCommandArgv performance 2.5x
158
159* Add support for SO_REUSEADDR
160
161* Fix redisvFormatCommand format parsing
162
163* Add GLib 2.0 adapter
164
165* Refactor reading code into read.c
166
167* Fix errno error buffers to not clobber errors
168
169* Generate pkgconf during build
170
171* Silence _BSD_SOURCE warnings
172
173* Improve digit counting for multibulk creation
174
175
176### 0.11.0
177
178* Increase the maximum multi-bulk reply depth to 7.
179
180* Increase the read buffer size from 2k to 16k.
181
182* Use poll(2) instead of select(2) to support large fds (>= 1024).
183
184### 0.10.1
185
186* Makefile overhaul. Important to check out if you override one or more
187  variables using environment variables or via arguments to the "make" tool.
188
189* Issue #45: Fix potential memory leak for a multi bulk reply with 0 elements
190  being created by the default reply object functions.
191
192* Issue #43: Don't crash in an asynchronous context when Redis returns an error
193  reply after the connection has been made (this happens when the maximum
194  number of connections is reached).
195
196### 0.10.0
197
198* See commit log.
199
200