| 1 |
Revision history for IO-Socket-IP |
| 2 |
|
| 3 |
0.44 2026-05-13 |
| 4 |
[CHANGES] |
| 5 |
* Can `use parent` rather than `use base` (thanks James Raspass) |
| 6 |
(RT177570) |
| 7 |
|
| 8 |
0.43 2024-11-25 |
| 9 |
[CHANGES] |
| 10 |
* Swapped unit tests from `Test::More` to `Test2::V0` |
| 11 |
* Fixed some documentation typoes |
| 12 |
(thanks Igor Sobrado Delgado) (RT157201) |
| 13 |
|
| 14 |
0.42 2023-07-25 |
| 15 |
[CHANGES] |
| 16 |
* Put error string in `$IO::Socket::errstr` as done by |
| 17 |
IO::Socket v1.45 |
| 18 |
* Updated for Perl v5.14 - use `package NAME VERSION` syntax |
| 19 |
|
| 20 |
[BUGFIXES] |
| 21 |
* Fix typo 'behvior' (RT133467) (thanks [email protected]) |
| 22 |
* Fix for DragonflyBSD and IPV6_V6ONLY from core perl (RT148293) |
| 23 |
|
| 24 |
0.41 2020-09-15 16:59 |
| 25 |
[CHANGES] |
| 26 |
* Updated to `=head2 barename` documentation style |
| 27 |
|
| 28 |
[BUGFIXES] |
| 29 |
* Remember to declare configure_requires Module::Build |
| 30 |
|
| 31 |
0.40 2020-09-15 16:41 |
| 32 |
[CHANGES] |
| 33 |
* Don't ship a Makefile.PL |
| 34 |
* Updated docs about `-register` option |
| 35 |
|
| 36 |
[BUGFIXES] |
| 37 |
* Ensure safety under default-strict (RT133322) |
| 38 |
|
| 39 |
0.39 2017/03/06 16:40:30 |
| 40 |
[BUGFIXES] |
| 41 |
* Handle EOPNOTSUPP in response of efforts to disable V6ONLY |
| 42 |
(RT119780) (thanks XENU) |
| 43 |
|
| 44 |
0.38 2016/08/02 18:33:50 |
| 45 |
[CHANGES] |
| 46 |
* Support setting custom socket options with new 'Sockopts' |
| 47 |
constructor parameter |
| 48 |
|
| 49 |
[BUGFIXES] |
| 50 |
* Restore blocking mode after ->connect errors (RT112334) |
| 51 |
|
| 52 |
0.37 2015/03/12 20:44:07 |
| 53 |
[CHANGES] |
| 54 |
* Earlier detection/reporting of erroneous Listen + Peer* argument |
| 55 |
combination; don't rely on resolveable hostname to do it (RT92930) |
| 56 |
|
| 57 |
[BUGFIXES] |
| 58 |
* Skip t/16v6only.t on irix (RT102662) |
| 59 |
* Remember to eval() $IO::Socket::VERSION in case it's a devel |
| 60 |
release (RT101954) |
| 61 |
* s/Port/LocalPort/ in unit tests (RT96064) |
| 62 |
|
| 63 |
0.36 2014/01/10 20:54:58 |
| 64 |
[BUGFIXES] |
| 65 |
* Always bind() on MSWin32 because Windows doesn't like it if we |
| 66 |
listen() having not done so (RT101396) |
| 67 |
|
| 68 |
0.35 2015/01/02 19:45:20 |
| 69 |
[BUGFIXES] |
| 70 |
* Restore blocking mode after timeout connect immediate success |
| 71 |
(RT100947) |
| 72 |
* Avoid CORE:: prefixing so global override modules work (RT101174) |
| 73 |
* Ensure that ->peer{host,port,hostname,service} never die even when |
| 74 |
unconnected (RT98759) |
| 75 |
|
| 76 |
0.34 2014/12/03 00:04:56 |
| 77 |
[BUGFIXES] |
| 78 |
* Don't select() for read-readiness when waiting for non-blocking |
| 79 |
connect() because it upsets VMS (RT100654) |
| 80 |
|
| 81 |
0.33 2014/11/21 16:17:39 |
| 82 |
[BUGFIXES] |
| 83 |
* Use SOL_SOCKET/SO_ERROR sockopt rather than 2nd connect() error |
| 84 |
hoisting to detect nonblocking connect/failure when implementing |
| 85 |
Timeout (RT98976) (thanks Arthur Axel 'fREW' Schmidt) |
| 86 |
|
| 87 |
0.32 2014/09/12 10:11:27 |
| 88 |
[CHANGES] |
| 89 |
* Implementation of Timeout for ->connect (RT92075) |
| 90 |
|
| 91 |
0.31 2014/07/11 01:24:49 |
| 92 |
[BUGFIXES] |
| 93 |
* A better fix for subclassing ->connect (RT95983) because 0.30 broke |
| 94 |
the way IO::Socket::SSL works (RT97050) |
| 95 |
|
| 96 |
0.30 2014/07/03 23:28:43 |
| 97 |
[CHANGES] |
| 98 |
* Invoke $self->connect fully-qualified in case of subclasses with |
| 99 |
override (RT95983) |
| 100 |
* Added extra diag() call in t/15io-socket.t (RT96914) |
| 101 |
|
| 102 |
0.29 2014/02/24 16:06:29 |
| 103 |
[BUGFIXES] |
| 104 |
* Workaround for OSes that disobey AI_ADDRCONFIG and yield AIs on |
| 105 |
families the kernel will not support anyway (e.g. HPUX) |
| 106 |
* Workaround for OSes that lack getprotobyname() (e.g. Android) |
| 107 |
|
| 108 |
0.28 2014/02/10 16:17:59 |
| 109 |
[CHANGES] |
| 110 |
* Renamed internal _configure method to _io_socket_ip__configure to |
| 111 |
avoid clashes with modules that try to subclass IO::Socket::INET |
| 112 |
|
| 113 |
[BUGFIXES] |
| 114 |
* Disable AI_ADDRCONFIG during one-sided 'v6 tests as sometimes it |
| 115 |
would otherwise fail |
| 116 |
* Skip the SO_BROADCAST test on OSes that fail with EACCES (RT92502) |
| 117 |
|
| 118 |
0.27 2014/01/20 18:08:31 |
| 119 |
[BUGFIXES] |
| 120 |
* Apply a short timeout to unit tests that probe for internet |
| 121 |
connectivity, in case of bad firewalls, etc... (Perl RT121037) |
| 122 |
* Defend against machines with IN6ADDR_LOOPBACK not being "::1" in |
| 123 |
unit tests, similar to the INADDR_LOOPBACK case (RT92295) |
| 124 |
|
| 125 |
0.26 2014/01/16 12:20:02 |
| 126 |
[CHANGES] |
| 127 |
* Set $VERSION in BEGIN block before 'use base' so that |
| 128 |
Acme::Override::INET still works |
| 129 |
* Various minor improvements to documentation |
| 130 |
|
| 131 |
0.25 2014/01/11 17:19:29 |
| 132 |
[BUGFIXES] |
| 133 |
* Allow both *Host with port and *Port arguments, letting port from |
| 134 |
Host argument take precedence (INET compat) |
| 135 |
* Ensure that a Host/Family-less constructor still constructs a |
| 136 |
socket, by using gai()s AI_ADDRCONFIG hint |
| 137 |
|
| 138 |
0.24 2013/09/19 14:17:22 |
| 139 |
[BUGFIXES] |
| 140 |
* Defend against INADDR_LOOPBACK not being 127.0.0.1 on machines with |
| 141 |
odd networking (e.g. FreeBSD jails) during unit testing |
| 142 |
|
| 143 |
0.23 2013/09/11 17:53:19 |
| 144 |
[BUGFIXES] |
| 145 |
* Check that kernel actually supports SO_REUSEPORT before |
| 146 |
unit-testing it (RT86177) |
| 147 |
|
| 148 |
0.22 BUGFIXES: |
| 149 |
* Always pass 'socktype' hint to getaddrinfo() in unit tests because |
| 150 |
some OSes get upset without it |
| 151 |
* Don't unit-test that sockaddr is updated with a local bind() on |
| 152 |
SOCK_DGRAM sockets because some OSes don't |
| 153 |
|
| 154 |
0.21 CHANGES: |
| 155 |
* Provide a downgrade method to turn an AF_INET-domain socket into a |
| 156 |
real IO::Socket::INET instance |
| 157 |
* Ensure that IO::Socket->new( Domain => ... ) definitely returns a |
| 158 |
socket in the right family |
| 159 |
|
| 160 |
0.20 CHANGES: |
| 161 |
* Skip the IPV6_V6ONLY tests on machines lacking that constant |
| 162 |
* Avoid Socket::inet_pton() in unit tests as Strawberry lacks it |
| 163 |
* Added ->join_addr |
| 164 |
* Respect subclassing argument to ->accept |
| 165 |
|
| 166 |
0.19 CHANGES: |
| 167 |
* More IO::Socket::INET compatibility fixes: |
| 168 |
+ Ensure ->connected is false after ->close |
| 169 |
+ Ensure that IO::Socket::IP->new(Family => $family) still creates |
| 170 |
an unbound, unconnected socket filehandle |
| 171 |
|
| 172 |
0.18 CHANGES: |
| 173 |
* Work around ->socktype being undef on IO::Socket versions before |
| 174 |
1.35 by fetching ->sockopt( SO_TYPE ) (RT81549) |
| 175 |
|
| 176 |
BUGFIXES: |
| 177 |
* Fix test skip counts in t/21nonblocking-connect-internet.t (RT79393) |
| 178 |
|
| 179 |
0.17 CHANGES: |
| 180 |
* Only pass AI_ADDRCONFIG as getaddrinfo() flag if no other flags |
| 181 |
specified, so the caller can disable it if required. |
| 182 |
|
| 183 |
BUGFIXES: |
| 184 |
* Don't count select() invocations during nonblocking unit test |
| 185 |
because this is too unreliable and races with the kernel/TCP stack |
| 186 |
|
| 187 |
0.16 BUGFIXES: |
| 188 |
* Don't ->close after all connect attempts fail, because |
| 189 |
IO::Socket::INET doesn't do that |
| 190 |
|
| 191 |
0.15 BUGFIXES: |
| 192 |
* (Hopefully) improved ->connect logic for better MSWin32 support |
| 193 |
* Skip testing AI_NUMERICSERV on OSes that don't support it |
| 194 |
|
| 195 |
0.14 BUGFIXES: |
| 196 |
* Second attempt at fixing test skip counts |
| 197 |
|
| 198 |
0.13 BUGFIXES: |
| 199 |
* Updated test skip counts in unit tests so they match the number of |
| 200 |
tests actually skipped |
| 201 |
|
| 202 |
0.12 CHANGES: |
| 203 |
* Ensure that all instances have a defined fileno after construction, |
| 204 |
even in non-blocking mode (RT77726) |
| 205 |
|
| 206 |
0.11 CHANGES: |
| 207 |
* Ignore unrecognised constructor arguments rather than complaining |
| 208 |
about their presence (RT77536) |
| 209 |
* Call $self->connect during constructor because IO::Socket::SSL |
| 210 |
relies on this to happen (RT77536) |
| 211 |
|
| 212 |
0.10 CHANGES: |
| 213 |
* Allow specifying other getaddrinfo flags using GetAddrInfoFlags |
| 214 |
constructor argument (RT75783) |
| 215 |
* Provide a convenient hostname+port string splitting utility method |
| 216 |
|
| 217 |
0.09 CHANGES: |
| 218 |
* Attempt to implement IPV6_V6ONLY sockopt wrapper - not all OSes can |
| 219 |
disable it, so detect and skip those |
| 220 |
* Provide ->sockaddr and ->peeraddr convenience accessors, same as |
| 221 |
IO::Socket::INET (RT75071) |
| 222 |
* Use new NIx_NOHOST and NIx_NOSERV flags to avoid redundant |
| 223 |
getnameinfo() lookups when only one of host or service name is |
| 224 |
required |
| 225 |
* Ensure that errno gets set to EINVAL on getaddrinfo() failures |
| 226 |
|
| 227 |
BUGFIXES: |
| 228 |
* Ignore existing-but-undefined constructor arguments |
| 229 |
* Avoid locale-specific error message testing of $! |
| 230 |
* Updates to unit tests for cygwin |
| 231 |
|
| 232 |
0.08 CHANGES: |
| 233 |
* Depend on Socket 1.95 now a full dual-life release has been made |
| 234 |
* Set the AI_ADDRCONFIG getaddrinfo hint |
| 235 |
* Fix some spelling mistakes/typoes |
| 236 |
* MSWin32 fixes: |
| 237 |
+ MSWin32 uses select() exceptfds rather than writefds to report |
| 238 |
on nonblocking connect() failure |
| 239 |
+ MSWin32 uses EWOULDBLOCK rather than EINPROGRESS to indicate |
| 240 |
nonblocking connect() |
| 241 |
|
| 242 |
BUGFIXES: |
| 243 |
* Ensure ->socket protocol argument is always defined |
| 244 |
* Bugfix for (e.g.) NetBSD, which gets upset at protocol hint without |
| 245 |
socktype hint to getaddrinfo() |
| 246 |
* Pass 'type' rather than 'proto' to constructor in t/11sockopts.t to |
| 247 |
avoid MSWin32 test failures |
| 248 |
* Localise $1/$2 to placate [perl #67962] |
| 249 |
|
| 250 |
0.07 CHANGES: |
| 251 |
* Prepare for Socket::getaddrinfo() in core; prefer it to |
| 252 |
Socket::GetAddrInfo::getaddrinfo() |
| 253 |
* Implement Family constructor arg |
| 254 |
* Optional registration with IO::Socket |
| 255 |
* Documentation rewordings |
| 256 |
* Bugfix to local socket connect tests - test using blocking rather |
| 257 |
than nonblocking sockets to guarantee synchronous packet delivery |
| 258 |
* Bugfix to t/04, t/05 - skip if unable to bind ::1 |
| 259 |
|
| 260 |
0.06 CHANGES: |
| 261 |
* Implement nonblocking connect |
| 262 |
* Implement LocalAddrInfo and PeerAddrInfo args |
| 263 |
* Example of nonblocking usage, using Net::LibAsyncNS |
| 264 |
|
| 265 |
0.05 CHANGES: |
| 266 |
* Implement combined ->bind and ->connect operations |
| 267 |
* Accept "host:service" as *Addr args, and PeerAddr as sole |
| 268 |
constructor argument |
| 269 |
* Accept "name(port)" as service arguments |
| 270 |
|
| 271 |
0.04 CHANGES: |
| 272 |
* Try to yield the most appropriate connect/bind/socket error |
| 273 |
* Implement sockhost/peerhost methods to match ::INET, moved hostname |
| 274 |
methods to sockhostname/peerhostname |
| 275 |
|
| 276 |
BUGFIXES: |
| 277 |
* Work around IO::Socket bug where ->accept'ed sockets do not get |
| 278 |
->sockdomain or ->socktype |
| 279 |
|
| 280 |
0.03 CHANGES: |
| 281 |
* Set $@ to raise error messages from constructor |
| 282 |
|
| 283 |
BUGFIXES: |
| 284 |
* Correctly implement 'Type' constructor argument |
| 285 |
* Test SO_BROADCAST on udp instead of tcp as some OSes forbid it |
| 286 |
|
| 287 |
0.02 BUGFIXES: |
| 288 |
* Not all OSes return true sockopts as 1; test simply for non-zero |
| 289 |
* Gracefully skip IPv6 tests if Socket6 unavailable |
| 290 |
|
| 291 |
0.01 First version, released on an unsuspecting world. |
| 292 |
|