2.19 2017-05-11T13:02:47Z
commit aac1b06c1e3b0d73ca964794bab0c444d454ebcc
Author: Yasuhiro Matsumoto <mattn.jp@gmail.com>
Date: Thu May 11 18:40:13 2017 +0900
don't run command when the port is empty
Test::TCP do hack on Windows for checking port used.
https://github.com/tokuhirom/Test-TCP/commit/0f4510d8
But this way always return 0 with succeeded with empty port. This is not
expected behavior. So this change fix to call check_port if the port is
not defined.
fixes#60
2.18 2017-04-24T06:00:25Z
- let the kernel select an empty port #59https://github.com/tokuhirom/Test-TCP/pull/59
(i110)
-------------------------------
2.16 2016-04-13T02:27:43Z
commit 3ae73a349f70f06c60de3f635916929bc8643429
Author: Alex Vandiver <alex@chmrr.net>
Date: Tue Apr 12 00:33:21 2016 -0700
Ensure that IO::Socket::IP is DESTROY'd before can_bind returns
be316f1f worked around a core perl bug[1] which causes Perl < 5.24.0
to DESTROY the IO::Socket::IP object too late, causing the empty port
to not actually be empty if it was used immediately.
Unfortunately, this change was lost in ba745fd1, which again put the
return of `can_bind` on the stack along with the actual return value,
re-instating the bug.
Split the statement across two statements again, and add a comment on
the importance of preserving it.
[1] https://rt.perl.org/Public/Bug/Display.html?id=124248
2.15 2016-03-15T00:25:52Z
- Add listen_socket function and listen option for race-free operation
Upstream changes:
2.14 2015-09-29T22:36:44Z
- Fix race condition in t/10_oo.t(exodist)
2.13 2015-07-24T02:30:17Z
- check whether the OS implements IPV6_V6ONLY before using it
2.12 2015-05-18T08:14:30Z
- Fixed spelling mistake
(Reported by gregor herrmann)
2.11 2015-04-07T00:07:25Z
- declare IO::Socket::IP as dependency #36
2.10 2015-04-06T19:23:43Z
- ensure the test object is DESTROYed when Net::EmptyPort::empty_port exits https://rt.cpan.org/Public/Bug/Display.html?id=103299
2.09 2015-04-02T21:55:18Z
- fix tests running for a long time on systems that do not support IPv6 #35
2.08 2015-04-02T04:04:33Z
- add `host` argument to various functions for binding to arbitrary address (incl. IPv6) #33
- add function `Net::EmptyPort::can_bind` #34
--------------
2.07 2015-01-22T10:44:21Z
Old versions of Test::SharedFork are incompatibles with the new
Test::Builder: I had "missing TB2::History" errors. It would be
helpful to upgrade the Test::SharedFork dependency to help the user to
avoid to encounter such hard to track deep errors.
(Reported by dolmen++)
Upstream changes:
2.06 2014-07-01T10:01:44Z
commit 8259d5eb28919bc766c8b500151d5be7e944b7f2
Author: Petr P.sa. <ppisar@redhat.com>
Date: Fri Jun 27 13:37:20 2014 +0200
Wait infinitely if max_wait is negative
The t/12_pass_wait_port_options.t will fail if the server process does not
start listening in max_wait limit. This can happen if the host is
loaded or just if the scheduler decides to postpone the process.
This patch adds possibility to wait infitely by passing a negative
max_wait value to the Test::TCP object and it changes the
t/12_pass_wait_port_options.t test to use this feature.
https://github.com/tokuhirom/Test-TCP/issues/28
Signed-off-by: Petr P.sa. <ppisar@redhat.com>
2.05 2014-06-24T00:49:45Z
- Release again with latest minil.
2.04 2014-06-23T23:42:28Z
- Release.
2.03_02 2014-06-23T23:37:07Z
- Release to CPAN.
2.03_01 2014-06-23T23:34:38Z
- Switch to ExtUtils::MakeMaker.
2.03 2014-06-23T10:18:53Z
- Re-packaging with Minilla v2.0.0-TRIAL
2.02 2013-10-30T03:22:39Z
- Fixed fork(2) error handling.
(tokuhirom)
2.01 2013-09-22T04:13:53Z
[Changes from MITHALDU (Christian Walde)]
- test waitport argument by running full code-chain, not with partial mocking
- The partial mocks cause stuck forks on win32 at times, causing the test to
hang. With the full code chain present the test runs reliably
Upstream changes:
2.00 2013-06-11T04:41:43Z
[INCOMPATIBLE CHANGES]
- Change wait_port() API to take max_wait seconds, and wait exponentially.
Dropped wait_port_retry argument introduced at 1.27.
(miyagawa)
Upstream changes:
1.30 2013-06-03T05:06:28Z
- Increase the default sleep and retry: 0.01 sec and 1000 times
(miyagawa)
1.29 2013-05-29T01:12:47Z
- Care the `test_tcp(port => undef)` case.
https://github.com/tokuhirom/Test-TCP/issues/13
1.28 2013-05-28T11:14:27Z
- Added wait_port_sleep and wait_port_retry option for
Test::TCP::wait_port, Test::TCP#new, test_tcp().
(tokuhirom)
Upstream changes:
1.27 2013-05-15T10:13:18Z
- export wait_port from Net::EmptyPort
(Pavel Shaydo)
- Make default wait_port() sleeping time shorter.
0.1sec => 0.0001sec.
(tokuhirom)
1.26 2013-03-29T08:46:20Z
- Remove unused deps from cpanfile.
1.25 2013-03-29T08:07:24Z
- re packaging, again & again
1.24 2013-03-28T02:49:32Z
- re-packing, again.
1.23 2013-03-27T01:37:08Z
- Just re-packaging
1.22 2013-03-27T01:17:38Z
- Add doc for Test::TCP::CheckPort
(Tatsuhiko Miyagawa)
- revert b4fc237697648234c960f6714d995210d4250e42 to fix Win32 breakage
Test::TCP::CheckPort must be used on win32 because it prints to STDOUT
(Christian Walde)
- Add UDP support to Net::EmptyPort
(Brendan Byrd)
Upstream changes:
1.18
- Added more documentation about empty_port()
1.17
commit 0f4510d83ae2f2ce112288c64289fa36d32865b3
Author: Christian Walde <walde.christian@googlemail.com>
Date: Sat Jul 28 07:04:40 2012 +0200
fix port checking on Win32 by performing it in another process
On Win32 fork is emulating by creating another thread in the same process.
This leads to a possible bug/race condition when a server tries to open a
port and listen on it, while in the same process a client tries to connect
to the same port. This manifests by the accept call of the server failing
with an error of "Bad file descriptor".
This is easily fixed by having another process perform the port checking,
since that will not interfere with the internals.
Upstream ChangeLog:
1.16
[TEST FIX]
- 1.15 breaks AIX fix on 1.14.
t/05_sigint.t skips on perl <= 5.8.8.
1.15
[TEST FIX]
${^CHILD_ERROR_NATIVE} is only available in perl 5.8.9 and later.
(kazeburo++)
1.14
- Fixed testing issue on AIX.
https://rt.cpan.org/Public/Bug/Display.html?id=72779
(w.phillip.moore)
- depend to Test::SharedFork 0.19 for better TB2 support
(tokuhirom)
Changes from previous:
1.13
- RT#67292: Tests are blocking in Windows 7.
https://rt.cpan.org/Ticket/Display.html?id=67292
- change port number range from 10000 .. 11000 to 50000 .. 60000 to
respect IANA.
https://rt.cpan.org/Ticket/Display.html?id=64012
1.12
- workaround for win32 test fails.
https://rt.cpan.org/Ticket/Display.html?id=66016
- more diagnostic messages
1.11
- localize $@ in Test::TCP::DESTROY
1.10
- added Test::TCP->stop method
1.09
- added document in FAQ section
1.08
- no feature changes
1.07_01
- new OO interface!
1.07
- allow forking in the client(lestrrat)
1.06
- depend to Test::SharedFork 0.14. It fixes issue with Test::Builder2.
1.05
- workaround for $@ issue in test case.
Data::Util's method modifier clears $@.
Pkgsrc changes:
- adjust dependencies
Upstream changes:
1.04
- fixed local $@ issue. this happens on some version of perl5.
1.03
- release to cpan
- fixed win32 issue(charsbar)
1.02_02
- use randomness on finding empty port(suggested by kazuhooku)
- try to connect the port before bind(Tatsuhiko Miyagawa)
1.02_01
- better cleanup code by RAII pattern.
https://rt.cpan.org/Ticket/Display.html?id=60657
(reported by dgl)
1.02
- lazy loading issue was fixed at Test::SharedFork 0.12.
Depend to it.
https://rt.cpan.org/Public/Bug/Display.html?id=60426
(reported by J.)
1.01
- remove unused deps for use_test_base().
1.00
- bump up version!
0.16_02
- oops. packaging miss.
0.16_01
- Do not depend to IO::Socket::INET 1.31.
Test::TCP works well with older IO, I hope.
(suggested by mst)
