Mailman 3 August 2016 - Twisted

[Twisted-Python] close to the finish line on hosted windows testing
by Glyph Lefkowitz Aug. 9, 2016

Aug. 9, 2016

Thanks to Craig and Adi's work, we now have tests sort of working on appveyor. However, they aren't actually reporting their pass/fail status right now as github statuses, I am assuming because of the persistent failures that were causing noise on unrelated tickets. However, it's very close to being a useful builder; there are really only 2 remaining issues, not counting turning on the reporting itself. I've made an umbrella ticket here: https://twistedmatrix.com/trac/ticket/8664 <https://twistedmatrix.com/trac/ticket/8664> and documented what's wrong with each one. So, if you care about Twisted on Windows, fixing the issues documented there should make it much easier for more folks to contribute easily :). -glyph

3 2

[Twisted-Python] INCOMPATIBLE CHANGE: twisted.python.dist renamed to twisted.python._setup
by Adi Roiban Aug. 8, 2016

Aug. 8, 2016

Hi, I would like to remove twisted.python.dist as a public API without following the 1 year deprecation processes. The module has a comment which informs people not to use it outside of Twisted, but I think it should be private module instead. The ticket is here https://twistedmatrix.com/trac/ticket/8622 The PR is here https://github.com/twisted/twisted/pull/467 The branch is here https://github.com/twisted/twisted/tree/8622-setup-coverage This is part of the setup.py cleanup effort for having most of the setup.py code outside of setup.py itself so that we can test it as part of the automated test suite. twisted.python.dist3 was kept as a private API as it is already used by https://github.com/habnabit/twisted-depgraph Please let me know if you are using twisted.python.dist outside of Twisted and are against this change or if you approve it. Thanks! -- Adi Roiban

8 11

[Twisted-Python] twisted.python.dist as a public API
by Adi Roiban Aug. 7, 2016

Aug. 7, 2016

Hi, Why is twisted.python.dist3.py a public module? Are there any external projects which need to get the list of the modules ported to Python3? ------------------ Why is twisted.python.dist a public module? The docstring starts with `Don't use this outside of Twisted.' but why not named it twisted.python._dist or twisted.python._setup to make it clear that it is private ? ------- I am trying to clean up the setup.py code as I feel that having the setup() definition split between setup.py and dist.py is not ideal ---------- Can I rename twisted.python.dist and remove unused code from twisted.python.dist without a deprecation dance? -------------- The dedicated ticket is here https://twistedmatrix.com/trac/ticket/8622 and is in the review queue Regards, -- Adi Roiban

2 3

[Twisted-Python] RFC: IPv6 multicast join/ticket 6597
by Jason Litzinger Aug. 6, 2016

Aug. 6, 2016

Hello, I'm looking at making the changes to support IPv6 multicast groups as described in ticket 6597 but wanted to get some feedback (and get a feel whether this is even desirable) before formally submitting any patches. Specifically: 1. I've read [1] and it alludes to udp hopefully disappearing, is that something in the works? Is there a new approach to solving this problem I should look at? A branch in the works where this (conceptual) change belongs? Note: The addressFamily attribute referenced already exists and is set properly for IPv6. 2. The attached change has the side effect that calls to ReactorBase.resolve() with IPv6 literals will now likely succeed where they may have failed in the past. That means clients counting on resolve raising an exception for an IPv6 literal will break. Not sure whether this is considered a compatibility issue, but I wanted to raise it. 3. One alternative to the above would be a complete API separation, via something like joinIPv6Group(), and a new resolve. Is that more appealing in this case? Caveats: 1. I have not finished all of the documentation related to developers, I will do so prior to formal submission. 2. I know I need tests and docs and will submit them with the final changes. On to the patches. With these changes, I can use the joinGroup API to add myself to an IPv6 multicast group on Linux (verified via /proc/net/igmp6). Additionally, trial reports the same two failures before and after these changes (twisted.python.test.test_release.APIBuilderTests.test_build and test_buildWithPolicy). These changes struck me as the obvious approach, but given the changes to resolve, not necessarily the best. diff --git a/twisted/internet/base.py b/twisted/internet/base.py index 4f2c862..e813741 100644 --- a/twisted/internet/base.py +++ b/twisted/internet/base.py @@ -567,6 +567,8 @@ class ReactorBase(object): return defer.succeed('0.0.0.0') if abstract.isIPAddress(name): return defer.succeed(name) + elif abstract.isIPv6Address(name): + return defer.succeed(name) return self.resolver.getHostByName(name, timeout) # Installation. diff --git a/twisted/internet/udp.py b/twisted/internet/udp.py index b5a5322..210b079 100644 --- a/twisted/internet/udp.py +++ b/twisted/internet/udp.py @@ -485,7 +485,10 @@ class MulticastMixin: def _joinAddr1(self, addr, interface, join): - return self.reactor.resolve(interface).addCallback(self._joinAddr2, addr, join) + if self.addressFamily == socket.AF_INET: + return self.reactor.resolve(interface).addCallback(self._joinAddr2, addr, join) + else: + return self.reactor.resolve(interface).addCallback(self._joinAddrIPv6, addr, join) def _joinAddr2(self, interface, addr, join): @@ -500,6 +503,18 @@ class MulticastMixin: except socket.error as e: return failure.Failure(error.MulticastJoinError(addr, interface, *e.args)) + def _joinAddrIPv6(self, interface, addr, join): + addr = socket.inet_pton(socket.AF_INET6, addr) + interface = socket.inet_pton(socket.AF_INET6, interface) + if join: + cmd = socket.IPV6_JOIN_GROUP + else: + cmd = socket.IPV6_LEAVE_GROUP + try: + self.socket.setsockopt(socket.IPPROTO_IPV6, cmd, addr + interface) + except socket.error as e: + return failure.Failure(error.MulticastJoinError(addr, interface, *e.args)) + def leaveGroup(self, addr, interface=""): """Leave multicast group, return Deferred of success.""" This does require the client specify the interface argument when calling joinGroup, e.g. self.transport.joinGroup("ff02::1", interface="::"). Thanks in advance for any feedback! -Jason Litzinger [1] http://twistedmatrix.com/pipermail/twisted-python/2016-March/030188.html

4 12