Get IO::Socket::IP into core for seamless v4/v6 support #12707
Comments
From @wolfsageCreated by wolfsage@gmail.comThere've been efforts to get IO::Socket::IP into the core distribution to Doing this would allow IPv4 and IPv6 support to be shipped with Perl. I believe that this is a good idea, and in #p5p we came up with the 1. Encourage people to switch to IO::Socket::IP now Perl Info |
From @wolfsageI've done some preliminary setup and testing, but I need to do more I hope to get to more soon (tomorrow, this weekend). Cheers, -- Matthew Horsfall (alh) |
From [Unknown Contact. See original ticket]I've done some preliminary setup and testing, but I need to do more I hope to get to more soon (tomorrow, this weekend). Cheers, -- Matthew Horsfall (alh) |
|
@wolfsage - Status changed from 'new' to 'open' |
From @leonerdOn Wed, Jan 16, 2013 at 07:17:57PM -0800, Matthew Horsfall via RT wrote:
Hi there, Thanks for taking a start at this one. How is progress? What's working currently...? Is there anything that -- leonerd@leonerd.org.uk |
From @wolfsageOn Mon, Feb 4, 2013 at 7:27 AM, Paul LeoNerd Evans
Paul, I hadn't got much done on this, but as discussed earlier in chat, I've There are 3 tests that fail that may or may not be important (at least one Failed 3 tests out of 2252, 99.87% okay. The full test output is attached. If you'd like to download and hack on the branch to try to fix these, you git://github.com/wolfsage/perl-1.git Something like: git clone git://github.com/wolfsage/perl-1.git If you clone this through github, feel free to push patches with pull Thanks, And again, I apologize for taking so long to get back to this. -- Matthew Horsfall (alh) |
From @leonerdOn Mon, Feb 04, 2013 at 07:52:34PM -0500, Matthew Horsfall (alh) wrote:
Will take a look...
This unit test employs Unwarranted Chumminess With Code Under Test. ok 2 I can't see a way to change INET.pm to handle this case; the entire unit
Oops; this I expect is due to a lack of use IO::Socket::IP -register; And a lack of Family => AF_INET in the @_-passing case And that IO::Socket::IP didn't construct a new unbound, unconnected
Fails on the ->connected test, because of IO::Socket::IP's subtly Now fixed upstream in IO::Socket::IP. Find attached a combined patch with newer IO::Socket::IP and fixes to -- leonerd@leonerd.org.uk |
From @leonerd0002-Newer-IO-Socket-IP-minor-bugfixes-to-fake-INET.pm-wr.patchFrom 7974d32e0d5c18a40d55b34201a0a3c011d96ca8 Mon Sep 17 00:00:00 2001
From: "Paul \"LeoNerd\" Evans" <leonerd@leonerd.org.uk>
Date: Mon, 11 Mar 2013 23:14:00 +0000
Subject: [PATCH 2/2] Newer IO::Socket::IP; minor bugfixes to fake INET.pm
wrapper to make more of IO's unit tests pass
---
cpan/IO-Socket-IP/lib/IO/Socket/IP.pm | 15 ++++++++++++---
dist/IO/lib/IO/Socket/INET.pm | 3 ++-
2 files changed, 14 insertions(+), 4 deletions(-)
diff --git a/cpan/IO-Socket-IP/lib/IO/Socket/IP.pm b/cpan/IO-Socket-IP/lib/IO/Socket/IP.pm
index c40bc67..26dc3de 100644
--- a/cpan/IO-Socket-IP/lib/IO/Socket/IP.pm
+++ b/cpan/IO-Socket-IP/lib/IO/Socket/IP.pm
@@ -1,7 +1,7 @@
# You may distribute under the terms of either the GNU General Public License
# or the Artistic License (the same terms as Perl itself)
#
-# (C) Paul Evans, 2010-2012 -- leonerd@leonerd.org.uk
+# (C) Paul Evans, 2010-2013 -- leonerd@leonerd.org.uk
package IO::Socket::IP;
@@ -9,7 +9,7 @@ use strict;
use warnings;
use base qw( IO::Socket );
-our $VERSION = '0.18';
+our $VERSION = '0.18_001';
use Carp;
@@ -512,6 +512,15 @@ sub _configure
}
}
+ if( !@infos and defined $hints{family} ) {
+ # If there was a Family hint then create a plain unbound, unconnected socket
+ @infos = ( {
+ family => $hints{family},
+ socktype => $hints{socktype},
+ protocol => $hints{protocol},
+ } );
+ }
+
# In the nonblocking case, caller will be calling ->setup multiple times.
# Store configuration in the object for the ->setup method
# Yes, these are messy. Sorry, I can't help that...
@@ -633,7 +642,7 @@ sub connected
my $self = shift;
return defined $self->fileno &&
!${*$self}{io_socket_ip_connect_in_progress} &&
- defined $self->peername;
+ defined getpeername( $self ); # ->peername caches, we need to detect disconnection
}
=head1 METHODS
diff --git a/dist/IO/lib/IO/Socket/INET.pm b/dist/IO/lib/IO/Socket/INET.pm
index 3c04b57..e98461d 100644
--- a/dist/IO/lib/IO/Socket/INET.pm
+++ b/dist/IO/lib/IO/Socket/INET.pm
@@ -1,5 +1,6 @@
package IO::Socket::INET;
+use IO::Socket::IP -register;
use base qw( IO::Socket::IP );
use Socket qw( PF_INET );
@@ -7,7 +8,7 @@ sub new
{
my $class = shift;
return $class->SUPER::new(Family => PF_INET, PeerAddr => shift) if @_ == 1;
- return $class->SUPER::new(@_);
+ return $class->SUPER::new(Family => PF_INET, @_);
}
1;
--
1.7.10.4
|
From @wolfsageAlright, I've updated https://github.com/wolfsage/perl-1/tree/rt-116433 I've removed multihomed.t since it's testing internal state of a module All tests pass on my system, though META.JSON and META.YAML aren't up to rjbs / porters - can we get this branch smoked now? What do I need to do? Thanks, -- Matthew Horsfall (alh) |
From @rjbs* Matthew Horsfall via RT <perlbug-followup@perl.org> [2013-04-05T15:39:33]
Here, the manifest test is broken, too. porting/manifest.t (Wstat: 0 Tests: 10765 Failed: 3) I fixed the two porting tests and pushed the branch to The result is not immediately useful to me: 1..26 There is a long pause between ok 11 and death. -- |
From @wolfsageOn Sat, Apr 6, 2013 at 10:59 AM, Ricardo Signes via RT <
I finally reproduced this and tracked it down. Here in io_sock.t line 172 on: # Then test UDP sockets The call to IO::Socket->new() is creating a handle and then blessing it to Currently discussing options with LeoNerd to get this moving along. -- Matthew Horsfall (alh) |
From @wolfsageWhoops, forgot cc... On Tue, Apr 23, 2013 at 12:34 PM, Matthew Horsfall (alh) <wolfsage@gmail.com
|
From @leonerdOn Tue, 23 Apr 2013 12:34:49 -0400
This I'm classing as a bug in use IO::Socket::IP -register; Because IO::Socket::new deletes the Domain argument, there is nothing -- leonerd@leonerd.org.uk |
From @wolfsage
LeoNerd pushed a dev branch to CPAN of IO::Socket::IP containing a fix for rjbs has pulled it into the smoke me branch and now we're waiting testing -- Matthew Horsfall (alh) |
From @leonerdOn Wed, 24 Apr 2013 12:25:41 -0400
FYI, that smoked OK so I've called it a real IO::Socket::IP 0.21 -- leonerd@leonerd.org.uk |
From @craigberryOn Wed, Apr 24, 2013 at 11:25 AM, Matthew Horsfall (alh) <wolfsage@gmail.com
I built this branch as of smoke-me/rjbs/io-socket-ip 2013-04-24.15:51:27 and got the following new test failures on VMS: cpan/IO-Socket-IP/t/11sockopts ................................ 11sockopts.t passes when run by itself outside of the test suite. The ones ok( !$socket->blocking, '$socket->blocking' ); Something internally is setting the blocking flag to on even though the $ perl blocking.pl where blocking.pl looks like: $ type blocking.pl use IO::Socket::IP; my $testserver = IO::Socket::INET->new( my $socket = IO::Socket::IP->new( print "Blocking? : " . ($socket->blocking ? 'yes' : 'no') . "\n"; The ones that end with "connect-internet" fail on the test that looks like: ok( !$socket->connected, '$socket not yet connected' ); and elsewhere as well, though the other failures may be fallout from the $ perl t/21nonblocking-connect-internet.t I believe everything that goes wrong does so somewhere |
From @wolfsageOn Thu, Apr 25, 2013 at 5:04 PM, Craig A. Berry <craig.a.berry@gmail.com>wrote:
Craig, Any chance I or LeoNerd could get access to one of the failing systems so Thanks, -- Matthew Horsfall (alh) |
From @craigberryOn Tue, Apr 30, 2013 at 9:30 AM, Matthew Horsfall (alh)
I'm not against the idea but it would take quite a bit of my time and yours I did more or less get to the bottom of the 11sockopts.t failure, which is DB<4> s If you put "setsockopt broadcast privilege" into your favorite search Side note: it seems wrong to me that IO::Socket::setsockopt does not check I will try to get to the nonblocking failures soonish. |
From @wolfsageOn Tue, Apr 30, 2013 at 8:24 PM, Craig Berry via RT <
Fair enough though now you've got me curious about VMS :)
If you have the time to do this that's great. Thanks a lot. -- Matthew Horsfall (alh) |
1 similar comment
From @wolfsageOn Tue, Apr 30, 2013 at 8:24 PM, Craig Berry via RT <
Fair enough though now you've got me curious about VMS :)
If you have the time to do this that's great. Thanks a lot. -- Matthew Horsfall (alh) |
From @leonerdOn Tue, 30 Apr 2013 19:23:44 -0500
Oh it's totally ancilliary; if this doesn't work on VMS it's quite I'm curious to know though - is this a new failure since 0.21, or has -- leonerd@leonerd.org.uk |
From @craigberryOn Wed, May 1, 2013 at 4:45 PM, Paul LeoNerd <leonerd@leonerd.org.uk> wrote:
The first hunk in the attached patch does that. It also has a second hunk, That gets all the IO::Socket::IP tests in the smoke-me/rjbs/io-socket-ip I'm sorry it took me so long to put together this trivial patch; things And the patch inline for human readability: Inline Patch--- cpan/IO-Socket-IP/t/11sockopts.t;-0 Wed Apr 24 10:51:26 2013
+++ cpan/IO-Socket-IP/t/11sockopts.t Fri May 3 15:50:14 2013
@@ -39,7 +39,9 @@ SKIP: {
ok( $sock->getsockopt( SOL_SOCKET, SO_REUSEPORT ), 'SO_REUSEPORT set' );
}
-{
+SKIP: {
+ skip "SO_BROADCAST requires elevated privileges on $^O", 1 if $^O eq
Inline Patch--- dist/IO/lib/IO/Socket.pm;-0 Wed Apr 24 10:51:26 2013
+++ dist/IO/lib/IO/Socket.pm Fri May 3 15:45:21 2013
@@ -167,7 +167,7 @@ sub blocking {
my $sock = shift;
return $sock->SUPER::blocking(@_)
- if $^O ne 'MSWin32';
+ if $^O ne 'MSWin32' && $^O ne 'VMS';
# Windows handles blocking differently
# |
From @craigberryio-socket-ip-vms.patch--- cpan/IO-Socket-IP/t/11sockopts.t;-0 Wed Apr 24 10:51:26 2013
+++ cpan/IO-Socket-IP/t/11sockopts.t Fri May 3 15:50:14 2013
@@ -39,7 +39,9 @@ SKIP: {
ok( $sock->getsockopt( SOL_SOCKET, SO_REUSEPORT ), 'SO_REUSEPORT set' );
}
-{
+SKIP: {
+ skip "SO_BROADCAST requires elevated privileges on $^O", 1 if $^O eq 'VMS';
+
my $sock = IO::Socket::IP->new(
LocalHost => "127.0.0.1",
Type => SOCK_DGRAM,
--- dist/IO/lib/IO/Socket.pm;-0 Wed Apr 24 10:51:26 2013
+++ dist/IO/lib/IO/Socket.pm Fri May 3 15:45:21 2013
@@ -167,7 +167,7 @@ sub blocking {
my $sock = shift;
return $sock->SUPER::blocking(@_)
- if $^O ne 'MSWin32';
+ if $^O ne 'MSWin32' && $^O ne 'VMS';
# Windows handles blocking differently
#
|
From @leonerdOn Fri, 3 May 2013 16:34:41 -0500
Excellent, that sounds good.
Oh that's quite alright - often it can take a while to gain the sudden
That sounds OK, but I wonder if we can provide a better detection
That sounds good. I mean, non-ideal, but perhaps unavoidable, and in -- leonerd@leonerd.org.uk |
From @rjbs* "Craig A. Berry" <craig.a.berry@gmail.com> [2013-05-03T17:34:41]
Great, thanks. -- |
From @craigberryOn Sun, May 5, 2013 at 2:45 PM, Ricardo Signes <perl.p5p@rjbs.manxome.org>wrote:
Thanks, it's in as d358b6c. |
From @wolfsageOn Sun, May 5, 2013 at 4:49 PM, Craig A. Berry <craig.a.berry@gmail.com>wrote:
Thanks a lot everyone. I think there's one failure left to look at on Darwin ( getaddrinfo() - service name not available for the specified socket type at But I notice that same failure is affecting some other oses on CpanTesters: There are some other failures at Thanks, -- Matthew Horsfall (alh) |
From @wolfsageOn Mon, May 6, 2013 at 12:59 PM, Matthew Horsfall (alh)
I haven't forgotten this. The plan is to figure out the failure on I should have some free time soon to play around and possibly figure out Cheers, -- Matthew Horsfall (alh) |
From @wolfsageOn Wed, Jun 19, 2013 at 3:40 PM, Matthew Horsfall (alh)
Soon came and went, and then I finally made the time. smoke-me/rjbs/io-socket-ip is running blead with IO::Socket::IP 0.22 Provided the looks good enough, we can finally move onto smoking the branch Thanks for your patience, -- Matthew Horsfall (alh) |
From michiel.beijen@gmail.comOn Thu, Aug 15, 2013 at 2:52 PM, Matthew Horsfall (alh)
|
From @wolfsageOn Mon, Sep 2, 2013 at 10:23 AM, Michiel Beijen <michiel.beijen@gmail.com>wrote:
I think so. None of the failures look related: http://perl.develop-help.com/?b=smoke-me%2Frjbs%2Fio-socket-ip I'd like to see it smoke across more OSes, but this is all we have so far. I think taking this to the next step now would be good. Ricardo? Thanks, -- Matthew Horsfall (alh) |
From @xdgAfter some discussion on IRC today about the current state of affairs, * The "drop in replacement for IO::Socket::INET" language in the docs * IO::Socket::IP (with revised language) will be added to blead ASAP * Separately, tests of the "drop in" assertion will continue outside This ensures that v5.20.0 will finally have full in-core support for IPv6. If the drop-in testing goes well enough, the "drop in" language in the -- |
From @rjbs* David Golden <xdg@xdg.me> [2014-01-08T13:05:19]
…with the (perhaps obvious) proviso that it would be removed before 5.20 if too -- |
From @nwc10On Thu, Jan 09, 2014 at 09:02:49AM -0500, Ricardo Signes wrote:
By using Acme::Overide::INET (thanks xdg, this makes things a lot easier)* https://rt.cpan.org/Public/Bug/Display.html?id=91982 In one case, we seem to be getting a silent constructor failure. Nicholas Clark * Not yet on CPAN, because the PAUSE hardware is very unhappy. See |
From @leonerdOn Thu, 9 Jan 2014 16:49:21 +0000
Having looked at them, they both fall into the category of "existing Hard to see what a sensible course of action is here. I'm not overly -- leonerd@leonerd.org.uk |
From @xdgOn Thu, Jan 9, 2014 at 11:49 AM, Nicholas Clark <nick@ccl4.org> wrote:
Available here for now: Once PAUSE is up, I'll ship it and blog about it. David -- |
From @xdgOn Thu, Jan 9, 2014 at 1:39 PM, Paul "LeoNerd" Evans
How about making such cases fatal? Then at least if someone tries to David -- |
From @leonerdOn Fri, 10 Jan 2014 11:00:48 -0500
Well, one of these can indeed do this (which is hinted at here However, the other bug relates more to what the caller does with the -- leonerd@leonerd.org.uk |
From @leonerdOn Thu, 9 Jan 2014 16:49:21 +0000
I believe both of these are now fixed. Patches in the bugs. Lacking a PAUSE still, you can find the current devel snapshot that http://home.leonerd.org.uk/local/IO-Socket-IP-0.24+bzr219.tar.gz Ifwhen PAUSE comes back (or some other suitable replacement appears) -- leonerd@leonerd.org.uk |
From @leonerdOn Sat, 11 Jan 2014 00:08:43 +0000
PAUSE appears to be up and running again: http://metacpan.org/release/PEVANS/IO-Socket-IP-0.25 -- leonerd@leonerd.org.uk |
From @wolfsageI'm finally going through the smoke run I did of blead with IO::Socket::IP drop-in replacement of IO::Socket::INET vs blead. IO::Socket::IP is at .26, both branches smoked are at: commit b373396 I'll keep adding issues as I find them so they have more time to be worked on. So far: Catalyst::Authentication::Credential::RemoteHTTP IO::Event IO::Socket::SSL -- Matthew Horsfall (alh) Lots of noise to go through (There are ton of modules that fail randomly due to Hash key order and timing issues....) -- Matthew Horsfall (alh) |
From [Unknown Contact. See original ticket]I'm finally going through the smoke run I did of blead with IO::Socket::IP drop-in replacement of IO::Socket::INET vs blead. IO::Socket::IP is at .26, both branches smoked are at: commit b373396 I'll keep adding issues as I find them so they have more time to be worked on. So far: Catalyst::Authentication::Credential::RemoteHTTP IO::Event IO::Socket::SSL -- Matthew Horsfall (alh) Lots of noise to go through (There are ton of modules that fail randomly due to Hash key order and timing issues....) -- Matthew Horsfall (alh) |
From @leonerdOn Wed, 19 Feb 2014 05:21:59 -0800
0.26? Latest on CPAN is 0.28 and that is mostly test fixes, though also [CHANGES]
Hmmm. I thought IO::Socket::SSL could optionally use IO::Socket::IP if -- leonerd@leonerd.org.uk |
From @wolfsagePerl 5.20.0 was released and contains IO::Socket::IP. I'm marking this ticket resolved, and will try to continue the work of getting it more tested (including the issues previously mentioned) for 5.22. -- Matthew Horsfall (alh) |
|
@wolfsage - Status changed from 'open' to 'resolved' |
Migrated from rt.perl.org#116433 (status was 'resolved')
Searchable as RT116433$
The text was updated successfully, but these errors were encountered: