test requires HTTP::Daemon 6.05+ and use 127.0.0.1 or [::1] according to server's sockdomain

[skaji]

We saw some test failures related IPv4/IPv6 (eg #278 #272 #101).

Now HTTP::Daemon 6.05 uses IO::Socket::IP, and IO::Socket::IP takes care of IPv4/IPv6,
so we should use "localhost" notations and let IO::Socket::IP select IPv4/IPv6 automatically.

Note:

• dzil build will append "test requires HTTP::Daemon 6.05" to META.json, so I didn't add it in this PR.
• t/TestServer.pm and t/cookies.t still use 127.0.0.1 notation. I hope we address them in another PR. ("localhost" is not a valid domain for cookies. So we cannot use "localhost" simply there)

[coveralls]

Pull Request Test Coverage Report for Build 273

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 92.711%

---

Totals
Change from base Build 265:	0.0%
Covered Lines:	725
Relevant Lines:	782

---

💛 - Coveralls

[oalders]

Or, we could also add this to TestRequires in dist.ini: https://github.com/libwww-perl/WWW-Mechanize/blob/master/dist.ini#L26-L27

[eserte]

Currently the prereq version for HTTP::Daemon is 0, but has be set to 6.05, otherwise tests will fail if older HTTP::Daemon is installed. Or alternatively skip tests if HTTP::Daemon version is < 6.05.

[skaji]

As I mentioned in the first comment, dzil build will bump HTTP::Daemon prereq.
I intentionally did not do it because it would also give us unrelated diff.

[skaji]

I manually bumped HTTP::Daemon in 68af39e

[oalders]

@skaji is this ready to merge?

[skaji]

I have cherry-picked a commit from patch2 branch 7d796b8.
jkeenan confirmed that tests passed with the commit libwww-perl/HTTP-Daemon#33 (comment).

@oalders If travis ci passes, I think this PR is ready to merge!

[jkeenan]

The CPANtester who originally reported the bad interaction between HTTP::Daemon and WWW::Mechanize (causing his rigs to hang indefinitely) has reported to me that skaji's branch patch2 corrects that problem. We are therefore eager to see a new WWW::Mechanize release to CPAN as soon as possible.

A new CPAN release is likely to resolve these 3 issues in your queue:
#253
#278
#281

Thank you very much.
Jim Keenan

[oalders]

Thanks @skaji, @eserte and @jkeenan!

[Corion]

Thank you all for fixing this! The log-server program at some time split off into its own CPAN module at Test::HTTP::LocalServer, where I added this fix as well!

I'll look into updating the program in WWW::Mechanize from Test::HTTP::LocalServer if I have features that the WWW::Mechanize distribution needs.