Load Net::DNS::Native before IO::Socket::IP

[Grinnz]

Summary

With the current load order Net::DNS::Native is never used in one setup because it fails to load due to IO::Socket::IP defining non thread safe symbols.

Motivation

Loading it first seems to fix the issue.

References

***********************************************************************
Some package defined non thread safe symbols which Net::DNS::Native uses internally
Please make sure you are not placed loading of modules like IO::Socket::IP
before this one and not called functions like getaddrinfo(), gethostbyname(),
inet_aton() before loading of Net::DNS::Native
************************************************************************

[Grinnz]

I can't think of any reasonable way to test this.

[Grinnz]

Through further testing both IO::Socket::IP and Mojo::IOLoop::TLS must be loaded after, probably due to Mojo::IOLoop::TLS loading IO::Socket::SSL.

[Grinnz]

15:30:13 <Grinnz> https://github.com/mojolicious/mojo/issues/716 some context
15:31:18 <Grinnz> i believe it's because this perl was built without pthreads
15:31:29 <Grinnz> which was not default until ...
15:31:52 <Grinnz> 5.22

#716

[kraih]

Calling for a vote @mojolicious/core.

[marcusramberg]

I'm not sure if optimizing for older Perls is worth it. Are there any downsides to the change?

[Grinnz]

The downsides are the introduced inconsistency in code layout, and that it can only be a partial solution, the setup can still break randomly (by break I mean it just won't use NDN) but at least Mojo::IOLoop::Client can load it in some cases.

[Grinnz]

For the record, I've worked around this in my code by loading Net::DNS::Native before anything from Mojo.

[jhthorsen]

I think since it will randomly break, then I'm negative to this change. I think a documentation update would be better.

[jberger]

As I mentioned on IRC, and agreeing with others, I don't think that because it cannot really address the issue I don't think this change is merited. I agree with @jhthorsen that a documentation patch would be better however I really wonder if this is an upstream issue with NDN and IO::Socket::*. 👎

[kraih]

Thanks all, with at least two negative votes this pull request cannot pass the vote.

[leonerd]

As author of IO::Socket::IP - is there anything for me to do with this, or that I should know about?

[Grinnz]

Only what is in the referenced issue #716.