Net::OpenTimeout is a Faraday::ConnectionFailed and not a Faraday::Ti…

[grosser]

…meoutError

according to

faraday/lib/faraday/adapter/net_http.rb

Line 34 in a217471

exceptions << Net::OpenTimeout if defined?(Net::OpenTimeout)

open timeout is considered a failed connection, so let's make net-http-persistent behave like net-http ... or I can open the opposite PR to make both raise timeout 🤷‍♂

In general I'd think it would be nice to consider Net::OpenTimeout a timeout so that the retry middleware handles it by default, it was added in #438

@mislav @iMacTia @mike-bourgeous @rusikf

[grosser]

wtf ... Method perform_request has a Cognitive Complexity of 6 (exceeds 5 allowed). Consider refactoring. ... want a PR to kick this out and use rubocop instead ?

[grosser]

also CI setup seems to be broken ? https://circleci.com/gh/lostisland/faraday/tree/master
... can anyone fix that ?
I had 1 error on master and this does not introduce any new errors 😓

[iMacTia]

Hi @grosser, thanks for raising this.
If I understand your point (and it is a very good point), I think it was extensively covered already here: #718

We also came up with a proposed solution, however there's a big issue here: your solution and the one proposed in that discussion are both breaking backwards compatibility.

For this reason, I'm afraid I'm unable to accept this PR at the moment as we decided that v1.0 will be fully backwards compatible and can only provide new features (like the streaming requests that you contributed to ❤️).

We'll address the issue with errors and attempt to standardise them in v2.0, but that work won't be limited to net-http-persistent

[grosser]

is anyone working on v2 ?
... would be nice to also make connections stateful by default, so we don't really need net-http-persistent at all since it would keep the connection open as long as the user uses the same connection object ... or at least make it not use this awkward global cache that erm someone contributed to fix it ...

[iMacTia]

No one actively working on v2 as the core team is focusing on getting v1 over the line at the moment. Works on v2 will likely start shortly after the release of v1, however we are beginning to collect the feature requests that will go into v2, as you can see from the milestone for the issue I linked above.

If you think there's something that we should seriously address in v2, you're more than welcome to open an issue for it and we'll mark it the same way so we won't forget.

Lastly, we really appreciate your contributions, and I'm sure you built some knowledge on Faraday, so I hope you'll be willing to lend some help once v2 works will start 👍