3.4.0 release

[samuelcolvin]

Please can we have a new release.

In particular, I need #3062, since it's a bug and there's no easy workaround for it.

The 3.3 milestone has a lot of stuff on it, if there's no plan to release for a while, perhaps #3062 would be backported as a patch?

[kxepal]

We already have 3.3.2 release. Did you mean 3.4.0?

[samuelcolvin]

sorry, yes.

[samuelcolvin]

@asvetlov?

[asvetlov]

Sorry, was pretty busy last weeks.
Yes, it's time for aiohttp 3.4
Please let me review open issues first.
Maybe we can close some of them quickly.

[samuelcolvin]

Great, sorry to hassle.

[webknjaz]

I think I've adjusted CIs to produce wheels for 3.7 as well ;)

[asvetlov]

Cool!

[samuelcolvin]

bump, any update on this? I'm still waiting on #3062.

[asvetlov]

The showstopper is: I'd like to publish a new multidict release with Python 3.7 wheels first.
Now we have problems with building macosx wheels on Travis.
@webknjaz is working on it but still no success (after that I expect the same fix for aiohttp itself is needed).
As an alternative, we can drop MacOSX wheels entirely (very tempting solution BTW)

[webknjaz]

@asvetlov that issue is related to me breaking Travis config during migration to tox. This one should be good to go.

[kxepal]

@webknjaz offtopic a bit, but why tox? No trolling, just curiouosity.

[webknjaz]

It's defacto standard. I know that there's some pitfalls and @asvetlov personally does not like it. That's why we're testing this in multidict, which is smaller. Andrew also asked me to investigate pyinvoke.
Other thing to take into account is that contributions are most likely to be proficient with tox and unfamiliar with other tools.
There's already some tox configs, which some contributors try to maintain on their own, which is sad, because it would be more straightforward if we followed some popular practices. Also, for Andrew, I've wrapped calls to tox via Makefile.

[kxepal]

Thanks! Going to join on watching this experiment. I had tried tox once upon a time, but things turned too much complicated with no oblivious profit /: May be today will be different picture or I'd missed something.

[webknjaz]

@kxepal tox still tends to be complicated :(
However, it would let us achieve the long-term goal of unification, where users will be able to run the same envs/tests as CIs.
Also, this makes it better portable across CIs whenever you want to migrate/add new ones.
Among nice things: virtualenv caching/reuse/isolation (different env for each task by default, can be changed; reuse of settings from other envs; env matrix generation (so-called factors).

[samuelcolvin]

Released

[asvetlov]

@samuelcolvin sorry for the long waiting

[samuelcolvin]

No problem, we all have busy lives.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a [new issue] for related bugs.
If you feel like there's important points made in this discussion, please include those exceprts into that [new issue].
[new issue]: https://github.com/aio-libs/aiohttp/issues/new