Fix connector

[asvetlov]

Prevent hang on connection acquiring after connection failure

[codecov-io]

Codecov Report

Merging #2567 into 2.3 will decrease coverage by <.01%.
The diff coverage is 87.5%.

@@            Coverage Diff             @@
##              2.3    #2567      +/-   ##
==========================================
- Coverage   97.35%   97.34%   -0.01%     
==========================================
  Files          39       39              
  Lines        8253     8260       +7     
  Branches     1437     1439       +2     
==========================================
+ Hits         8035     8041       +6     
  Misses         96       96              
- Partials      122      123       +1

Impacted Files	Coverage Δ
aiohttp/connector.py	97.34% <87.5%> (-0.16%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4e7a2c4...8d6a0ca. Read the comment docs.

[webknjaz]

@asvetlov is this double__underscore notation intentional?

[asvetlov]

Yes. It is a test for _drop_acquire_per_host method.

[pfreixes]

Why the clients that are waiting for an available connection hang in case of the previous one had an error? I cant see this bad behavior in the current code. Could you please point me to the right path?

Also, I don't get the point of the test_error_in_performing_request test, how is it testing that situation?

[asvetlov]

self._waiters was not cleaned up, https://github.com/aio-libs/aiohttp/blob/2.3/aiohttp/connector.py#L357 said that available is not positive.
connect() method created a waiter but no other connection was exist, nobody set resumed the waiter future.

That's why test_error_in_performing_request failed.

I need review the logic for available, looks like len(self._waiters) is a wrong check.

[asvetlov]

Dropped a check for len(self._waiters), self._acquired is enough.

[pfreixes]

Oks, got it. LGTM can be merged once all checks passes

[pfreixes]

IMHO this kind of bug is enough important to have a bugfix release for the 2.X series

[asvetlov]

@pfreixes will publish 2.3.4 just after merging

[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