On Ubuntu Impish (development version), we're currently facing a
problem where thin's testsuite fails when it runs on an armhf system.
After some investigation, I noticed that the reason for this failure
was that the server was still running even after the loop on
spec/spec_helper.rb:stop_server.

Given that the loop is actively checking to see if the server has
stopped, I'd like to propose that the timeout be increased from 1
second to 10 seconds.  Although I'm not a fan of the
"let's-just-increase-the-timeout" type of solutions, in this
particular case I think it's justified and makes sense (1 second may
indeed be too quick for the server to properly stop).

Signed-off-by: Sergio Durigan Junior <sergio.durigan@canonical.com>

Bump version to 1.8.1

Fix possible HTTP Response Splitting

Ignore any response header with \r or \n in their value

See GHSA-84j7-475p-hp8v

Update Rake

Old URL doesn't exist anymore. Closes #373

This test recently started to fail on Fedora. I suspect that either
Kernel or systemd increased RLIMIT_NOFILE by default to 512k.

There should probably be additional logic, which would initialize the
maximum_connections variable I refactored. `prlimit` can display the
actual value.

One-line fix, as suggested in https://stackoverflow.com/a/25631436/446106

Resolves #275

…server is finished.

Ruby changed the way Kernel#warn reports the error [1], but mocking the
method on the class, which is calling it, should be universal [2].

[1] https://bugs.ruby-lang.org/issues/12944
[2] https://www.reddit.com/r/ruby/comments/2nki9q/rspec_how_do_you_mock_or_stub_kernel_methods_like/