client: make Redis::Client#ensure_connected handle fork reconnects

[yaauie]

If @reconnect is enabled and #ensure_connected is called in the forked child of
the process that established the connection, a reconnect attempt should be made;
handle this case like any other ConnectionError.

[saizai]

+1

[saizai]

See also #389 and #364.

[yaauie]

It looks like the build failure is unrelated, and only on jruby-19mode conn=ruby REDIS_BRANCH=2.8:

 1) Failure:
test_resolves_localhost(TestInternals) [/home/travis/build/redis/redis-rb/test/internals_test.rb:347]:
Exception raised:
<#<Redis::CannotConnectError: Timed out connecting to Redis on localhost:6381>>.

[yaauie]

Can I do anything to help this move forward?

[Bertg]

I was wondering if @pietern is not moving on this because of the work he is doing with the hiredis project.
Maybe this type of patch is not compatible with it; and he'd like to keep them in sync?

[pietern]

@Bertg This is compatible. The functionality in redis-rb is just a superset of the functionality in hiredis-rb.

@yaauie Thanks for this patch. It's the most elegant one I've seen so far to address this problem. The test failure you have seen is a flaky one. The JRuby ones fail every now and then, usually related to I/O errors or timeouts.

[Bertg]

Sweet. Thx @pietern and @yaauie.
Any indication when this will be released as a gem?

[rrauenza]

+1 on the gem -- can we get a 3.0.8?

[yaauie]

If there's anything I can do to assist in getting this into a release (any other bugs that need squashing or features that need completion/verification), I'd be more than willing to lend help.

[eljuanchosf]

Guys, we are in desperate need of a 3.0.8 release with this fix. How can we help?

[yaauie]

This feature was released in 3.1.0.