Handle read_timeout for queries executed within EventMachine

[gmodarelli]

Hello, I'm not sure if this is the best approach to handle a read_timeout when a query is being run within an EventMachine reactor loop, but I wanted to give it a try and pick your brain on the topic :)

The read_timeout option is only used when a query is executed synchronously, so it will be completely ignored for EM clients.

mysql2/ext/mysql2/client.c

Lines 782 to 792 in b3fe727

	rb_rescue2(do_send_query, (VALUE)&args, disconnect_and_raise, self, rb_eException, (VALUE)0);
	if (rb_hash_aref(current, sym_async) == Qtrue) {
	return Qnil;
	} else {
	async_args.fd = wrapper->client->net.fd;
	async_args.self = self;
	rb_rescue2(do_query, (VALUE)&async_args, disconnect_and_raise, self, rb_eException, (VALUE)0);
	return rb_ensure(rb_mysql_client_async_result, self, disconnect_and_mark_inactive, self);

In order to expose a similar behaviour I've introduced a Deferrable Timeout which will raise a Mysql2::EM::ReadTimeout.

In case of a read timeout from the normal non-evented client, a Mysql::Error is raised, but I am not sure it would be ok to raise that exception in this case.

I also understand that a change like this one would break compatibility. Maybe there is a better way to expose this timeout logic to the client code. On the other hand it is also a bit counter intuitive to have a read_timeout option that doesn't do anything when the client is evented :)

Please let me know what you think about it.
Have a nice day!

[gmodarelli]

I think the failing test on Ruby 2.3.4 is just a random failure. I've run the tests locally with Ruby 2.3.4 and they are green

[sodabrew]

Can you switch to the defer.errback pattern see in the tests below? It is problematic to throw exceptions outside the scope of EM.run (throwing exceptions like this in fact would crash EM 1.2.x until I finally finally figured out the problem in EM 1.2.6)

[gmodarelli]

Sure, I'll switch to defer.errback :)