Save last_error on init and emit

[cabiad]

Uses a threadlocal variable to store errors, which a user of this library can optionally access and act on after initializing the sender (initially connecting to the fluentd server) or emitting an event.

Also includes relevant tests.

[cabiad]

whoops, I'll resubmit in a sec without this change

[cabiad]

@kiyoto - I just wanted to follow up on our earlier email discussion. Do you think this PR can be merged?

[cabiad]

🔔 - Any update on this?

[kiyoto]

Ah sorry. @yyuu and @kzk do you think you can review and merge the changes?

[yyuu]

@kiyoto The test code won't success if fluentd isn't listening on 24224/tcp. It should be rewritten with using mock to pass tests on travis-ci.org.

// I found travis-ci.org is not working on fluent/fluent-logger-python. See also #32

[cabiad]

@yyuu and @kiyoto - The existing tests also fail in that case, but because error information was discarded, the failure was not detectable. I added comments to indicate this where appropriate to existing tests.

I believe that this is therefore an important fix. In my opinion, it makes sense to take this increment of value right away. We can update all tests to work properly without a running fluentd instance in another PR.

[ghost]

Hiding the connect or send error is crazy, this library needs to be fixed. Look how the ruby library returns a success or failure by bubbling up a true or false on the post() method here:
https://github.com/fluent/fluent-logger-ruby/blob/master/bin/fluent-post

This lib should ideally do the same, beside storing the error, a new event should have a post() method and that should return a true/false on success or not. That makes usage far more clear IMO.

[repeatedly]

I merged #59.
This patch is baesd on this PR with additional changes.
Thanks for the contribution!