Undefined method "method_class" for AMQ::Protocol::BodyFrame about amqp HOT 12 CLOSED

@alexnorthsoul in what context does this happen? A BodyFrame is not supposed to be the first frame in the buffer (from my understanding). I was fighting this issue myself today.

For me, an exception was thrown on the callback block of the first message received, causing the channel's buffer to not be cleared by this line:

clear_frames_on(frame.channel) if @frames[frame.channel]

in #receive_frame. So each subsequent message received on that channel would call first_frame.method_class on the leftover BodyFrame.

I was about to submit an issue myself but just saw yours.

My thought is that #receive_frameset should have some sort of assurance of cleanup. Even if one of my messages blows out the top of the call stack, AMQP shouldn't leave itself in a bad state.

Anywhere you call user specified code without a guaranteed cleanup there can be bugs like the one outlined above.

from amqp.

@Inanepenguin's hypothesis sounds realistic

from amqp.

@michaelklishin any chance the "graceful cleanup" in #receive_format could make it in? I can look into trying to submit a PR, I'd have to come up to speed on the style/best practice for the repo. Like I said, I was about to open an issue myself as this just happened to me today. Very coincidental timing!

from amqp.

Feel free to submit a PR, thank you.

On 1 oct 2015, at 11:32, Peter Wilson [email protected] wrote:

@michaelklishin any chance the "graceful cleanup" in #receive_format could make it in? I can look into trying to submit a PR, I'd have to come up to speed on the style/best practice for the repo. Like I said, I was about to open an issue myself as this just happened to me today. Very coincidental timing!

—
Reply to this email directly or view it on GitHub.

from amqp.

The issue is still exists in gem 1.6.0 but solved in 1.5.1. Do you plan to update 1.6.x gem ?

from amqp.

@relonger that's curious. What specific commit is in 1.5.x-stable but not master? Happy to release a new 1.6.x version later today or tomorrow.

from amqp.

This commit dd0b2de seems to solve the problem.

from amqp.

@relonger I cherry-picked 02a9298 to master, please give it a try.

from amqp.

Could you release a 1.6.1 with this commit included ?

from amqp.

@maattdd do you confirm that the issue is gone in master?

from amqp.

@michaelklishin yes I confirm this bug is in 1.6.0 but fixed in master.

from amqp.

@maattdd I released a 1.7.0.

from amqp.