`rabbit_common` dependency doesn't compile on newer rebar versions about amqp HOT 3 CLOSED

@mschae @mobileoverlord Thanks for reporting this.

d0rc/amqp_client compiles beause it has the the flag warnings_as_errors commented in rebar.conf. In jbrisbin/amqp_client`, which is the one used to build the community hex package, the flag is on.

I actually submitted a couple of PR's to jbrisbin's repo to get rid of the warnings. I'm still not sure what changed in the latest rebar to cause the warning about the behaviour gen_server2 being undefined. I will look into this and update the amqp_client hex package.

A similar issue, with the supervisor2 behaviour was discussed in the RabbitMQ forum (https://groups.google.com/forum/#!topic/rabbitmq-users/ThCiHBOXMlE). It was later fixed upstream, so I'll check if they already handled the gen_server2 behaviour warning.

from amqp.

@mschae
I updated to rebar 2.5.1 but so far was unable to reproduce the problem.

Here's my log after compiling your mschae/rebar_problem: https://gist.github.com/pma/bff0f180b32d28d092b4

I noticed that in the failed Travis build log, gen_server2.erl should have been the first file to be compiled, but it's not. This is enforced in rebar.conf by https://github.com/jbrisbin/rabbit_common/blob/master/rebar.config#L2.

Edit: After updating .travis.yml I now see the build failing on Travis, but still can't reproduce the problem locally.

from amqp.

@mschae
Could be related to this issue in rebar (rebar/rebar#445). As I understand the issue can cause rebar to discard the file compilation order set by erl_first_files in rebar.config and use the order of the filesystem.

When using container-based infrastructure for the Travis build, the file order of the filesystem could be different. Would explain why I'm not able to reproduce this locally and why my previous travis.yml did not fail (fe10f9f).

from amqp.