JRuby fails to bundle

[aemadrid]

When bundling in JRuby for development it fails because of sys-proctable

/usr/local/var/rbenv/versions/jruby-1.7.18/bin/jruby --2.0 /usr/local/var/rbenv/versions/jruby-1.7.18/bin/bundle install
WARN: Unresolved specs during Gem::Specification.reset:
      ffi (>= 0)
WARN: Clearing out unresolved specs.
Please report a bug if this causes problems.
Fetching gem metadata from http://rubygems.org/.Retrying dependency api due to error (2/3): Bundler::MarshalError ArgumentError: marshal data too short
Retrying dependency api due to error (3/3): Bundler::MarshalError ArgumentError: marshal data too short


Resolving dependencies.....
Could not find sys-proctable-0.9.4 in any of the sources

Process finished with exit code 7

[bschwartz]

Hi Adrian,

That's unfortunate. It's because of the dependency on nsq-cluster for testing. Ironically we moved to sys-proctable there for get better cross platform support (see: wistia/nsq-cluster@e46160d).

Looks like sys-proctable isn't planning to support JRuby anytime soon: djberg96/sys-proctable#18.

Any ideas for how to get around this? PR's welcome, of course!

Brendan

[aemadrid]

I'll give it a try. My initial idea is to remove the hard dependency, already in my fork, and then in the tests check if it can be loaded. If it can't then display a warning and skip that test. Now I haven't checked if you can run any of the tests without it. What do you think?

[bschwartz]

I think that sounds reasonable. It might be easier to try to do this at the nsq-cluster level though. We need nsq-cluster for pretty much all the nsq-ruby tests.

But nsq-cluster just needs sys-proctable for stopping nsqadmin, which none of these tests really need.

Hmm, this is a tricky problem :)

[aemadrid]

Hmh, it seems then that we need to change the hard requirement at nsq-cluster and follow the whining idea when you try to stop nsqadmin? Basically try to load sys-proctable when you actually try to stop nsqadmin and raise the exception there. Or add a warning message at nsq-cluster load time and also raise the exception?

[aemadrid]

Created this PR for nsq-cluster:

wistia/nsq-cluster#15

Please let me know what you think.