the Activity Monitor traceback: https://gist.github.com/a616e5afea7eafc8c979 fix is to synchronize calls, so shutdown can't happen while a call is in-flight. Conflicts: ext/zookeeper_base.rb
…x in the 1.0 release
spin up a new thread, join it if we're on a non-dispatch thread, otherwise just trust that somehow, our thread will get cleaned up.
Use a bunch of phony Gemfile links, so bundler can store a Gemfile.lock for each ruby impl. under tmp. after the first run, it took ~ 1:30 to run the tests for all interpreters
…d calling. It hadn't occurred to me, but there's no reason to: malloc, assign, call, free. The only place the arg struct's lifespan begins and ends inside the zkrb_call_* function, so we should allocate it on the stack, pass it to the zkrb_gvl_* function via zkrb_thread_blocking_region, and then use the return code. No malloc, no free, no problem.