Ulf Wiger

Ah, very good! :) I will concede that I must have forgotten to stop both nodes before making a new attempt. I see an opportunity for some more tests here,...

Refactoring into a helper function seems like a good first step. Using `function_exported/3` would probably work. The only thing to ensure is that an initial check is done to see...

Sorry, I now realize I was a little bit hasty before: my suggestion obviously doesn't address the deprecation warnings. @tsloughter, specifically which compile macro did you have in mind? A...

The `.travis.yml` /really/ needs to be updated, since it actually only tests versions between R15B01 and 17.1, none of which arguably should be tested at all. I'm fine with supporting...

Have you tried starting each node with `-gproc gproc_dist all`?

Good point. This should be added.

The function `gproc:reg_or_locate/1` is only intended for unique names, and should crash if called with a (non-unique) property: ```erlang Eshell V7.3.1 (abort with ^G) 1> application:ensure_started(gproc). ok 2> gproc:reg_or_locate({p,l,foo}). **...

No, this is not intended. Nor does it make sense, of course, to allow you to register an object, but not unregister it. I note also that the documentation is...

I agree that it should be clarified. Currently, one can infer it from [the documentation of `gproc:select/2`](https://github.com/uwiger/gproc/blob/master/doc/gproc.md#select2). That is: gproc defines a logical view of the registry as `[{Key, Pid,...

First benchmarks indicate that the changes incur an overhead of a few percent, up to almost 25% for `update_counter()`. While the extra cost seems acceptable for custom types, some special-casing...