single-artifact fix for AOT bug

[martinraison]

Here it is. This should fix plumatic/plumbing#74. I ran the tests against Clojure 1.6.0 and 1.7.0-RC1, everything passes in both cases.

[w01fe]

Thank you!!

My one concern is that emit-impl is private, and as such could cease to exist or change arbitrarily at any point in the future, which would create a new compatibility issue.

Two potential solutions that ensure forward compatibility:

• Check the version number instead: https://github.com/Prismatic/plumbing/blob/master/src/plumbing/core.cljx#L42
• Rather than checking for the issue being fixed, check if it's broken. Wrap the check in a try/catch, and assume it's fixed if an exception is thrown.

Thoughts?

Thanks again, much appreciated!

[martinraison]

Thanks for your feedback!

I updated the request with a more robust check. I basically try to call extend-protocol with fn excluded, and revert the exclusion if an exception is thrown. Is that close to what you had in mind?

I usually prefer avoiding direct version checks, especially since in this case we would need to check the version qualifier as well (1.7.0-RC1 is fixed, 1.7.0-alpha1 is not). This may not play well with forks of the Clojure compiler that use different version names. The only drawback I can think of is the unused "CLJ1195Test" interface lying around. I don't think this will impact anyone, though.

[w01fe]

Agree a functionality check is better than a version check, and this looks like a safe way to do it. Thanks again for your contribution, much appreciated!

Did you also test with AOT to confirm that the root issue is fixed? Either way, I'll merge now, cut a snapshot and ask for feedback on the plumbing thread, and then cut a real release as soon as we have confirmation that everything looks good.