kristian: yeah I realized I had forgotten your update. I'll pull them in.
Woodie: I started to make the native split, but now I'm having second thoughts.
Warbler currently just pulls in core and stdlib jars from jruby-jars. Without an update, anyone using existing warbler versions will see a behavioral difference if we split out the jruby-native stuff into a separate jar. Specifically, anyone with webapps using FFI (e.g. nokogiri) or any system-level features (symlink, stat, etc) will see those features suddenly break when they update jruby-jars to 1.6 and do nothing else.
We also have so many release artifacts now that it's getting difficult to manage them. I think we should look at some other options for GAE JRuby. Maybe one of these:
- A separate gem jruby-jars-appengine that's specifically tailored for GAE. It could be even more aggressive about stripping things out, like removing the whole FFI subsystem, jline/readline, and more.
- Pull from Maven; all deps are there, and there's various ways we could automate pulling in dependencies. In addition, if we pulled dependency libs from Maven and deployed with GAE we'd also benefit from GAE's own caches of those jars.
If the Maven approach isn't appropriate, perhaps we should see about transitioning the appengine-specific jar stuff to a separate gem you guys can depend on? I'd be happy to spin up a project that we can both maintain that's structured specifically for GAE's limits on jar size. How does that sound?