On Feb 26, 2008, at 10:27 AM, Chad Woolley wrote: > Scenario 1: Existing rubygems install, with some gems installed in the > default system location. I want to programatically (such as > Gem::GemRunner) install some additional gems in a user-writeable > location (~/gems). After additional gems are installed, they should > all (system and user-writable) be loadable with the 'gem' method > and/or require.
According to the documentation of your shell, set GEM_PATH to `gem env gempath` with ~/gems added set GEM_HOME to ~/gems At this point, `gem install` will install gems into ~/gems. > Scenario 2: Same as above, but on mac OS X Leopard, with all gems > under default locations /System/Library/... and /Library/... still > loadable Same. > That's it for now. What I'm looking for is a specific set of steps to > accomplish the above scenarios. I don't need details on how to > programatically install gems, I know that, but managing the env vars, > default dir, and path is where I'm confused. I should add a page to `gem help` for this. I'm switching `gem env gempath` to be usable by shell scripts for the next release, so you won't have to manually join them with ':'. > I have some hacks to do this, but it involves manually setting up > GEM_HOME and GEM_PATH before rubygems is ever required. This is > hard in some environments - such > as irb on Leopard, where rubygems is already required, and > ENV['GEM_PATH'] is frozen. Maybe hard on others, such as ruby 1.9 > with bundled rubygems? I haven't played with this under 1.9... Gem.clear_paths will force RubyGems to re-read GEM_HOME and GEM_PATH. _______________________________________________ Rubygems-developers mailing list [email protected] http://rubyforge.org/mailman/listinfo/rubygems-developers
