Several sets of changes

[temujin9]

First commit fixes a bug with rvm::system and mysql::client (which runs gem_package at compile time).

Second commit is a workaround for an internal bug resulting in GID collision between recipes. It's less likely to be generally useful, but probably still worth including.

Third commit fixes an (intermittent?) bug found in first commit.

Fourth commit uses gem_package's native code if ruby_string == 'system', which will be a bit more useful once rvm::gem_package can use multiple environments.

[fnichol]

These are great adds and fixes, thanks! There's only one thing I had a question about. I was looking at pushing the rvm gem install into the execute phase to be more in line with the other opscode cookbooks. It looks like most are being updated to clear out these kinds of problems (chef-boneyard/cookbooks@9597eb7). But you're right that if there are cookbooks running gem_package in the compilation phase then we don't have an RVM environment setup yet.
Other than the mysql cookbook (which looks updated now currently), can you think of other cookbooks with this kind of problem? I'm hoping I can get a few comments back (https://twitter.com/#!/fnichol/status/82127542496149504).
Otherwise, the other commits look ripe for pulling in.

[temujin9]

I can see legitimate need for other cookbooks to put gem_package calls in the compile phase, but MySQL's the only cookbook that we encountered this problem with so far.

[fnichol]

Pulled in and updated a couple of things. Mainly:

• We only need to install RVM in the compilation phase if recipe[rvm::gem_package] is included in the run_list. So default behavior can stay the same, but will be altered on inclusion of the gem_package recipe.
• The default RVM ruby and list of rubies don't need to be installed in the compilation phase. The RVMRubygems package provider calls the rvm_environment resource just before installing the gem, so whatever RVM ruby/gemset will get "lazily" created on demand at that point.
• Adding support for multiple rubies in gem_packge's rvm_string parameter meant some adjustments to the fourth commit, but I believe the desired behavior remains.

Thanks again for your help, I think we have a more robust gem_package solution as a result.

[temujin9]

My pleasure. Some more bug-fixes on this coming shortly.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.