Speed up install by disabling RDOC generation about rvm

[yfeldblum]

Who needs rdoc/ri generation on a server? Especially on an automated server?

Skipping rdoc/ri generation would be fantastic.

[fnichol]

Agreed, I hate that as well (rubygems/rubygems#42). Only thing I'm iffy on is going against the rubygems default (principal of least surprise), even if it's kinda lame. Besides, it's possible that the cookbook could be used for development workstations as well.

What do you both think about an attribue, something like:

default['rvm']['gem_docs'] = true

which could be overridden to false (that's what I'd drop in my role definitions that tweak rvm defaults).

To be honest, I forgot about this build delay because sooner or later I drop in a /etc/gemrc file and it's done. Good catch!

[rosenfeld]

I think it is a good idea, but I think the default should be false because it is the most common option for most automated install environments.

[fnichol]

I don't think my arm needed too much twisting there. Will do (make false the default) then those who want the docs can at least ask to get them. Now to update...

[rosenfeld]

Thanks!

[yfeldblum]

+1

[fnichol]

So, it looks like the Chef::Provider::Rubygems provider already kills docs on gem install (https://github.com/opscode/chef/blob/0.9-stable/chef/lib/chef/provider/package/rubygems.rb#L420). Can either of you confirm with -l debug on a chef execution? Here is an example when I bundler gets installed:

[Sun, 01 May 2011 21:12:18 -0700] DEBUG: Processing rvm_gem[bundler] on cb-tester
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: Executing grep -q "^bundler" "/usr/local/rvm/gemsets/global.gems"
: stdout
[default] [Sun, 01 May 2011 21:12:18 -0700] DEBUG: ---- Begin output of grep -q "^bundler" "/usr/local/rvm/gemsets/global.gems" ----
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: STDOUT: 
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: STDERR: 
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: ---- End output of grep -q "^bundler" "/usr/local/rvm/gemsets/global.gems" ----
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: Ran grep -q "^bundler" "/usr/local/rvm/gemsets/global.gems" returned 1
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: Executing echo "bundler" >> "/usr/local/rvm/gemsets/global.gems"
[default] [Sun, 01 May 2011 21:12:18 -0700] DEBUG: ---- Begin output of echo "bundler" >> "/usr/local/rvm/gemsets/global.gems" ----
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: STDOUT: 
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: STDERR: 
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: ---- End output of echo "bundler" >> "/usr/local/rvm/gemsets/global.gems" ----
[Sun, 01 May 2011 21:12:18 -0700] DEBUG: Ran echo "bundler" >> "/usr/local/rvm/gemsets/global.gems" returned 0
[Sun, 01 May 2011 21:12:18 -0700] INFO: Ran execute[Add bundler to /usr/local/rvm/gemsets/global.gems] successfully
[default] [Sun, 01 May 2011 21:12:19 -0700] DEBUG: sh(bash -c "source /etc/profile.d/rvm.sh" && rvm ruby-1.8.7-p334@global gem env gempath)
: stdout
[default] [Sun, 01 May 2011 21:12:20 -0700] DEBUG: No installed version found for bundler (>= 0, runtime)
[Sun, 01 May 2011 21:12:20 -0700] DEBUG: sh(bash -c "source /etc/profile.d/rvm.sh" && rvm ruby-1.8.7-p334@global gem env)
: stdout
[default] [Sun, 01 May 2011 21:12:21 -0700] DEBUG: Found gem bundler version 1.0.12 for platform ruby from http://rubygems.org/
[Sun, 01 May 2011 21:12:21 -0700] INFO: Installing gem_package[bundler] version 1.0.12
[Sun, 01 May 2011 21:12:21 -0700] DEBUG: sh(bash -c "source /etc/profile.d/rvm.sh" && rvm ruby-1.8.7-p334@global gem install bundler -q --no-rdoc --no-ri -v "1.0.12")
: stdout
[default] [Sun, 01 May 2011 21:12:28 -0700] DEBUG: Processing execute[Add bundler to /usr/local/rvm/gemsets/global.gems] on cb-tester
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: Executing grep -q "^bundler" "/usr/local/rvm/gemsets/global.gems"
: stdout
[default] [Sun, 01 May 2011 21:12:28 -0700] DEBUG: ---- Begin output of grep -q "^bundler" "/usr/local/rvm/gemsets/global.gems" ----
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: STDOUT: 
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: STDERR: 
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: ---- End output of grep -q "^bundler" "/usr/local/rvm/gemsets/global.gems" ----
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: Ran grep -q "^bundler" "/usr/local/rvm/gemsets/global.gems" returned 0
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: Skipping execute[Add bundler to /usr/local/rvm/gemsets/global.gems] due to not_if
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: Processing gem_package[bundler] on cb-tester
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: Found installed gem bundler version 1.0.12 matching bundler (>= 0, runtime)
[Sun, 01 May 2011 21:12:28 -0700] DEBUG: Doing nothing for gem_package[bundler]

[rosenfeld]

Hi Fletcher, I think we are talking about different things. You are talking about the 'gem_package' definition, while I'm talking about installing Ruby through RVM. I believe that gem_package won't generated RDoc, but while installing RVM Ruby and watching 'top', I've noticed that 100% of the CPU was being used for generating the docs for the initial/default gems installed by RVM.

I talked with Waynee and told him that we would like to avoid changing gemrc for disabling RDoc generation for the initial gems while installing RVM Ruby. He told me that we could add 'export rvm_gem_options="--no-rdoc --no-ri"' to rvmrc.

What do you think about this option?

[fnichol]

Great, I saw your patch to RVM. I'll get this added in the next couple of days, thanks for your help!

[fnichol]

Done! Looks okay?

[rosenfeld]

Yes, it looks ok. I just didn't understand why using "enable"/"disable" instead of true/false. Another approach also would be something like:

default[:rvm][:rvm_gem_options] = "--no-rdoc --no-ri"

What do you think about it?

[yfeldblum]

I would at least name it skip_docs_on_gem_install because it affects only the installation of gems, not the installation of rvm or of any rubies.

I might also tend to name it in the positive with a zero/null/empty default, as the following examples:

default[:rvm][:gem_install_docs] = false # option 1
default[:rvm][:gem_install_docs] = [] # option 2

# can be overridden:
node[:rvm][:gem_install_docs] = true # option 1
node[:rvm][:gem_install_docs] = %w[rdoc ri] # option 2

[fnichol]

How does this look? I agree, that feels better. The reason for those crazy "enable"/"disable" values is because I didn't realize you could use boolean value in attributes and the rvm/install_rubies was already in that form. Oh well, it's pre-1.0 so API changes are allowed (so that's updated too).

[yfeldblum]

+1

[rosenfeld]

Awesome! :)

[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.