Wrong variable name - rvm_custom_path

[edbond]

It seems variable name should be rvm_path not rvm_custom_path:
set :rvm_custom_path, '~/.myveryownrvm'
See https://github.com/capistrano/rvm/blob/master/lib/capistrano/tasks/rvm.rake#L10
custom_path doesn't works for me, rvm_path do:

set :rvm_path, '~/.rvm'

[kirs]

Exactly. Could you make a PR, please?

[Kriechi]

I think this functionality is implemented in the init task.
The rvm_custom_path is fetched, then checked against the system/user dirs and stored as rvm_path which is then used from there on.
Maybe there is a error somewhere in there, but based on the idea one should only use and set rvm_custom_path.

[edbond]

Yes, I think there is a bug somewhere. I had this config:

set :rvm_type, :user
set :rvm_ruby_version, '2.0.0'

and task:

  task :update_crontab do
    on roles(:app) do
      within current_path do
        execute :bundle, "exec whenever -i cr"
      end
    end
  end

Running the task:

 INFO [21b760d9] Running /bin/rvm 2.0.0 do bundle exec whenever -i cr on server
.........
DEBUG [21b760d9]    bash: /bin/rvm: No such file or directory

It doesn't detect user rvm here.
Setting rvm_custom_path results in the same error

set :rvm_custom_path, '~/.rvm'

rvm_path works fine:

set :rvm_path, '~/.rvm'

 INFO [33b1d923] Running ~/.rvm/bin/rvm 2.0.0 do bundle exec whenever -i cr on server

Probably the source of this is user rvm detection failure.
HTH

[edbond]

ok, it seems culprit is here:

namespace :deploy do
  after :starting, 'rvm:hook'
end

I don't do deploy, just run a specific task:

cap production deploy:update_crontab

[Kriechi]

yes - as stated in the README you have to hook your custom tasks.
See https://github.com/capistrano/rvm#custom-tasks-which-rely-on-rvmruby

[Kriechi]

@kirs maybe we should do invoke rvm:hook before this line: https://github.com/capistrano/rvm/blob/master/lib/capistrano/tasks/rvm.rake#L10

This would ensure that rvm is hooked prior running one of the binaries (bundler, gem, ruby, rake).
Is that possible?

[kyledecot]

Any update this?

[Kriechi]

The main issue @edbond was having, should be solved by hooking RVM, as stated in the README (see above comment).

As for the cleaner approach, this is currently in development in combination with #22

[kirs]

This issue is supposed to be fixed in master now!