Conflict between system ruby and rvm1-capistrano3 installed ruby

[Rudikza]

Hi,

I am trying to use rvm1-capistrano3 on a Centos server which already has system ruby installed.
The system ruby is needed by the sysadmin for puppet and he has asked that I dont mess with it.

The problem is that most capistrano tasks use /usr/bin/env and when that happens it defauls to the system, ruby 2.0.0p598, instead of the specified version of jruby-9.0.1.0.

Here is the output from cap staging rvm1:check

DEBUG [e6ce521c] Command: /app/rvm1scripts/rvm-auto.sh rvm version
DEBUG [e6ce521c]  rvm 1.26.11 (latest) by Wayne E. Seguin <[email protected]>, Michal Papis <[email protected]> [https://rvm.io/]
DEBUG [e6ce521c] Finished in 0.586 seconds with exit status 0 (successful).
rvm 1.26.11 (latest) by Wayne E. Seguin <[email protected]>, Michal Papis <[email protected]> [https://rvm.io/]
DEBUG [16772925] Running /app/rvm1scripts/rvm-auto.sh rvm list as [email protected]
DEBUG [16772925] Command: /app/rvm1scripts/rvm-auto.sh rvm list
DEBUG [16772925]
DEBUG [16772925]  rvm rubies
DEBUG [16772925]
DEBUG [16772925]  =* jruby-9.0.1.0 [ x86_64 ]
DEBUG [16772925]
DEBUG [16772925]  # => - current
DEBUG [16772925]  # =* - current && default
DEBUG [16772925]  #  * - default
DEBUG [16772925]
DEBUG [16772925] Finished in 0.754 seconds with exit status 0 (successful).
rvm rubies

=* jruby-9.0.1.0 [ x86_64 ]

# => - current
# =* - current && default
#  * - default
DEBUG [3b561ebc] Running /app/rvm1scripts/rvm-auto.sh rvm current as [email protected]
DEBUG [3b561ebc] Command: /app/rvm1scripts/rvm-auto.sh rvm current
DEBUG [3b561ebc]  jruby-9.0.1.0
DEBUG [3b561ebc] Finished in 0.279 seconds with exit status 0 (successful).
jruby-9.0.1.0
DEBUG [56ed4fd2] Running /usr/bin/env if test ! -d /app/releases/20151019141431; then echo "Directory does not exist '/app/releases/20151019141431'" 1>&2; false; fi as [email protected]
DEBUG [56ed4fd2] Command: if test ! -d /app/releases/20151019141431; then echo "Directory does not exist '/app/releases/20151019141431'" 1>&2; false; fi
DEBUG [56ed4fd2] Finished in 0.042 seconds with exit status 0 (successful).
DEBUG [a0eb5962] Running /usr/bin/env ruby --version || true as [email protected]
DEBUG [a0eb5962] Command: cd /app/releases/20151019141431 && /usr/bin/env ruby --version || true
DEBUG [a0eb5962]  ruby 2.0.0p598 (2014-11-13) [x86_64-linux]
DEBUG [a0eb5962] Finished in 0.061 seconds with exit status 0 (successful).
ruby 2.0.0p598 (2014-11-13) [x86_64-linux]

The problem here is that /usrbin/env is using ruby 2.0.0p598 instead of jruby.

I have tried using 'rvm1:hook' but that doesn't seem to make a difference.

Is there some way to force all commands to use rvm-auto.sh instead of /usr/bin/env or am I holding this thing wrong?

[Rudikza]

Ok so the problem here is that rvm1-capistrano3 is not setting rvm1_map_bins correctly when used with capistrano-puma.

Capistrano-puma sets rvm_map_bins to ["puma", "pumactl'] and then rvm1-capistrano3 uses the following code to merge what ever is in rvm_map_bins with "rake", "gem", "bundle", "ruby".

set :rvm1_map_bins,   -> { fetch(:rvm_map_bins, %w{rake gem bundle ruby}) }

The problem is that it only returns back what is already in rvm_map_bins and does not append the new string elements in the array.

For example:

pry(main)> set(:rvm_map_bins, %w{peach apple})
=> ["peach", "apple"]
pry(main)> fetch(:rvm_map_bins, %w{rake gem bundle ruby})
=> ["peach", "apple"]

One way to do this would be to use the same method as capistrano-puma by doing a to_a.concat

pry(main)> set(:rvm_map_bins, %w{grape banana})
=> ["grape", "banana"]
pry(main)> fetch(:rvm_map_bins).to_a.concat(%w{ rake gem bundle ruby })
=> ["grape", "banana", "rake", "gem", "bundle", "ruby"]

So the new code in lib/rvm1/tasks/capistrano3.rake:41 would look like this:

set :rvm1_map_bins, -> { fetch(:rvm_map_bins).to_a.concat(%w{ rake gem bundle ruby }) }

I would be happy to write the code, just let me know if you are happy with this way of doing it?