I have built a node.js app that listens for webhooks. Currently it is used to build a jekyll website.

I have configured it on my server and jekyll build works perfectly when I run it in the root of my jekyll website (which is sending the hooks). When I run the node.js app over ssh in a shell everything works fine as well when a git hook is triggered.

However, when the node.js app is run from an upstart script (shown below) it doesn't seem to find the gems. It keeps asking for dependencies which I am sure I have installed (globally as well as for my user).

Inside the script I have put echo`which jekyll` and this shows that it is indeed pointing to the locally installed jekyll bin: /home/christophe/.gem/ruby/2.0.0/bin/jekyll. But right below that I execute the jekyll command and it fails:

/usr/lib/ruby/2.0.0/rubygems/dependency.rb:296:in `to_specs': Could not find 'jekyll' (>= 0) among 31 total gem(s) (Gem::LoadError)
 from /usr/lib/ruby/2.0.0/rubygems/dependency.rb:307:in `to_spec'
 from /usr/lib/ruby/2.0.0/rubygems/core_ext/kernel_gem.rb:47:in `gem'
 from /home/christophe/.gem/ruby/2.0.0/bin/jekyll:22:in `<main>'

How can I execute this bash script to properly execute jekyll?

Upstart

# /etc/init/libservice.conf
# Task to automatically start the library service.

author "Christophe De Troyer"
description "Run the githook for the blog."

# Path of the configuration files
env PROJ="/home/christophe/jekyll-builder"

# Configure to run as `christophe`
setuid christophe
setgid christophe

script
 export PATH=/home/christophe/.gem/ruby/2.0.0/bin:$PATH
 cd $PROJ
 gulp run
end script 

start on startup
#Respawn the process if it crashes
#If it respawns more than 10 times in 5 seconds stop
respawn limit 10 5

Build script

#!/bin/bash 

######################## 
# Parameters from Node # 
######################## 

giturl=$1 
reponame=$2 
branch=$3 
ownermail=$4 
reporoot=$5 
htmlsink=$6 
www=$7 

########## 
# Script # 
########## 

# Check to see if reponame exists. If not, git clone it
if [ ! -d $reporoot ]; then 
 mkdir -p $reporoot 
 git clone $giturl $reporoot 
fi 

# Checkout and pull branch. 
cd $reporoot 
git checkout $branch 
git pull origin $branch 
cd - 


echo `which jekyll` 
jekyll # fails 
# Run jekyll 
jekyll build -s $reporoot -d $htmlsink # fails too

Update:

gem env while logged in as a user:

 RubyGems Environment:
 - RUBYGEMS VERSION: 2.0.14
 - RUBY VERSION: 2.0.0 (2014-01-12 patchlevel 384) [x86_64-linux-gnu]
 - INSTALLATION DIRECTORY: /var/lib/gems/2.0.0
 - RUBY EXECUTABLE: /usr/bin/ruby2.0
 - EXECUTABLE DIRECTORY: /usr/local/bin
 - RUBYGEMS PLATFORMS:
 - ruby
 - x86_64-linux
 - GEM PATHS:
 - /var/lib/gems/2.0.0
 - /home/christophe/.gem/ruby/2.0.0
 - /usr/share/rubygems-integration/2.0.0
 - /usr/share/rubygems-integration/all
 - GEM CONFIGURATION:
 - :update_sources => true
 - :verbose => true
 - :backtrace => false
 - :bulk_threshold => 1000
 - REMOTE SOURCES:
 - https://rubygems.org/

gem env from within the script, executed from the node.js app running via upstart gives:

RubyGems Environment:
 - RUBYGEMS VERSION: 2.0.14
 - RUBY VERSION: 2.0.0 (2014-01-12 patchlevel 384) [x86_64-linux-gnu]
 - INSTALLATION DIRECTORY: /var/lib/gems/2.0.0
 - RUBY EXECUTABLE: /usr/bin/ruby2.0
 - EXECUTABLE DIRECTORY: /usr/local/bin
 - RUBYGEMS PLATFORMS:
 - ruby
 - x86_64-linux
 - GEM PATHS:
 - /var/lib/gems/2.0.0
 - /.gem/ruby/2.0.0
 - /usr/share/rubygems-integration/2.0.0
 - /usr/share/rubygems-integration/all
 - GEM CONFIGURATION:
 - :update_sources => true
 - :verbose => true
 - :backtrace => false
 - :bulk_threshold => 1000
 - REMOTE SOURCES:
 - https://rubygems.org/

Notice that the GEM_PATHS is missing the home directory prefix in the second entry. I have tried resolving this by putting env GEM_PATH="/home/christophe/.gem/ruby/2.0.0" in the upstart script but that didnt change anything.

In the meanwhile I have solved it by intalling a list of deps manually as root. However, I don't think this is a good approach as the upstart is explicitly running as my user. And secondly, this software needs to run on a server I don't have root permissions on. So I would still like to know a fix.

sudo gem install jekyll
sudo gem install jekyll-gist
sudo gem install jekyll-cite
sudo gem install jekyll-scholar
sudo gem install addressable -v 2.3.5
sudo gem install yajl-ruby -v 1.2.0
sudo gem install pygments.rb
sudo gem install posix-spawn

Did you try setting the GEM_HOME environment? I was able to at least get an upstart service listing a custom install dir via something like:

cd
gem install rake --install-dir grepable

And then using a /etc/init/footest.conf service of:

author "Nobody"
description "Echo some details"

env GEM_HOME="/home/jdoe/grepable"

setuid jdoe
setgid jdoe

script
 /usr/bin/gem environment
end script 

start on startup
respawn limit 10 5

And after a reboot,

# fgrep -1 grep /var/log/upstart/footest.log
 - RUBY VERSION: 1.9.3 (2013-11-22 patchlevel 484) [x86_64-linux]
 - INSTALLATION DIRECTORY: /home/jdoe/grepable
 - RUBY EXECUTABLE: /usr/bin/ruby1.9.1
 - EXECUTABLE DIRECTORY: /home/jdoe/grepable/bin
 - RUBYGEMS PLATFORMS:
--
 - GEM PATHS:
 - /home/jdoe/grepable
 - /.gem/ruby/1.9.1

For a production service one presumably would use an --install-dir/GEM_HOME of somewhere suitable not in the vendor space nor your home directory.

After a few hours of agony I have finally fixed it.

The trick was to set the PATH to: /home/cdetroye/.rbenv/shims:/home/cdetroye/.rbenv/bin:/usr/local/bin:/usr/bin:/bin.

# /etc/init/libservice.conf
# Task to automatically start the library service.

author "Christophe De Troyer"
description "Run the githook for the blog."

# Path of the configuration files

env PATH=/home/cdetroye/.rbenv/shims:/home/cdetroye/.rbenv/bin:/usr/local/bin:/usr/bin:/bin

# Configure to run as `christophe`
setuid cdetroye
setgid cdetroye

script
 cd $PROJ
 gulp run
end script 

start on startup
#Respawn the process if it crashes
#If it respawns more than 10 times in 5 seconds stop
respawn limit 10 5