Running Travis CI Locally

Dec. 18, 2017, 5:02 a.m. (running-travis-ci-locally)

TL;DR


I accidentally pushed a change to kilonull that added the word "test" in the site's title tag. I pushed another commit to remove the extra text and pulled on my server. About 5 minutes later I received an email telling me my build on Travis had failed.

# ...snip...
Error: could not determine PostgreSQL version from '10.1'

    ----------------------------------------
Command "python setup.py egg_info" failed with error code 1 in /tmp/pip-build-wq2uqxzp/psycopg2/

The command "pip install -r requirements.txt" failed and exited with 1 during .

Your build has been stopped.

Well at least my code didn't break anything. But hey, it's a Sunday and I have chores to ignore. Let's look into this. I googled the error and stumbled upon a comment on Github stating that the fix was to update the psycopg2 requirement to 2.7.1 (the current latest version). Great, that should be an easy fix. But hang on, I have all these chores to ignore. I can probably run Travis locally before pushing just to verify the fix. Let's look into this.

Someone else (Ibrahim Ulukaya) was kind of enough to document how to run Travis tests from a Docker container. I followed his instructions but I kept getting an error about being unable to load travis/support when I ran travis compile.

$ travis compile
/home/travis/.rvm/rubies/ruby-2.4.3/lib/ruby/2.4.0/rubygems/core_ext/kernel_require.rb:120:in 'require': cannot load such file -- travis/support (LoadError)
    from /home/travis/.rvm/rubies/ruby-2.4.3/lib/ruby/2.4.0/rubygems/core_ext/kernel_require.rb:120:in 'require'
    from /home/travis/.travis/travis-build/lib/travis/build.rb:1:in '<top (required)>'
    from /home/travis/.rvm/rubies/ruby-2.4.3/lib/ruby/2.4.0/rubygems/core_ext/kernel_require.rb:120:in 'require'
    from /home/travis/.rvm/rubies/ruby-2.4.3/lib/ruby/2.4.0/rubygems/core_ext/kernel_require.rb:120:in 'require'
    from /home/travis/.travis/travis-build/init.rb:11:in 'setup'
    from /home/travis/.rvm/gems/ruby-2.4.3/gems/travis-1.8.8/lib/travis/cli/command.rb:197:in 'execute'
    from /home/travis/.rvm/gems/ruby-2.4.3/gems/travis-1.8.8/lib/travis/cli.rb:64:in 'run'
    from /home/travis/.rvm/gems/ruby-2.4.3/gems/travis-1.8.8/bin/travis:18:in '<top (required)>'
    from /home/travis/.rvm/gems/ruby-2.4.3/bin/travis:23:in 'load'
    from /home/travis/.rvm/gems/ruby-2.4.3/bin/travis:23:in '<main>'
    from /home/travis/.rvm/gems/ruby-2.4.3/bin/ruby_executable_hooks:15:in 'eval'
    from /home/travis/.rvm/gems/ruby-2.4.3/bin/ruby_executable_hooks:15:in '<main>'

I found another useful comment on Github that suggested specifying the path to the travis script from travis-support. Maybe I did something wrong when I followed the instructions on Medium but this was working for me.

Here are the steps that worked for me. I hope this is useful for someone else someday.

First off, we'll need to decide on one of Travis's docker containers to run from. Available containers are listed on Quay. We'll want one of the containers named travis-<some language>. I copy-pasted from the instructions in the Medium article so I ended up running everything under the travis-jvm container. In retrospect, I should have used travis-python since I was dealing with a Python project. The command docker run -it -u travis quay.io/travisci/travis-jvm /bin/bash can be used to run the container (replace travis-jvm with whatever container is desired).

Before setting up travis-build we can choose which version of Ruby to work with. The latest stable version was 2.4.3 when I checked so I decided to go with that.

rvm install 2.4.3
rvm use 2.4.3

Once Ruby is set up to our liking we can set up travis-build:

# We'll clone travis-build inside of ~/.travis for convenience.
cd ~/.travis
git clone https://github.com/travis-ci/travis-build.git
cd travis-build
bundle install
bundler add travis
bundler binstubs travis

We'll use travis-build to generate a shell script for our build. The generated shell script respects the configurations defined in .travis.yml. If your project is online somewhere like Github or GitLab but not public, you'll need to generate an SSH key in order to clone your repo. For Github/GitLab, I'd recommend adding a deploy key to your project rather than adding a key to your account.

Now that we have our code in our container we can generate the shell script for our build. Navigate to your project and watch Travis run its magic

.~/.travis/travis-build/bin/travis compile > ci.sh
# We need to make the script executable
chmod u+x ci.sh

Great now we have the script for our build we can go ahead and run it

$ ./ci.sh
# ...snip...
$ git clone --depth=50 --branch='' https://github.com/vacuus/kilonull.git vacuus/kilonull
Cloning into 'vacuus/kilonull'...
warning: Could not find remote branch  to clone.
fatal: Remote branch  not found in upstream origin
Unexpected end of command stream

The command "eval git clone --depth=50 --branch='' https://github.com/vacuus/kilonull.git vacuus/kilonull " failed. Retrying, 2 of 3.

Cloning into 'vacuus/kilonull'...
warning: Could not find remote branch  to clone.
fatal: Remote branch  not found in upstream origin
Unexpected end of command stream

The command "eval git clone --depth=50 --branch='' https://github.com/vacuus/kilonull.git vacuus/kilonull " failed. Retrying, 3 of 3.

Cloning into 'vacuus/kilonull'...
warning: Could not find remote branch  to clone.
fatal: Remote branch  not found in upstream origin
Unexpected end of command stream

The command "eval git clone --depth=50 --branch='' https://github.com/vacuus/kilonull.git vacuus/kilonull " failed 3 times.


Failed to clone from GitHub.
Checking GitHub status (https://status.github.com/api/last-message.json):
good
Everything operating normally.
2017-11-28T00:03:47Z

Unfortunately, it didn't work for me. Removing the branch flag from the clone command fixed it for me though. I suspect the actual Travis CI service inserts the branch name based on what branch contained the commit triggering the build. We're doing this manually though so we'll have to make a small adjustment to our build script. Go ahead and open it up in your favorite text editor (vim). Find the definition for the travis_run_checkout function. Remove the branch flag (or specify a branch name if you want to pull from something other than master) from the clone command inside the if block.

Run the ci.sh script with our modification again and you should be able to successfully clone your project and continue with the rest of the build process. This is nice and all but the whole point in running Travis locally for me was so I could test changes without having to make a commit. But travis_run_checkout is called every time our build script executes. We can make another modification so we can test local changes without committing. Open ci.sh up again and go back to the definition of the travis_run_checkout function. Before the travis_fold end git.checkout line there will be a cd command that tells the build script to go to the location we pulled our repository to. Copy this line then go to the bottom of the script. Comment out the call to travis_run_checkout and paste the cd command on the next line. Your hacked build script should look something like this:

EOFUNC_FINISH
# END_FUNCS
source $HOME/.travis/job_stages
travis_run_setup_filter
travis_run_configure
#travis_run_checkout
travis_cmd cd\ vacuus/kilonull --echo
travis_run_prepare
travis_run_disable_sudo
travis_run_export
travis_run_setup
travis_run_setup_casher
travis_run_setup_cache
travis_run_announce
travis_run_debug
travis_run_before_install
travis_run_install
travis_run_before_script
travis_run_script
travis_run_before_cache
travis_run_cache
travis_run_after_success
travis_run_after_failure
travis_run_after_script
travis_run_finish
echo -e "\nDone. Your build exited with $TRAVIS_TEST_RESULT."

travis_terminate $TRAVIS_TEST_RESULT

Now we have two copies of our repo in the container which might get a bit confusing. We can remove the repo at the location that we manually cloned (not the one cloned as part of the build script inside ~/build). Just remember to move the ci.sh script before deleting your project's directory. We can now make changes to the copy of the repo inside the ~/build directory and run our hacked build script to test any changes. I updated my requirements.txt's version for psycopg2 and my build succeeded just as I had hoped :) .

This process is pretty convoluted but I think I can automate this and include it a container for my project. But, maybe I'm better off using something like Jenkins for CI if I'm so concerned with running my builds locally. At least I can feel like I did something productive while avoiding my chores.