Changes are sent to Travis CI for integration testing after pushes, and you can see the current test status here.
The testing configuration is controlled by two files:
travis.yml
and run_tests.sh
.
This file defines the prerequisites required for testing, and the list of directories to be tested. (The directories listed are tested in parallel.)
The same script (run_tests.sh) is executed for all directories.
This file is run once for each directory defined in
.travis.yml
. The directory name being tested is passed in the environment
variable TEST_DIR
.)
run_tests.sh
checks the value of TEST_DIR
, and then runs
commands to test the relevant implementation. The commands that do the testing
must return zero on success and non-zero value on failure. Tests that
return zero, even if they output error messages, will be considered by the
testing framework as a success.
Add your directory to the .travis.yml
file:
# Define the list of directories to execute tests in.
env:
- TEST_DIR=js
- TEST_DIR=go
- TEST_DIR=your directory goes here
Then add the necessary code to run_tests.sh
:
# Your language goes here
if [ "$TEST_DIR" == "your directory goes here" ]; then
cd directory && run something && run another thing
exit # Exit immediately, returning the last command status
fi
Note the use of &&
to combine the test commands. This ensures that if any
command in the sequence fails, the script will stop and return a test failure.
The test commands must be followed by an exit
statement. This ensures that
the script will return the same status as the tests. If this status is zero,
the test will be marked successful. If not, the test will be marked as a
failure.
Travis CI assumes that each github project has only
a single language. That language is specified in the .travis.yml
file (language: node_js
).
This shouldn't be a problem, since prerequisites can still be loaded in the
before_script
section, and then commands executed in
run_tests.sh
. However in the event that you can't resolve a
problem, leave a comment in the issue or push request and we'll see if someone
can figure out a solution.