In shell #1:
$ cd $MONGOOSEIM $ ./rebar3 compile $ make devrel
In shell #2:
$ cd $MONGOOSEIM/_build/mim1/rel $ ./bin/mongooseimctl live
In shell #3:
$ cd $MONGOOSEIM/_build/mim2/rel $ ./bin/mongooseimctl live
In shell #4:
$ cd $MONGOOSEIM/_build/mim3/rel $ ./bin/mongooseimctl live
In shell #5:
$ cd $MONGOOSEIM/_build/fed1/rel $ ./bin/mongooseimctl live
Back to shell #1:
$ cd test.disabled/ejabberd_tests $ make quicktest
Wait for the tests to finish and celebrate in joy (or despair in grief)!
Step by step breakdown
make devrel builds required server nodes:
These are preconfigured for breadth of features and compatible
with as many test suites as possible.
There are other four of them:
$MONGOOSEIM/_build/mim1/rel, for most test SUITEs
$MONGOOSEIM/_build/mim*/rel, in order to test cluster-related
$MONGOOSEIM/_build/fed1/rel, in order to test XMPP federation
(server to server communication, S2S).
In general, running a server in interactive mode (i.e.
live) is not required to test it, but it's convenient as any warnings and
errors can be spotted in real time.
It's also easy to inspect server state or trace execution (e.g. using
in case of anything going wrong in some of the tests.
To run server in the background instead of interactive mode,
mongooseimctl start && mongooseimctl started.
quicktest configuration is a relatively comprehensive one,
giving good overview of what does and what doesn't work in the system,
without repeating tests.
Why would we want to ever repeat tests?
In order to test different backends of the same parts of the system.
E.g. a message archive might store messages in MySQL/PostgreSQL or Riak KV
- the glue code between the XMPP logic module and database is different in each case,
therefore repeating the same tests with different databases is necessary
to guarantee a truthful code coverage measurement.
Testing a feature in development / TDD
The whole suite takes a significant amount of time to complete. When you focus on a new feature, fast iteration speed is crucial to maintain the flow (who doesn't like the feeling?!) and not lose focus.
$MONGOOSEIM/test.disabled/ejabberd_tests/ we have:
$ tree test/ejabberd_tests/ -L 1 -F test/ejabberd_tests/ ├── Makefile ├── README.md ├── default.spec ├── test.config ├── tests/ └── ...
tests/ is where the test suites reside.
*.config files are suite configuration files - they contain predefined
XMPP client specifications, server addresses and XMPP domains to use
and options required by test support libraries
*.spec files are test specifications - they define the configuration
file to use, the suites, test groups or individual test cases to run
or skip and some less important things.
default.spec is, well, the default when running
but it can be overridden with
# make sure we're in $MONGOOSEIM/test.disabled/ejabberd_tests/ cd $MONGOOSEIM/test/ejabberd_tests/ make quicktest TESTSPEC=my-feature.spec
It's customary to create a per-feature (or per-project, if you're cloning
.spec file and only enable
the suites / test groups you want to test - this speeds up the iteration
cycle by not testing parts of the system that you know have not changed.
It's worth running
default.spec once in a while to check for
Have a look into
default.spec file to see how to pick only the
interesting tests to run.
If you're sure that none of the test dependencies have changed
and you only edited the test suites, it's possible to speed up
the test run even a bit more by skipping Rebar dependency and compilation
checks by providing
PREPARE= (i.e. an empty value):
make quicktest PREPARE=
Have a look inside the
Makefile to see how it works.
Reloading node(s) code
When working on a feature or a bug fix often you modify the code and check if it works as expected.
In order to change the code on already generated dev nodes (
fed*) it's enough to
recompil the code for specific node.
Let's say you need to update the code on
mim1 node, all you have to do is:
./rebar3 as mim1 compile
Similar command applies to other nodes, the important thing here is rebar3's profile.
When the above command finished, the code can be reloaded on the server by:
1. loading new module(s) in the node's shell, f.e.
1. restarting the node