rebar ct replacement

Why "topcat"?

Well, it replaces "rebar ct". "ct" reversed is "tc". "T.C." is Top Cat.

What does it give me?

Progress reporting

It prints out what it's doing as it goes along, so you can better interpret your test output and results.

Test summaries

It prints out a nice test summary and list of skipped/failing tests after all your applications have been run.

This means you won't miss a failing test when the output scrolls past, making it easier to see what's working and not working.

Working code coverage

It's careful about ensuring that paths are fully-qualified whenever possible. This means that, in particular, code coverage actually works.

Including it with rebar.config

Add the following dependency to your top-level rebar.config:

{topcat, ".*",
  {git, "git://github.com/rlipscombe/topcat.git", {tag, "0.9.1"}}}

rebar compile will build it for you.

Then, where you would have run rebar ct, use deps/topcat/topcat.

Usage

Run everything

./deps/topcat/topcat

Runs all suites, groups and tests in all applications.

Specified application

./deps/topcat/topcat -app my_app

Runs all suites, groups and tests from the specified application. You can specify multiple applications if you prefer.

Specified test cases

Note that you have to specify the suite as well:

./deps/topcat/topcat -app my_app -suite the_suite -case a_test another_test

Miscellanea

What's the ".topcat" directory for?

Because of the way that topcat injects itself into the Erlang node that's running your Common Test suites, it needs to provide a set of .beam files to the Erlang runtime.

These are packaged into the top-level topcat escript, but need to be extracted for the Erlang runtime to find them later.

They go into the .topcat folder. You should add it to your .gitignore file and to your make clean target. Other than that, feel free to ignore it.

Why not just use rebar's escriptize command?

rebar has an escriptize command, which creates a stand-alone escript executable from your source. It does this by packaging your BEAM files into a ZIP archive, and then jamming a simple escript header on the front.

Why not use that, instead of unpacking to the ".topcat" directory?

Because, while the Erlang loader supports loading code from ZIP archives, and while escript will load code from appended ZIP archives, the Erlang loader won't support loading code from ZIP archives appended to escript.

And we need that, so that the modules can be loaded into the topcat_slave node.

This means that the BEAM files have to be extracted somewhere temporarily.

I guess that I could have opted to just write out the ZIP file, rather than extract the contents, though...

Licensing

Apache 2.0