Re: [Boost-users] Boost Test usage

From: Gennadiy Rozental (gennadiy.rozental_at_[hidden])
Date: 2006-07-14 10:20:36

"Matt Schuckmann" <matt_at_[hidden]> wrote in message 
news:44B6DA44.8020305_at_schuckmannacres.com...
> I've been looking into using the boost test library to start testing
> some of our code and it looks pretty good. I've gone through the
> documentation and several of the examples and there are a couple of
> things I don't really get

So I  gather you are using boost 1.33.1, right?

> I don't see how the test_suite class lets you create do anything special
> with regards to building multi-level tests, none of the examples really
> show multilevel test hierarchy and how it really helps me.

There is a unit_test_example3 that builds multilevel hierarhy.

> It looks like
> it's really just a way of using a class to group test functions instead
> of a series of free functions, am I missing something here?

test_suite is subclass of test_case. You could use method "add" to build 
hierahies.

> I really expected more from a feature like this, I'm thinking of things 
> like
>    Command line control over which test suites to run, I mean you name
> the test suites but you don't appear to do anything with that name

Run by name is planned to be implemented in 1.35

>    Reporting success and failure grouped by test suite

You could use --report_level=detailed for that

> Am I missing something here, is this stuff possible and the examples
> don't show it or are these feature that are yet to come or haven't been
> thought of?
>
> I don't see a good example for how I should setup a unit test for a real
> world type project you know an extensive example that covers inputs,
> outputs, multiple test suites, etc all in one example.

Take a look how Boost.Test internal test are build. I personally rarely used 
multilevel hierarhies, but that minor feature IMO unless you are into 
hardcore TDD..

> In my case I've a got a library with several independent classes and
> functions that I'd like to test and I'm not sure if I should be
> creating  multiple executables (1 for each test case) or a single
> executable that tests everything, I don't understand the pros and cons
> of either case. It seems like I want the single executable that can test
> everything or that can test just want I tell it to test.

It's really up to you. There several guidelines on how to split your unit 
tests by test modules, but iun the end it just your preference:
* each program unit (possible single class or several cooperating) has 
separate test module
* performace test goes in separate module
* feature that may cuase compile time failure should go in separate module
* Each feature goes in separate test case: construction/destruction/access 
methods/printing etc

> Any info would be appreciated.

Also I really recomment consider using auto registration facilities, 
especially if you could use latest cvs version of Boost.Test.

> Matt S.

Gennadiy