kicad-developers team mailing list archive

[John Beard <john.j.beard@xxxxxxxxx>]

On 29/04/2019 12:41, John Beard wrote:

> Then you might start running into consistency issues with multiple
> users of A_C getting different answers.

Notably, there is a specific case that illustrates where global state 
can cause Bad Things (TM) right now: unit tests.

Say you had a mutable ADVANCED_CFG, and you have a test suite like this:

    BOOST_AUTO_TEST_CASE( Test1 )
    {
        CONNECTIVITY_FOO thing;
        ADVANCED_CFG::GetCfg().m_whatever = true; // default false

        thing.Frobnicate(); // cares about m_whatever
    }

    BOOST_AUTO_TEST_CASE( Test2 )
    {
        // what is ADVANCED_CFG::GetCfg().m_whatever?
        // It depends on the test case ordering.
        // Test2 does not *have* to come second, and might even be 
concurrent
        // so AC::m_whatever could even change *during* this test

        CONNECTIVITY_FOO thing;
        thing.Frobnicate(); // what will this do?
    }

Also, if core functions depend on A_C, it's very hard to test them with 
good coverage. It's better to pass the A_C as a higher level parameter, 
so you can do this:

    BOOST_AUTO_TEST_CASE( TestWithFeatureOn )
    {
        bool enable_feature = true;
        CONNECTIVITY_FOO thing( enable_feature );
        thing.Frobnicate(); // test with feature
    }

    BOOST_AUTO_TEST_CASE( TestWithFeatureOff )
    {
        bool enable_feature = false;
        CONNECTIVITY_FOO thing( enable_feature );
        thing.Frobnicate(); // test without feature
    }

If CONNECTIVITY_FOO only used A_C internally, you cannot test the 
feature reliably as it depends on the config the user has set.

Cheers,

John