CFE_EVS_SendEvent stub should provide full context

[gilliganeric]

Describe the bug
The CFE_EVS_SendEvent stub for ut-assert does not provide its full context. It only provides the first argument, EventId, providing no way to for unit tests to verify the correct EventType or event message text was sent.

To Reproduce
Steps to reproduce the behavior:

1. See the CFE_EVS_SendEvent stub. Only EventID is registered with the context and copied.

Expected behavior
The CFE_EVS_SendEvent stub should provide its full context for unit testing purposes, including the EventID, EventType, and the event message string. My preference would be that the event message text provided would not include the format specifiers and instead be the resulting string with the format specifiers replaced with the appropriate text. I've been convinced and also talked myself out of that preference.

Code snips

int32 CFE_EVS_SendEvent(uint16 EventID,
                        uint16 EventType,
                        const char *Spec,
                        ...)
{
    int32 status;

    UT_Stub_RegisterContext(UT_KEY(CFE_EVS_SendEvent), &EventID);
    status = UT_DEFAULT_IMPL(CFE_EVS_SendEvent);

    if (status >= 0)
    {
        UT_Stub_CopyFromLocal(UT_KEY(CFE_EVS_SendEvent), (uint8*)&EventID, sizeof(EventID));
    }

    return status;
}

System observed on:

• cFE 6.7.11

Additional context
n/a

Reporter Info
Eric Gilligan NASA/GSFC-5820

[gilliganeric]

Thinking about this more, I'd also be fine with the Context getting the string as passed into CFE_EVS_SendEvent with the format specifiers still in it, as long as the extra arguments are put into the Context as well.

[skliper]

I think @jphickey has examples for getting format and arguments.

[jphickey]

Yes, this stub will need an update to pass through more context if UT test cases need this.

The framework has the facility to do this, via the UT_DefaultStubImplWithArgs routine, which can pass through the full context in a va_list to a hook implementation. Then in the hook implementation one can do whatever is needed.

I'd still recommend against passing it to [v]snprintf, as that is now crossing the bridge of no longer validating the direct outputs of the unit, but rather mimicking/duplicating something that the FSW might do and validating that. Very different thing. (stubs shouldn't duplicate FSW logic).

[gilliganeric]

I agree. Disregard that preference in the original description.

Would this look like ArgPtr's last element in the hook function's UT_StubContext_t being a va_list containing the variable number of arguments?

[asgibson]

I completely agree with not passing it to [v]snprintf, but I would like to see the default implementation of the CFE_EVS_SendEvent stub make the arguments available for checking without having to use a hook.

[jphickey]

The "Fixed" arguments, up to and including the spec string, can be registered using calls to UT_Stub_RegisterContext().

The use va_start to capture all the remaining variable arguments, and invoke UT_DefaultStubImplWithArgs with va list as the final argument, then do va_end after it returns.

For an example of a stub that uses this mechanism, see OS_printf in: https://github.com/nasa/osal/blob/master/src/ut-stubs/osapi-utstub-printf.c

Usage with hook function example is in sample_lib: https://github.com/nasa/sample_lib/blob/master/unit-test/coveragetest/coveragetest_sample_lib.c

[skliper]

CCB 20200415 - Ok to add and pass it through

[skliper]

@gilliganeric - can you confirm #673 does what you need, or clarify what would still be necessary?

[ericgilligan-nasa]

It looks like #673 provides the Event Type and Spec, but not the variable arguments as a va_list as described by @jphickey earlier. Is that correct?

[skliper]

Correct, we can leave this open until that's implemented:

The use va_start to capture all the remaining variable arguments, and invoke UT_DefaultStubImplWithArgs with va list as the final argument, then do va_end after it returns.

For an example of a stub that uses this mechanism, see OS_printf in: https://github.com/nasa/osal/blob/master/src/ut-stubs/osapi-utstub-printf.c

Usage with hook function example is in sample_lib: https://github.com/nasa/sample_lib/blob/master/unit-test/coveragetest/coveragetest_sample_lib.c

[jphickey]

Ahh yes, sorry I overlooked that during my review of the change. If you want to have the actual data available in your hook this needs to use the UT_DefaultStubImplWithArgs() function to pass them all through.