Progressive Coverage is a method of coverage collection that’s highly applicable in communication devices, but may as well be applied elsewhere. Now before I start to babble about my philosophy on coverage collection why don't I just give you an example?
Let’s recall our beloved struct from a recent post - packet - and make it a bit more interesting:
Our sample packet can now have various types of errors. First off, it may be invalid (as indicated by the packet_valid field). If the frame is valid, it might have a corrupt header (header_crc_ok == FALSE), in which case it would be impossible to parse the payload because whoever receives this packet doesn’t know what kind of packet it is. But that’s not all folks, the plot thickens. Even if the header is ok, the payload itself might still contain errors, in which case the payload_crc_ok would be FALSE. Got it?
Ok, so now we want to write a reference model for the Receiver part which - how surprisingly - receives packets of that particular type, tries to parse them and then do something with them - not so important. What I’d like to focus on today is the coverage collection within the reference model.
So let’s get started!
Roughly, the reference model will look something like this:
So this is the skeleton of the reference model, and the important thing to notice is that the nature of the processing is progressive - you first process the packet as a whole, then the header, and last - the payload. And now ladies and gentlemen, we must add some coverage. We’ll make it really simple and assume we want coverage on the header value, and the payload size.
So now we cover all we want upon receipt of each packet in the reference model. But is that all? Of course not. What’s the point of covering the payload size if the header is invalid? So let’s use the WHEN command to make sure we only cover items when relevant.
Now the coverage will only be collected when relevant. Great! we’re finished with the introduction. Phew
In modern network devices we might bump into packets or frames that are a bit more complicated than this one, and may have more than one layer of information in them. Try to imagine how easy it would be to come up with the appropriate WHEN expression. It might get really ugly guys. But, and there’s always a but, the solution is just a minute away.
Time to unveil the curtain - Progressive Coverage is all about avoiding WHEN commands, and instead - using multiple coverage events. So in our sample reference model we’ll do this:
So we got rid of the WHENs, and created a sequence of coverage events, that either occur or not, according to the parsing progress - hence the term Progressive Coverage. The nice thing about it is that in large systems it really makes life easier to deal only with the coverage items that are relevant to the current phase. The alternative - complicated WHEN expressions - is not only harder to write, but also more error-prone so that you might end up with missing coverage buckets, or God forbid - false ones.