Re: Is Aunit helpful?

[Simon Wright <simon@pushface.org>]

Midoan <midoan.ses@gmail.com> writes:

> On Aug 14, 10:38 pm, Yannick Duchêne (Hibou57)
> <yannick_duch...@yahoo.fr> wrote:
>> Le Sat, 14 Aug 2010 23:33:32 +0200, Midoan <midoan....@gmail.com> a
écrit:
>> > This can be done to achieve branch, decision or MC/DC coverage as
>> > desired.
>>
>> What is MC/DC please ?
>
> MC/DC stands for Modified Condition/Decision Coverage. Informally, it
> forces deep testing of conditional expressions as in "IF A and B or C
> THEN ..." where several tests would be required for the "A and B or C"
> expression whereas branch and decision testing would only require two
> (one for a True outcome and one for a False outcome). It is a more
> thorough, if somewhat badly defined, coverage criterion often required
> for high integrity systems.

Reverting to the use of gcov: even if you got full gcov coverage, that
wouldn't tell you whether the False outcome had actually
occurred. (well, there are options to check for that -- but, as far as I
can see on a quick check, they aren't very helpful, because they report
on the branches in the object code, which includes (for instance) all
the range checks that always succeed in a well-written Ada program!).

Most often, tools such as Midoan is discussing will involve
automatically instrumenting the source code, so that they can report at
the appropriate (source) level.