Re: Tests in a software release

["Randy Brukardt" <randy@rrsoftware.com>]

"Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> wrote in message 
news:ouguv4$1h92$1@gioia.aioe.org...
> On 15/11/2017 01:06, Randy Brukardt wrote:
>
>> I'd consider making an argument for a pragma to *require* an ELSE part 
>> for
>> all IF statements.
>
> Only if return and raise statements had the forms:
>
>    return Result when Condition;
>
>    raise Exception with Text when Condition;

You're right that sometimes the comment isn't very enlightening:

     if Condition then
         return Result;
     -- else nothing needed.
     end if;

...but note that often in such cases it is useful to say how the result will 
be determined if Condition is False:

    -- else continue looking for the result.

which does provide useful information to a reader (especially one that 
hasn't looked at a particular subprogram in a decade, something that happens 
a lot when debugging a long-lived program like Janus/Ada).

The place where this comment is annoying is for tracing:

    if J2Trace.Trace(This_Unit) then
        Put_Line ("Type " & Type_Ptr'Image(The_Type) & " is unconstrained");
    -- else no trace needed.
    end if;

Here, the else comment is noise. I personally usually leave it out (having 
an exception in the style guide for trace messages), but others (especially 
Isaac) always followed the rule religiously, so Janus/Ada is full of this 
comment.

                              Randy.