Re: [Spark] How to terminate information flow?

[Lutz Donnerhacke <lutz@iks-jena.de>]

* Peter Amey wrote:
> The short answer is that you can't make those messages go away because
> they are true!  We do not recommend slavishly trying to eliminate all
> flow errors as long as the reason for them is understood.  The Examiner
> is telling you what is _really_ happening - you have to interpret that
> information.

Ok, understood.

> What we do in these cases is:
>
> (1) Choose a good naem for the unused parameters.  e.g. UnusedRes and
>     UnusedOk.  You then get a flow error "assignment to unused is
>     ineffective" which is self-documenting.

Done.

> (2) Add a comment to line 29 saying "expect ineffective .... because ...".
>     Which maintain a justifation along with the code.

I add a comment type (--! regex) which can be parsed automatically, in order
to suppress the examiner output on this cases by a wrapper script. This
wrapper script is even necessary here to provide a editor friendly output.
I'll put the wrapper script at
  ftp://ftp.iks-jena.de/pub/mitarb/lutz/ada/spark/spark.pl
in a few hours.

> I do have a tentative plan for a a "justify" annotation which can be
> placed at the point where any flow error is produced and which will
> suppress it at that point but add it to a summary file of justified
> errors.

That would be fine, but hat very low priority. A practial workaround exists.