Re: [sv-bc] Issue 548: .* and implicit nets

Steven,

    If you are changing text there, would you consider making the "types match"
rule refer explicitly to either section 5.9.1 (my favorite), or to one of
the weaker sections that follow, so there is less ambiguity about which
connections must be implicitly produced.  I favor 5.9.1 because I think
the compiler should not assume that a packed struct and a packed array with
the same name and width implicitly connect.  Granted that in practice they probably
should connect, but I'd like the user to intervene in some way to reassure
me that this was an intentional type divergence, not accidental.

   I also have a data-types on nets question: Under port connections (18.12?)
we're told that Assignment compatibility rules apply.  But for full SV data
types those rules are asymmetric.  So for an inout net with, say, enum data
type, we mean to require assignment compatibility in BOTH directions, right?

Greg


Steven Sharp wrote:
> Another difference between .* and a list of implicit .name connections is
> that .* will not implicitly declare nets.  Implicit net declarations only
> occur upon finding an undeclared identifier in a port list.  With .*, there
> are no identifiers in the port list.  This is much like the import issue.
> 
> I will try to modify Dave's proposal for this.
> 
> Steven Sharp
> sharp@cadence.com
> 
>