[sv-bc] RE: Mantis 1111, omitting range on port declaration

Actually, I think that today many tools do not allow one to specify
different ranges for the port and net/variable declaration. Allowing
that is a super-gotcha.

The question is about omitting the range entirely on the port
declaration. Even today, some tools do not allow this. The question came
up because some people interpret the LRM as implicitly allowing that. It
was not the intention of the LRM text to allow that.

In my mind, allowing that is also a gotcha. It means you cannot even be
sure of the bit range by looking at the port declaration. 

It is true that many tools still allow it, but the proposal would not
really change that, as all tools have their own extensions to the
language. 

The proposal is intended to make explicit the intent of previous
versions of the LRM, not change that intention, in order to avoid
misinterpretation. It would mean that a tool is not required to support
that case, and that a person and a tool should not produce such code as
it is not guaranteed to be portable. It still would not really prevent
tools from accepting that in order to be backward compatible. It should
also be noted that no Verilog LRM ever explicitly permitted that
behavior.

Shalom


> > SVDB 1111 ___Yes   _X__No
> > http://www.eda.org/svdb/bug_view_page.php?bug_id=1111
> > I think this will cause a lot of backwards compatibility
> > issues. Verilog-XL did not check this, and other tools
> > have been forced to follow Verilog-XL. I think the
> > standard should reflect what most tools actually
> > have implemented.
> >
> 
> If my failing memory serves me correctly, I recall discussing this
> when we
> created Verilog-1995.  Everyone, including the company that owned
> Verilog-XL, agreed that that product had a bug, and that the standard
> show
> require that the port size and net/variable size should match.  Thus,
> we
> specified the sensible rule in the standard, rather than basing the
> standard
> on one tool's bug.  But Mark is right.  Twelve years later, and now
> that bug
> is the de facto standard.  It would be fully backward compatible to
> change
> the standard to say what tools actually do.  Specifically, if the port
> packed range and the net/variable packed range differ, the
> net/variable
> packed range is used.

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.