RE: [sv-champions] Champions conference call May 21st, 2009 -- 2735

Dave,

In the following BNF erratum/proposal from 2003

      http://www.eda-twiki.org/sv-ec/hm/1519.html

the first bug I described was exactly a lack of method call chaining

     "Only accepts method calls for identifiers, instead
      of for any expression that yields a type that allows
      a method call."

That initial fix, early in Accellera 3.1a, was then built on by

    http://www.eda-twiki.org/sv-ec/hm/1623.html
    http://www.eda-twiki.org/sv-ec/hm/1643.html
    http://www.eda-twiki.org/sv-ec/hm/1715.html

and most recently, during work on 1800-2009, by

    http://www.eda-stds.org/svdb/view.php?id=1480

which was discussed here

      http://www.eda-twiki.org/sv-ec/Minutes/SV-EC_Meeting_September_11_2006_Minutes.txt
      http://www.eda-twiki.org/sv-ec/Minutes/SV-EC_Meeting_October_9_2006_Minutes.txt

Because the BNF has long accepted method call chaining and there seems to be no prohibition on it in the text, I believe it's allowed unless the committee adds an explicit deprecation.  The proposed resolution is not accurate.

For completeness, here's an example BNF derivation of obj.method1(5).method2(6)

       expression ::= primary
                        ::= function_subroutine_call
                        ::= subroutine_call
                        ::= method_call
                        ::= method_call_root . method_call_body
                        ::= primary . method_identifier ( list_of_arguments )
                        ::= function_subroutine_call . method2(6)
                        ::= subroutine_call . method2(6)
                        ::= method_call . method2(6)
                        ::= method_call_root . method_call_body . method2(6)
                        ::= primary . method_identifier ( list_of_arguments ) . method2(6)
                        ::= obj . method1(5) . method2(6)

-- Brad


________________________________________
From: Rich, Dave [Dave_Rich@mentor.com]
Sent: Wednesday, May 20, 2009 3:55 PM
To: Brad Pierce; sv-champions@eda.org
Subject: RE: [sv-champions] Champions conference call May 21st, 2009 -- 2735

Brad,

The example in 8.11 has (relatively) well defined namespace search
rules, because after the first name, everything is a field or property
select.

What happens if you allow method calls, and that name is not found
locally? Do normal Verilog hierarchical search rules apply? In the
middle of dotted name? What about static variable names inside
functions?

There are not very hard problems to solve, but just not in the timeframe
we have for the ballot.

Dave


> -----Original Message-----
> From: owner-sv-champions@server.eda.org [mailto:owner-sv-
> champions@server.eda.org] On Behalf Of Brad Pierce
> Sent: Wednesday, May 20, 2009 3:21 PM
> To: sv-champions@eda.org
> Subject: RE: [sv-champions] Champions conference call May 21st, 2009
--
> agenda
>
> Neil,
>
> On Mantis 2735, I don't understand the ballot issue or its resolution.
> Can't method calls be chained?  For example, as in 8.11
>
>    Third, instance qualifications can be chained as needed to reach
into
> objects or to reach through objects:
>
>              p.next.next.next.val // chain through a sequence of
handles to
> get to val
>
> -- Brad
>

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