[sv-bc] Another ballot comment related to 232

Ballot comment 232 (Mantis item 652) about automatic variable lifetimes is
closely related to my Mantis item 409 (ballot comment 10?).

Both of these are about subprocesses being able to reference an automatic
variable via ref arguments, after the lifetime of the automatic has expired.
In Mantis item 652, the subprocesses are join_any/join_none, while in Mantis
item 409, the subprocesses are nonblocking assignments and procedural
continuous assignments.

Since the approach to fixing 652 was to put restrictions on the subprocesses
rather than the variables passed by reference, presumably a similar approach
is required for 409.  A comparable set of restrictions would be to forbid
ref argument formals in the same places that automatics are forbidden in
1364-2001 (LHS of nonblocking assignments or procedural continuous
assignments, RHS of procedural continuous assignments or force statements,
intra-assignment event controls of nonblocking assignments, and as arguments
to such system tasks as $monitor and $dumpvars).

However, it could be considered that the restriction for ballot issue
235 (Mantis item 666) already resolves Mantis item 409, as a side effect.
That restriction ensures that all ref argument formals are automatic
variables.  If we conclude that they are therefore illegal anywhere
automatics are forbidden, then the problem situations are already forbidden.

Is this restriction sufficiently clear, or is there a need for a more
explicit restriction?  One concern I would have with relying on this is
that we might end up relaxing the restriction in Mantis 666.  If that
happened, someone would have to remember that we were relying on this
restriction to resolve 409 also.  We would then have to go back and put
in more explicit restrictions on ref argument formals.

Steven Sharp
sharp@cadence.com