Another question about AssertionInstance

[HungMingWu]

Hi Mike, I have another question.
Here is my test code

module test;
	sequence s1(reg aa, reg bb);
		(bb, $display("s1"), $display("s1"));
	endsequence
endmodule

Here is my visitor

ExpressionVisitor {};
void visit(const SequenceWithMatchExpr& expr) {
    expr.expr.visit(*this);
    for (const auto& item : expr.matchItems)
        item->visit(ExpressionVisitor{});
 }

I trace expr.expr until InvalidExpression, and I think it discard AssertionPortSymbol info

Expression& AssertionInstanceExpression::bindPort(const Symbol& symbol, SourceRange range,
                                                  const ASTContext& instanceCtx) {
    // ignore
    auto [propExpr, argCtx] = it->second;
    if (!propExpr) {
        // The expression can be null when making default instances of
        // sequences and properties. Just return an invalid expression.
        return badExpr(comp, nullptr);
    }
}

Have any good idea ?

[MikePopoloski]

The sequence is not instantiated so you're just looking at the "default" instantiation, which of course has no way to fill in the arguments because... what would you put in there? If you want there to be a proper argument replacement you need to actually instantiate it.

[HungMingWu]

I don't need proper argument, I just need the expr is related to which AssertionPortSymbol
For this example

module test;
	sequence s1(reg aa));
		(aa, $display("s1"), $display("s1"));
	endsequence
endmodule

Since it just have one AssertionPortSymbol, I can know the expr is related to aa
But

module test;
	sequence s1(reg aa, reg bb);
		(bb, $display("s1"), $display("s1"));
	endsequence
endmodule

I cannot find the relation since I need to know it.
I think a better way is define another expression like ValueExpressionBase which refer to AssertionPortSymbol
How do you think?

[MikePopoloski]

You can't use a ValueExpression because an AssertionPort is not a value and can't necessarily be used in a general expression (what would the result type of that expression be?) Perhaps it can instead be replaced with a HierarchicalReferenceExpression (even though it's not actually hierarchical, it should be ok) and then set the type to be an ErrorType so that there won't be any follow on errors reported.