07-158

To:         J3
From:       Dick Hendrickson
Subject:    Clause 3; Don't add MACRO ERROR
Date: 2007 February 06
References: J3/07-007

A straw vote at meeting 178 was mildly in favor of adding an ERROR
statement to the macro capability.

This paper doesn't propose a MACRO ERROR facility.  Based on a private
communication from Malcolm, I now believe that a MACRO ERROR isn't
necessary and is difficult or impossible to get right.

The problem with trying to specify some sort of pre-processor error
mechanism is getting existing tools, like MAKE, to do the right thing.
Fortran doesn’t have a good way to tell a pre-processor to do the
right thing when it finds an error.  Whereas, existing compilers are
required to do some sort of diagnosis when they find syntax-like errors.
Rather than add a new class of error, it's easier to just have a macro
generate something like

       ERROR   inconsistent macro arguments xxx  yyy  zzz

and let the existing compiler rules do the right thing.