Re: [gfortran] Fix RETURN matcher was: Re: bug fc110

[Richard E Maine <Richard dot Maine at nasa dot gov>]

On Apr 15, 2005, at 3:54 PM, Tobias Schlüter wrote:

> Yes, that's the place that I had quoted :-) in reply to which Walt 
> claimed
> that the blank between RETURN (keyword) and 1 (constant) is not 
> required,
> because it's not a constraint.  But maybe I misunderstood him?

Yes, I think you did. At least the above is sure confused.

There are *LOTS* of things that are required even though they are not 
constraints. Pretty much the whole standard is a bunch of requirements, 
and most of it is not constraints. In this regard, it is absolutely 
critical to distinguish between requirements on the compiler and 
requirements on the programmer - the two are fairly often exactly the 
opposite in that requirements on the programmer correspond to freedoms 
of the compiler.

The blank is required. Period. That's what the cited text says. But 
remember that requirements of the standard are requirements on 
programmers - that's actually quite important. The requirements on the 
compiler are quite different. The compiler is, of course, required to 
accept the blank. That's because a compiler is required to accept all 
standard-conforming syntax.

The business about constraints comes in because constraints are one of 
the things that compilers are required to be able to diagnose 
violations of. This is not a constraint and, therefore, the compiler is 
not required to diagnose the absence of the blank. Note the difference 
here: the programmer is required to put the blank there, but the 
compiler is not required to diagnose its omission.

--
Richard Maine                |  Good judgment comes from experience;
Richard.Maine@nasa.gov       |  experience comes from bad judgment.
                            |        -- Mark Twain