[Patch, fortran] PR19262 more than thirty-nine continuation lines should issue a std-warn

[Brooks Moses]

Manfred Schwarb wrote:
> I just checked, ifort does not issue warnings with 300 continuation 
> lines. 

That's just because you didn't check a high enough number.  Ifort 8.1 
issues an error (not a warning) in fixed-form with 513 continuation 
lines, and in free form with 7201 continuation lines.

> I'm really not good in interpreting the fortran standards, but I can
> imagine that the "shall not" is not meant to be a restriction to the
> compiler, but to the user who wants to write standards conforming code. 
> I really can not imagine any reasons why a standard should not allow
> a compiler to accept more continuation lines as an extension.

That is correct.  The standard, in general, refers to what programs are 
allowed to include.  When it specifically lists something as a 
"Constrant:", that means that the compiler is required to emit an error 
if a program is in violation.  "Shall not" is not a constraint; the 
compiler is not required to emit an error.

However, in many cases it is useful for the compiler to emit a warning 
or error that is not strictly required by the standard, so "the compiler 
isn't required to emit an error for excess continuation lines" is merely 
a necessary argument, not a sufficient one, for making the case that the 
compiler shouldn't emit at least a warning.

- Brooks