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

[Jerry DeLisle]

Tobias Burnus wrote:
> Jerry DeLisle wrote
>> The attach patch modifies scanner.c to keep track of the number of
>> continuation lines in a statement.  The F95 standard 3.3.1.4 says the
>> number shall not exceed 39. F2003 changes this to 255.  I have set the
>> limit to default to 255.  If -pedantic is given, it is set to 39.  If
>> the limit is exceeded an error is issued.
>>
>> I chose an error rather than warn because the standard states "shall
>> not".  If others feel strongly that should be a warn, I can easily change.
> 
> How about:
> * default and -std=f2003: 255 (free form)
> * -std=f90 and -std=f95: 39 (free form)
> * -std=f2003: 255 (fixed form)
> * -std=f77, -std=f90, -std=f95: 19 (fixed form)
> * default (fixed form): 19 ?
> 
> I don't see what it has to do with -pedantic.
> 
> If you go for an error (and maybe also otherwise), could you add an
> option like -maxcontin=N. At least for fixed-form code one easily
> exceeds the 19 lines, even if one wants to compile with the -std=* option.
> 
> Tobias
> 
Going into this patch I took the view that gfortran is primarily a Fortran 95 
compiler with F2003 features.  So I thought the default behavior should be the 
least restrictive of the two with 255 for free form.

I had not thought about fixed form checks, so I will add that.

The -pedantic is used to signify the more restrictive case.  I will adjust that 
so that the limits of 39 and 19 are used for "-pedantic -std=f95" meaning the 
user does not want to use the F2003 limits.

Gfortran does not have -std=f77 or -std=f90 as options.  Error checking is more 
or less lumped into legacy, gnu, f95, or f2003.

Regardless, I will try to take into account the numerous comments on this thread.

Jerry