This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [gfortran] Support INTEGER<->LOGICAL conversion as GNU extension

On Saturday 28 May 2005 16:57, Roger Sayle wrote:
> (...) if I
> can't convince you both to change your minds (or soften your opinions),
> I'll concede and rewrite my patch to use -fugly-int.

So convince us :-)

I really don't have a very strong opinion about this, but I have two
reasons why I don't like having this extension enabled by default:

1) Just my general opposition against Fortran extensions.  It is
   IMHO really, _really_ bad to enable extensions by default, and
   even more so in a language that is already so mutilated by too
   many vendor extensions.

2) I would very much like to follow the default option settings
   that g77 used, so that people can write F77 code for gfortran
   that will almost certainly also compile with g77 without extra
   flags and options (i.e. source compatibility).

I suppose the first argument is a lost battle already, because most
compilers already allow this implicit conversion by default (at least
the HP f90, Cray f90 and Absoft f95 compilers don't complain).  And
I understand it would be nice if gfortran can compile existing "real
world code", such as the code you have to deal with, without additional

But the second argument still stands.  It will take quite some time
before g77 has been entirely replaced by gfortran.  At least until then
I believe we should make it as easy as possible to write Fortran that
will compile with both.  The easiest way IMO would be to make gfortran
behave like g77 in cases like this.  But I'd be just as happy with a
well-documented -fg77 option or something like that to mimic g77 where
needed, and make the defaults more relaxed like other fortran compilers.
I just think that would be more work.

> Hopefully OpenEye's expectations of a f77 compiler aren't too
> unreasonable?

It is sad enough that they are not :-/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]