Re: [PATCH, Fortran] Four small DEC extensions

[Jerry DeLisle <jvdelisle at charter dot net>]

On 10/13/2016 01:01 PM, Fritz Reese wrote:
> On Thu, Oct 13, 2016 at 2:24 PM, Jerry DeLisle <jvdelisle@charter.net> wrote:
> > On 10/12/2016 05:30 AM, Fritz Reese wrote:
> > > Here I submit for review four small extensions to the GNU Fortran
> > > frontend for compatibility with legacy code. I figure it might be a
> > > nice change of pace from my larger patches. Never fear, for I have
> > > more large patches to come, which I will continue to submit
> > > one-at-a-time.
> > >
> > > Each extension is enabled with its own compile flag, and each are
> > > enabled along with -fdec. They are fairly self-explanatory.
> ...
> > By the way, you may be thinking of -fdec-all that will grab all these dec
> > extensions, or maybe just use one for all like -fdec and not have so many flags.
> >
> > I wonder if we should add a flag called -floose-as-a-goose and just turn off
> > all checks. ;)
>
> Funny you mention it.... An earlier stage of my extensions contained
> (by request) an "-flazy-types" which basically did just that. I'm
> pretty sure I ripped out half the resolution phase for it, from errors
> in type conversions to argument type mismatches and more... Yuck!
> Thankfully I was able to toss that flag in place of some more
> reasonable ones eventually.
>
> ...
> > > OK for trunk?
> >
> > Not yet, generally, I don't think we need so many compiler flags. We should
> > probably think about this as far as the other extensions you have already
> > committed.
>
> The goal of these extensions is to allow GNU Fortran to compile a
> legacy codebase which compiles "fine" with another legacy compiler
> (e.g. pfc or ifort). Therefore in my mind the typical use case of the
> extensions is just to enable them all, which is why I made "-fdec"
> from the beginning. (Not that I encourage such crippling ignorance in
> a compiler- rather I am just responding to demands for these
> compatibility extensions, where they seem reasonable.)
>
> When I originally proposed the extensions on the mailing list here I
> got the impression that some folks were concerned about some
> extensions affecting default behavior of the compiler and wanted the
> ability to selectively enable them. My gut reaction was to just use a
> new flag for every extension to make such people happy - maybe that is
> overkill.
>
> I certainly don't mind squashing some flags and rolling them in with
> -std=legacy/gnu or even just -fdec if there is a consensus on the
> issue. I am curious to see what others think.
>
> ---
> Fritz Reese

Of the four in this patch, TYPE is the only one that I am uncomfortable with so 
maybe it should be behind -fdec, The other three seem minor to me, especially 
the LF, I dont think LF needs a flag.

I am sure others will chime in.

For clarity, so far we have:

flag_dec_intrinsic_ints
flag_dec_structure
flag_dec_static
flag_dec_math
flag_dec

and you were thinking about:

-fdec-feed] Treat form feed characters as whitespace.
-fdec-type-print  Treat TYPE as an alias for PRINT where applicable.
-fdec-loc-rval]
-fdec-logical-xor

I would go with putting latter 3 just under -fdec and no flag for LF whitespace.

Jerry