an option to enable C9X features

an option to enable C9X features

[Jim Wilson]

We need one.  Preferably, we want something that the gcc2 group will also
accept, so that we can avoid command line incompatibilities between gcc2 and
egcs.  I'm volunteering to take up this issue with the gcc2 group, but first,
I thought I would ask for comments here.

I believe that there have been two previous suggestions.  Ulrich Drepper
suggested these options
	-iso-c (defaults to c89)
	-iso-c=89
	-iso-c=9X
Paul Eggert suggested a more complete set of options
	-std=c9x		C9x
	-std=gnu		GNU C (the default)
	-std=iso9899:1990	ANSI C X3.159-1989 / ISO/IEC C 9899:1990
	-std=iso9899:199409	ANSI C as modified by Normative Addendum 1
	-std=iso9899:199?	equivalent to -std=c9x, whenever it comes out
	-std=lia1		LIA-1
	-std=traditional	K&R C

	-ansi			equivalent to -std=iso9899:1990
	-traditional		equivalent to -std=traditional

Both of them seem reasonable, but I'm inclined to propose the -std=X options
first, because they give us more flexibility, and would allow us to also
handle for instance C++ and Fortran standards the same way.  I don't like
the colons though.  If we can't get the gcc2 folks to accept that, though,
then I think that the -iso-c=X options are a reasonable compromise that
would allow us to start adding C9X features.

Jim

Re: an option to enable C9X features

[Richard Earnshaw]

> 	-std=iso9899:1990	ANSI C X3.159-1989 / ISO/IEC C 9899:1990
> 	-std=iso9899:199409	ANSI C as modified by Normative Addendum 1
> 	-std=iso9899:199?	equivalent to -std=c9x, whenever it comes out
...
> I don't like the colons though.  

-std=iso9899-1990 ...?

Re: an option to enable C9X features

[Richard Henderson]

On Tue, Mar 10, 1998 at 10:12:13AM +0000, Richard Earnshaw wrote:
> > 	-std=iso9899:1990	ANSI C X3.159-1989 / ISO/IEC C 9899:1990
> > 	-std=iso9899:199409	ANSI C as modified by Normative Addendum 1
> > 	-std=iso9899:199?	equivalent to -std=c9x, whenever it comes out
> ...
> > I don't like the colons though.  
> 
> -std=iso9899-1990 ...?

Except that colons are how ISO writes it, eg `ISO/IEC 9945-1:1996(E)'
is POSIX 1.  Not that it would really be confusing if we changed it...


r~

Re: an option to enable C9X features

[Jim Wilson]

	Except that colons are how ISO writes it, eg `ISO/IEC 9945-1:1996(E)'
	is POSIX 1.  Not that it would really be confusing if we changed it...

I have concerns about the colons because I usually think of them as special
characters not appropriate for normal use.  However, trying out a few shells,
it seems that colons are OK for all of them, so we may as well stick with
them for now.

Jim

Re: an option to enable C9X features

[Ulrich Drepper]

rearnsha@arm.com (Richard Earnshaw) writes:

> > 	-std=iso9899:1990	ANSI C X3.159-1989 / ISO/IEC C 9899:1990
> > 	-std=iso9899:199409	ANSI C as modified by Normative Addendum 1
> > 	-std=iso9899:199?	equivalent to -std=c9x, whenever it comes out
> ...
> > I don't like the colons though.  
> 
> -std=iso9899-1990 ...?

The colons are the official notation and we should switch this without
real reasons.  `iso9899-1990' would mean ISO 9899, part 1990.

-- Uli
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------

Re: an option to enable C9X features

[Mike Stump]

> To: egcs@cygnus.com
> Date: Mon, 09 Mar 1998 18:53:32 -0800
> From: Jim Wilson <wilson@cygnus.com>

> 	-std=gnu		GNU C (the default)
> 	-std=iso9899:1990	ANSI C X3.159-1989 / ISO/IEC C 9899:1990
> 	-std=iso9899:199409	ANSI C as modified by Normative Addendum 1

I like the flexibility of the longer options, but I wonder if we
should force dumb users to type such long names, or are we secretly
discouraging the use of any ISO standard?  :-)/2

Re: an option to enable C9X features

[Jim Wilson]

	> 	-std=iso9899:1990	ANSI C X3.159-1989 / ISO/IEC C 9899:1990
	> 	-std=iso9899:199409	ANSI C as modified by Normative Addendum 1

	I like the flexibility of the longer options, but I wonder if we
	should force dumb users to type such long names, or are we secretly
	discouraging the use of any ISO standard?  :-)/2

We can add aliases if necessary.  We already have -ansi (and -pedantic?) which
is equivalent to -std=iso9899:1990 for C code.  We could perhaps add a -std=na1
for the Normative Addition #1.  I suspect that there will be lots of different
standards that people will be interested in, so having the generality at the
begining will be useful.

Jim