Re: suggestion about g77

[Paul Brook <paul at codesourcery dot com>]

[I think this discussion should be kept on the list]

>   My point, ( which I have now sufficiently occluded ), was only to keep
> the f95 and f77 portions separate, and default to as strict behavior for
> both as we can, this way we make it harder for someone to shoot them self
> in the leg.

gfortran doesn't have a strict f77 mode, and I don't think it's worth adding 
one. Almost all f77 code is also legal f95.

g77 supports a lot of things that are not standard fortran. IMO we should 
treat these just like any other language extensions. If they have unexpected 
interactions with other features (standard or otherwise) this should be 
documented, and a compiler warning/error issued if practical.

Obviously the decision whether to support a particular extension should be 
influenced by how much benefit it provides users. In the absence of other 
information it seems reasonable to assume that features implemented in g77 
are required by existing GNU toolchain users.

>   Extensions and interesting options would then have to be specified by
> flags.

This is what I meant by categorising the extensions. It would allow users 
finer control over which extensions are enabled, and allow us to just enable 
the "good" extensions by default.

In theory this could provide individual control over each extension. In 
practice most users probably don't need/want that level of detail. gcc 
already has way too many knobs.

I've no particular opinion on which extensions should be enabled by default, 
but I think users would complain if gfortran were pedantically strict by 
default. 

Paul