Re: suggestion about g77

[Paul Brook <paul at codesourcery dot com>]

On Tuesday 22 March 2005 08:50, Lars Segerlund wrote:
>  Hi,
>
>   I have been doing a bit of thinking, and it seem's we are starting to
> gather a lot of g77 compability code and functions, and I have some worries
> that we will bring a lot of g77 peculiarities into the fortran frontend (
> f95 that is ), perhaps we should have the g77 compability contained by a
> compile time flag ? fx. -fg77 which at the same time would disable the f95
> extensions ?
>
>   The reason for this is what if we have things like the g77 %loc %val %ref
> mixed with the 'proper' frotran 95 mechanisms ? Also a mix of new and old
> library functions doesn't seem so nice, with the time we could separate the
> g77 stuff into it's own lib ?

I can guarantee there are people who want to use both a the same time. 
It is already be possible to disable the nonstandard extensions. Any 
extensions that aren't disables by the appropriate commandline options are a 
bug.

It may be desirable to categorise supported extensions into logical groups, 
and give the user more fine-grained control over which extensions are 
enabled. The first step is to decide what categories are useful, and which 
categories each extension belongs to. "Implemented in g77" doesn't seem a 
particularly useful category.

I don't understand your point about "old" and "new" library functions. Having 
additional intrinsics available doesn't really cost the user anything, the 
linker will discard them if they aren't being used.

gfortran was written as a Fortran 95 compiler, and as such "f95 extensions" 
isn't a meaningful concept. The core features implemented by the compiler are 
those defined in the fortran 95 standard, everything else in an extension. 
Adding a pure Fortran 77/g77 mode would be quite a bit of work, and IMHO a 
waste of time. It is desirable to have gfortran correctly compile everything 
g77 can compile, but I don't think making gfortran behave exactly like g77 
gains us anything.

Paul