equivalence, common and initialization was: Re: [gfortran] patch

[Tobias Schlüter]

Actually there still are some restrictions that we don't implement: e.g.
COMMON /A/X
EQUIVALENCE (X,Y)
REAL :: Y = 1.
DATA X / 1. /
END
is not rejected by the frontend, even though there are several errors.

But as you said, accepts-invalid is better than rejects-valid, so I 
think you should just assert that you don't have multiple 
initializations for objects when building the trees, and hope that the 
frontend will be fixed to give an error in such cases. That way valid 
code can be made working without much delay. Once I can connect to 
gcc.gnu.org again, I will open a PR, where we can keep track of 
restrictions that we missed. I will then fix the frontend once I have my 
other patches out of the way.

- Tobi