Meissner examples - the last one to fall

[Paul Thomas]

Tobi,

>Even with Paul's latest patch there are a huge number of accepts-invalid (or
>ice-on-invalid) bugs, so I don't believe "production quality" is not yet true.
> "Much better than 4.0" is definitely true.
>  
>
I would agree with that: - In starting on the accepts-invalid and the 
ice-on-invalids, I have found that there is a lot lurking there that is 
not very nice.  I have held back a few more patches that deal with a 
further eight such PRs until the equivalence constraints were out of the 
way.

What is impressing me is what gfortran will do with correct code.  I 
have just spent the day adding an interface to the fusion community's 
common data interface, MDSplus (find it with Google).  With a bit of 
work, I was able to read data from General Atomic's DIIID database and 
that of the Joint European Torus.  I have also compiled and run a free 
boundary equilibrium code, Proteus, and a neoclassical transport code 
NCLASS.  I see from the octave lists that gfortran has been used to 
build octave.  These things were definitely not feasible, even a few 
months back.  I think that all the contributors have a lot to be proud 
of.  Many have contributed many small but significant improvements over 
the last year and Richard Sandiford injected a quantum leap or two with 
his recent patches.

Has anybody taken the time to put together a collection of all the 
examples of the outstanding PRs?  I have noticed that more than a few 
are getting fixed un-noticed.

How long do we have before 4.2?  I have been of the opinion for a little 
while that we should assemble a priority list, divide the work on the 
more serious and try to work towards the "production quality" gfortran.

Well done everyone!

Paul T