Object-oriented Fortran

[Tobias Burnus]

Hi Damian, hi Daniel, hello all,

Damian wrote:
> Since the rest of our library is object-oriented and we want to maintain
> that philosophy throughout, I exploit every new object-oriented feature you
> check in, e.g., I am already using type-bound procedures. In case it's of
> interest, it would really be nice to have allocatable derived types (do you
> support that already?).  Alternatively, type finalization would be nice, but
> I could very likely do without final procedures if I had access to
> allocatable derived types.

I'm not quite sure what you mean by "allocatable derived types". If you
mean "type t2; type(t), allocatable :: a(:)", i.e. TR 15581, this is supported.
If you mean allocatable scalars: No that is not yet supported, but planned
for 4.5 (esp. as it is useful for CLASS). And finally, if you mean Fortran
2008's "type t; type(t), allocatable :: a" that is also not yet supported.

And also finalization did not make it due to lacking review time and some
remaining issues of the almost finished patch. I think this is also a good
candidate for 4.5 since it is almost ready.

Daniel Kraft wrote:
> Abstract types yes, but unfortunatelly no deferred bindings. They
> would've been the next project for me (and probably will be the first
> to work on again), but they didn't make it for the 4.4
> "fixes and documentation changes only" deadline.

> Depending on how bugfixing goes and how the release-plan really is,
> I think however we could eventually start working on new features as
> patches a little before branching of 4.5 so we've got the patches quite
> ready when we can check in. What do others think about this? Is this a
> good idea or has experience (maybe) shown that it's not so good for
> various reasons? I don't think there will be many conflicts for
> stage3 check-ins.

The tendency is that the number of bug-fix patches decrease over Stage 3
as the number of important and simply fixable bugs decreases and,
additionally, towards the end of Stage 3 one increasingly does not
accept large bug-fix patches, which potentially may break other things.
And close to the release, only regression fixes are allowed; so far
gfortran had no (or no important [wrong-code, rejects-valid,
ice-on-valid-code] or easily fixable) regression. Thus one can really
start to doing 4.5 work then.

In principle, one can start to do 4.5 already earlier, but I think
Stage 3 is a good opportunity to to Stage-3 work, namely bugfixes.

Tobias