Re: GSoC 2008

[Tobias Burnus <burnus at net-b dot de>]

Hi Janus,

Janus Weil wrote:
> Now the most obvious thing for me to do would be to continue working 
> on last year's project, namely procedure pointers

I think this is something which should really be done for GCC 4.4.0: The 
procedure pointer complex including the related C binding parts. Reading 
comp.lang.fortran this is one of the missed things and also Joost misses 
it for Car-Parinello 2000 (CP2K).

> and type-bound procedures. As you probably remember I implemented most 
> of the basic PROCEDURE syntax, which went into the 4.3, but didn't 
> really finish procedure pointers. So I could try to finish it this 
> year and see how far I get into type-bound procedures.
>
> First question: Do you think this would be helpful at this point, or 
> is there any other project you would prefer me to do instead (may it 
> be more urgent/more important/more appropriate for SoC/etc)? I 
> remember Paul making some kind of plan for F2003 some weeks ago 
> (connected to parametrized derived types). I don't know how far that 
> has gone now, and how my efforts would fit in there.

I think the OOP part is useful. Paul is drafting the internal structure 
- especially in view of parameterized derived types and type extensions. 
The parameterized-derived types will be implemented by Sa Liu/IBM. And 
Daniel Kraft also plans to do OOP work as GSoC. Nonetheless, I think the 
task is big enough that more than one person can work on it - and one 
can split it in separate projects.

(In any case, it would be bad if your early patch for parsing type-bound 
procedures you sent in September in a private mail were lost.)

However, there are enough other useful projects, looking at, e.g., 
http://gcc.gnu.org/wiki/Fortran2003Status

IEEE is missing. (I think Fran,cois-Xavier and Asher planned to do some 
work there, but I don't know how much preliminary work they did.)

The ASSOCIATE construct is useful and I think quite independent.

Assignment to an allocatable array (see 3.3 in 
ftp://ftp.nag.co.uk/sc22wg5/N1601-N1650/N1648.pdf ) and pointer 
assignment are also missing - and I think useful.

(Fortran 2008 features could also be implemented, though I think picking 
a Fortran 2003 feature is more useful.)

Other than that, see also http://gcc.gnu.org/wiki/GFortran44

> So, second question: Do you think a project concerning documentation 
> would be worthwhile to think about? Would this be appropriate for 
> Summer of Code? And would it be satisfying to only work on 
> documentation instead of doing real coding? I guess for me it would be 
> more interesting to work on a F2003 feature like procedure pointers, 
> but if people think setting up some solid documentation would be more 
> helpful, I would also be willing to do that.

I think it is useful to devote time for documentation, but I think there 
should also be some coding - last but not least because it is more fun. 
Additionally, the GSoC FAQ has: "While we greatly appreciate the value 
of documentation, this program is an exercise in developing code; we 
can't accept proposals for documentation-only work at this time."
But documentation patches are always welcome!

> And then, third question, if I would decide to apply for GSoC and find 
> a suitable project, would there be anyone willing to be my mentor? 
> Last year this was a bit of a problem, and in the end Steven Bosscher 
> did it, just because no one else would do it, and although he's not 
> very active on gfortran any more. But he basically ended up doing only 
> the 'official' stuff (which is not a lot), and I was communicating 
> much more with other people about the project itself and the technical 
> details. It would be great if this year someone could be mentor who is 
> closer to the project itself. I think the 'official' stuff for Google 
> is really not a lot of work at all. There is just some midterm 
> evaluation and a final evaluation, each consisting only of a short 
> questionnaire. Apart from that it would be just me getting on your 
> nerves and asking lots of stupid questions, which I'll do anyway to 
> anyone who is subscribed to the mailing list ;)

I would do it, but I would not mind having a backup/co-mentor. I signed 
up last year, but somehow I was not approved a mentor (or I overlooked 
the emails?) - nonetheless I still got a SoC T-shirt :-) This year I got 
approved and now I'm already swamped by GSoC-mentors-list emails ;-)

Tobias