Re: SoC fortran procedure pointers

[Tobias SchlÃter <tobias dot schlueter at physik dot uni-muenchen dot de>]

Hi,

since implementation of these advanced feature would best happen on a 
branch, I think both RamÃn and Janus should take care of the copyright 
assignment formalities ASAP, so that they can be granted write access 
timely once their projects are accepted for the SoC.

RamÃn GarcÃa wrote:
> As for me there would be no problem in working together. However, in
> that case the project would be slightly small, and we should go beyond
> it and implement something else.
>
> My approach to the problem is the typical of reverse engineering
> situations, that is, working with code that I am not familiar with. I
> try running the software with a simple case, and look for the code
> paths. With this approach I proposed a small patch that is a small
> progress towards the implementation of procedure pointers.

Looks good, but please don't forget that in order for us to accept 
patches for new features good testsuite coverage, which can also take a 
lot of time, is also required.

> For debugging it has been useful to find that
> gfc_current_locus.nextc contains the text not yet parsed.

Indeed.  Once you're used to it, it's quite obvious, but nevertheless, I 
think we should record it in gfc-internals.texi.

> diff -x '*~' -u ./decl.c ../../../../gcc-4.1.2/gcc/fortran/decl.c
> --- ./decl.c 2006-11-06 18:18:03.000000000 +0100
> +++ ../../../../gcc-4.1.2/gcc/fortran/decl.c 2007-03-26 
> 08:10:46.298296687 +0200
> @@ -1699,24 +1699,32 @@
>     }
>
>   m = gfc_match (' type ( %n )', name);
> - if (m != MATCH_YES)
> - return m;
> -
> - /* Search for the name but allow the components to be defined later. */
> - if (gfc_get_ha_symbol (name, &sym))
> - {
> - gfc_error ('Type name '%s' at %C is ambiguous', name);
               ^                                 ^
Your mailer seems to fsck up double quotes and indentation.  If it can't 
be forced to behave sanely, please attach patches instead of copying 
them inline.  I find your patch a bit difficult to follow with all 
indentation removed, but you're certainly not changing the right part of 
the code for the testcase.  You want to change the parsing of 
component-decls, but you're modifying the parsing of type-specs.

Anyway, I would very much welcome if both your projects were accepted 
for the SoC.

Cheers,
- Tobi