Re: [Fortran-experiments] Bind(c) first experiences

["Christopher D. Rickett" <crickett at lanl dot gov>]

> Hi Chris,
> 
> Christopher D. Rickett wrote:
> > thanks for sending me your code.  i tried it out, and the output from 
> > gfortran is similar to that of NAG.
> Thanks. I probably have an too old version.
> Your patch, http://gcc.gnu.org/ml/fortran/2007-01/msg00301.html is not
> checked in yet, is it?

not that i know of.  Steve was planning on looking at it soon.

> I probably should apply the patch first, before continue doing some tests.
> Are there further changes or is that patch up to date?

not anymore.  i just made a couple of minor changes to add a test for the 
second arg of c_f_pointer to make sure it has the pointer attribute.  
also, i mark c_ptr and c_funptr as bind(c) since it's required for them to 
be interop and i'm now enforcing that.  there is a new testcase for this.  
i'll try and submit this small patch today.

> What is actually the procedure for applying patches to the branch?

i'm not sure.  i create a diff with svn and submit it to the list.  is 
there anything else i need to be doing?

> 
> (We should by the way also merge the trunk up to the branch and
> consolidate the VALUE implementation.)

i'm hoping to look at this soon, but if anyone else wants to volunteer 
that would be great.  :-)  there is some other work that needs done to the 
patch before entering the trunk, including finishing the c_f_pointer 
versions in iso_c_binding.c/h for the types/kinds not listed yet.  this 
should be fairly straight-forward since they all have the same pattern.

> Regarding to your last patch,
> http://gcc.gnu.org/ml/fortran/2007-01/msg00301.html
> 
> + * to generate.         If <code>c_ptr</code> is given, then
> <code>c_null_ptr</code>
> 
> I think the coding convention is not to use "<code>...</code>"  (at
> least none of the other gfortran files do)

i'll go through and remove them and change my function comment blocks to 
the style used in gfortran now.

> 
> +static void gen_special_c_interop_ptrs (const char *kinds[],
> +                                       const char *module_name)
> 
> If you use
> 
> +static void
> +gen_special_c_interop_ptrs (const char *kinds[],
> 
> one can grep for    ^gen_special_c_interop_ptrs  to find the function
> itself (and not the use of the function). This seems to be the coding
> convention in gfortran.

thanks.  i'll change that too.

Chris