Re: Fwd: [Patch, Fortran] PR 43199 - Handle class%component references when reading .mod files

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

Dear Paul,

On 03/01/2010 09:26 AM, Paul Richard Thomas wrote:
> Anyway, since there are already so many instances of this, perhaps it
> is better to continue as you were.
>   
OK - I have now checked in my patch as Rev.

> Maybe we need a macro to obtain the $data field?  I´m mainly after
> something thats a bit easier to read and interpret, rather than
> expecting changes in the API.
>   

I was wondering whether the CLASS struct should not be removed for good
and the extra field be incorporated into the array descriptor - as it is
done, e.g., on Cray. While that only directly affects the tree code
(trans*.c), I think one could also change the gfc_symbol part. But
first, we need a new array descriptor such that we have space for the
extra argument - then one can slowly convert the CLASS code in trans*.c
and (afterwards/simultaneously?) the gfc_symbol part of the code.

 * * *

Talking about the new array descriptor: As mentioned, there is a TR
draft which also defines an interoperable array descriptor. Although
gfortran does not have to use internally the interoperating one and
although the draft *will* be modified, it might be still useful to have
a look. Latest TR draft of the subgroup (pre-meeting version) is
available ftp://ftp.nag.co.uk/sc22wg5/N1801-N1850/N1808.pdf - however, I
think it is only minutely updated compared with the older .txt version.
Note, however, that the draft will be reworked (schedule:
ftp://ftp.nag.co.uk/sc22wg5/N1801-N1850/N1812.txt). The objectives for
the TR - as agreed on last month - can be found at
ftp://ftp.nag.co.uk/sc22wg5/N1801-N1850/N1820.txt (there is also a
version at J3: http://j3-fortran.org/doc/year/10/10-142.txt - though I
have heard that 142r2.txt is the latest, though neither r1 nor r2 are on
the webserver). For the old draft, there exists a reply to comments at
http://j3-fortran.org/doc/year/10/10-122.txt

Tobias