[Bug fortran/39427] F2003: Procedures with same name as types/type constructors

[burnus at gcc dot gnu dot org]

------- Comment #6 from burnus at gcc dot gnu dot org  2010-05-04 16:10 -------
Created an attachment (id=20557)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20557&action=view)
First draft patch

(In reply to comment #5)
> (In reply to comment #4)
> > What a horrible rule...

What do you mean? Allowing "name()" as (generic) constructor (function)? Why
not? That matches the natural use. If you mean the constraint in comment 3:
That just ensures that one cannot have use a structure constructor and generic
function simultaneously.


> Intel has promised to support it in its next release.

Well, I can also promise that _one_ of the next releases will have it :P

At least there are plans to improve OOP and there is also a Summer of Code
projects for OOP, cf. http://gcc.gnu.org/wiki/SummerOfCode

 * * *

Attached is a draft patch which should work in simple cases.

TODO:

* Call structure constructor if no generic function matches; that means:
Converting the actual arguments into a constructor and do the checking (cf.
FIXME in resolve.c)

* Add a is-function check: "A generic name may be the same as a derived-type
name, in which case all of the procedures in the interface block shall be
functions." (Note: That's not a constraint thus one does not need to catch all
cases.)

* Fix the case of first defining the generic name and then the type:
interface t
  procedure t2
end interface t
type t
end type t

* Add a new "%X" (X = some letter) to finally solve the "@" printing problem in
error messages for PPC and for constructors (this PR): gfc_error should then
simply strip the @... from the message for arguments to %X while for %s no
extra stripping is done as it is currently done. One then only needs to change
all relevant %s to %X and one is done. (using a more-intuitive different latter
than %X would be helpful; maybe %S = Symbol String?.)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39427