This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RFC: patch to add ISO_FORTRAN_ENV intrinsic module


François-Xavier,

I am having some difficulty with this proposal; perhaps because I am ignorant on the format of the F2003 intrinsic modules. Why is it necessary to have custom code for each new intrinsic module? Would it not be better, as I wrote elsewhere, to have .mod files and a library in a special directory for intrinsic modules? That way, the existing module code could be used, with the addition of your bits to sort out intrinsic and non-intrinsic.
Hi all,

Attached patch adds the intrinsic ISO_FORTRAN_ENV module to gfortran,
as well as general support for intrinsic modules and the F2003-style
USE statement. I first thought this might wait until the ISO_C_BINDING
module is added, but seeing that Christopher encounters problems with
his paperwork, we agreed it's best to submit it now, and then it will
be relatively easy to change his patch (I volunteer to do it) to use
the framework I'm proposing here.
What structure has Christopher adopted?


 -- I've used the attr.intrinsic bit for the module symbol to
specifiy that it's an intrinsic module. Except one line in resolve.c
(see diff), it seems to be safe. However, I'd like
   a) to be sure that, for every user module symbol created, the
intrinsic bit is set to 0, so that we check that a non-intrinsic and
an intrinsic module of the same name aren't used
This seems OK to me. However, are we perhaps being a bit parsimonious with attribute bits? Maybe an intrinsic_module bit would be better?
   b) the code USEing non-intrinsic modules (in module.c) to issue an
error if an intrinsic module of the same name was already USEd; ie,
I'd like the following to error out (while, with my patch, it does
not):
You'll have to find the module symbol, check if it's intrinsic and throw on the new module being non-intrinsic.

module iso_fortran_env end module iso_fortran_env

program foo
use, intrinsic :: iso_fortran_env
use, non_intrinsic :: iso_fortran_env ! This should issue an error
end program foo
...and the other way round, I suppose!

Anyway, it's impressive to see you launching into module.c - convince me that what you are doing is OK!

Best wishes

Paul


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]