(j3.2006) Interop observation
Mon Feb 6 18:50:37 EST 2012
> possibility (if anything is done at all) is simply
Nothing is a good option since its a topic for the linker notes. But (assuming action) rather than giving advice about how to avoid the problem on some class of processors, I'd put in a note for a "typical" POSIX style Processor which fails.
Sent from my iPad
On Feb 6, 2012, at 4:33 PM, Van Snyder <Van.Snyder at jpl.nasa.gov> wrote:
> As was observed in an earlier discussion on this topic, we don't want to
> add anything normative, or at least not imperatively normative, that
> might result in essentially all vendors needing to revise their
> processors just to try to outwit people who are trying to break the
> One possibility (if anything is done at all) is simply to advise that a
> binding name for a Fortran procedure ought to be in mixed case, in the
> hope that most processors will produce their linker names in a single
> case. I've seen all upper case, all lower case, with and without
> various numbers of underscores, both before and after the Fortran name,
> but never a mixed-case linker name.
> Less desirable, but perhaps more helpful, would be to augment that note
> with a recommendation (not a requirement) that processors do not use
> mixed case in their linker names for non-BIND procedures or common
> On Mon, 2012-02-06 at 15:15 -0800, Keith Bierman wrote:
>> Obviously highly platform dependent. But unix flavored environments
>> are common enough that a note wouldn't be unreasonable (IMNSO )
>> Sent from my iPad
>> On Feb 6, 2012, at 4:07 PM, Van Snyder <Van.Snyder at jpl.nasa.gov> wrote:
>> Which fails
>>> A colleague has observed that it's not terribly difficult to construct a
>>> duplicate definition using interop features.
>>> The example was
>>> SUBROUTINE AnyNamesubf(i)BIND(C,NAME="csubf_")
>>> USE, INTRINSIC :: iso_c_binding, ONLY: c_int
>>> INTEGER(c_int) :: i
>>> CALL Csubf(i)
>>> END SUBROUTINE AnyNamesubf
>>> where Csubf was a non-interoperable Fortran external subroutine.
>>> I don't think this is a normative issue for the standard, but should we
>>> put a note somewhere that might help people not create these? Maybe
>>> something like advising not to give binding names to Fortran procedures
>>> that are all lower case (or all upper case) and are sufficiently similar
>>> to the external names of non-interoperable Fortran external procedures
>>> that the linker might be sent the same name for both of them.
>>> J3 mailing list
>>> J3 at j3-fortran.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the J3