Re: user backtraces

["N.M. Maclaren" <nmm1 at cam dot ac dot uk>]

On Sep 6 2012, Manfred Schwarb wrote:
> the recent post of Nick MacLaren reminded me of the old
> thread about user backtraces:
>
> http://gcc.gnu.org/ml/fortran/2012-06/threads.html#00131

That's not an OLD thread - this one goes back a LOT longer, and
not just in gcc/gfortran :-)

> AFAIK, this thread died, and nothing was actually implemented.
> What is the state of things?
>
> I really would appreciate such a feature.
> There recently was a thread at the gcc mailinglist, perhaps
> this gives some new momentum?

I should be happy to be (peripherally) involved, but absolutely
cannot take anything on at present.  In addition to Tobias'
suggested approaches, the standard allows compilers to extend
the standard modules; ISO_FORTRAN_ENV would be the best, but is
not ideal.

I would favour creating a new intrinsic module, say GNU_EXTENSIONS,
as the most future-resistant solution.  This would break no code
and, as far as I can see, would conform with the standard.  There
wouldn't even need to be any changes to the warning code, as the
use of that module would be enough!

The biggest question is whether it is worth defining a compiler-
independent interface, with one eye on it being put into the actual
standard.  It's moot, and I would vote one way today and another
tomorrow :-)

Incidentally, ERRTRA started with IBM on the System/360, to update
one of the posts I noticed.  Please don't adopt it, as it was
superseded as an interface in the 1960s ....


Regards,
Nick Maclaren.