-frecord-marker

[Keith Bennett]

Hi Adrian,

On Thu, Jul 20, 2006, Adrian Umpleby wrote:
> Keith, FYI I've also made -frecord-marker work on a per-object basis. So if you create an object
> with -frecord-marker=4, it's as if you added RECORD_MARKER='4BYTE' to all OPEN statements (well,
> apart from any that might already have a tag) in just that particular object. I can imagine it
> could also easily be implemented as a runtime environment variable (executable runtime, that is,
> not gfortran runtime) - e.g. GFORTRAN_RECORD_MARKER, or some such thing - thus allowing control
> over an executable's I/O behavior without the need to recompile it.

Thanks. That would certainly address the problem.

Apologies if I came across a little harsh in any of my previous posts. I
didn't fully appreciate the merits of the patch and I was a bit concerned
that it might be incompatible with any future attempt at allowing
4-byte or Intel markers to be specified at compile time.

Keith.