Re: [open-cobol-list] CVS updated (3)

[Roger While]

Yes, a good idea.
Actually, my fault that I did not state that the necessity for
recompilation/link.
I would make it a global integer that each developer
would increment if recompilation/link is necessary.
It's not only structure changes; for instance I have in test
some optimization changes that move stuff out of libcob into
the generated cobol code.
The drawbacks are not overwhelming.
(1) is acceptable.
(2) is not a problem as we do it in the program initialization
      ("if !initialized" in the generated C code)

Let me think about it.
With some of the things that I envisage implementing,
it will become a common occurrence to have to recompile/link
after a CVS pull.

Roger

At 14:57 01.11.2004 +0100, you wrote:
> That seems fine, except maybe one point: the reordering of fields
> in the cob_file struct in fileio.h.
>
> It means that ALL programs MUST be recompiled and (relinked
> if static) against the newer libcob.
>
> I wonder if we could not add a version identification in such
> structures and let the runtime check the availability of new
> features before using them.
>
> Drawbacks:
> 1) Incompatible the first time
> 2) More time spent at execution.
>
> Or at least keep a note for the next release pointing to that
> incompatible change.
>
> Your opinion ?
>
> Bernard Giroud
> Credit Lyonnais (Suisse) SA
> ----- Original Message -----
> From: "Roger While" <address@hidden>
> To: <address@hidden>
> Cc: <address@hidden>; <address@hidden>
> Sent: Sunday, October 31, 2004 12:29 PM
> Subject: [open-cobol-list] CVS updated (3)
>
>
> >
> > CVS updated.
> > Always activate I/O exceptions.
> >
> > Case 1 : FILE STATUS is defined.
> > Do not terminate and do not print the exception.
> > Case 2 : FILE STATUS is not defined :
> > Print the exception and exit the execution
> > (at the moment with return status 1. We maybe could
> > /should return the actual status to allow scripts to check).
> >
> > Any USE declaratives for the file will be executed prior to following
> > the above rules.
> >
> > Roger
> >