[m-dev.] fixing the bug reported on m-users

[Zoltan Somogyi]

On Tue, 2 Nov 2021 13:16:46 +1100 (AEDT), Julien Fischer <jfischer at opturion.com> wrote:
> What are the disadavantages (if any) of enabling error reporting when
> generating interface before you have completed your current set of
> changes?

The main issue with reporting errors at interface-generation-time is that
it has been tested only minimally, and therefore switching it on
may result in problems. That is a price we have to pay if we *ever*
switch it on, so you are right, that is not really an argument against switching now.

In the short term, I expect switching now will require nontrivial changes
to some of the test directories, both to control whether or not we build
.int files before attempting code generation, and to handle situations
in which a set of errors that used to be all reported at code-generation time
now have a subset of those errors reported at interface-generation time.

The annoying part of that would be making those changes to the test suite
work both with and without reporting errors at interface-generation time,
just in case we have to turn that back off temporarily while we (I)
fix a glitch.

But since you agree that this is the preferred approach, I will proceed
along these lines.

Zoltan.