[m-rev.] for review: changes to error messages

[Julien Fischer]

On Fri, 15 Oct 2021, Zoltan Somogyi wrote:

> I have just about finished a change that stores type, inst and mode
> definitions in parse_tree_module_srcs in checked form. This implies
> that any errors in those definitions are now diagnosed by the code in
> check_type_inst_mode_defns, not the code in add_type.m and add_mode.m,
> and indeed I intend for my next change to make the code
> in those two modules, which right now still process items individually,
> to take checked definitions as input, and just record them in the HLDS.
>
> The two sets of diagnostics (one in check_type_inst_mode_defns.m,
> the other in add_{type,mode}.m) were developed mostly independently.
> Sometimes they report different info, and usually they use different wording.
> I thought I would get your feedback, for each test case involving these
> diagnostics, whether
>
> - you prefer the old diagnostic
> - you prefer the new diagnostic
> - or you would prefer a blend of the two.

The new diagnostics are better.

Julien.