[m-rev.] for post-commit review: fix more tests/invalid failures

[Zoltan Somogyi]

On Mon, 15 Sep 2025 12:48:12 +1000, Julien Fischer <jfischer at opturion.com> wrote:

> On Sun, 14 Sept 2025 at 14:28, Zoltan Somogyi <zoltan.somogyi at runbox.com> wrote:
> > Fix some more tests/invalid failures.
> >
> > tests/invalid/illtyped_compare.err_exp2:
> >     Add an expected output for target languages for which
> >     unification pretests cast the input arguments to c_pointers,
> >     not ints.
> 
> illtyped_compare.m needs to be updated with a comment explaining what
> the different expected outputs are for.

Yes, but we would first need to figure out what the test case is for
*exactly*. It seems to have been added by Simon in 2003, with
an explicit mmake rule for it that tests making its .opt file.
Then a commit by Fergus in 2004 specifies --no-intermod-opt for it,
which seems strange :-)

Since I have no idea why Simon wanted to test this type error when
making .opt files, and we can't ask him, I would be inclined to simply delete
its special make rule, and test it with just the general rule.

> > tests/invalid/pragma_c_code_dup_var.err_exp3:
> > tests/invalid/pragma_c_code_no_det.err_exp3:
> > tests/invalid/test_may_duplicate.err_exp3:
> > tests/invalid/test_may_export_body.err_exp3:
> > tests/invalid/try_detism.err_exp2:
> > tests/warnings/foreign_singleton.err_exp3:
> >     Update these expected output files to expect color in diagnostics,
> >     and (in some cases) updated wording as well.
> 
> We should probably rename test cases that refer to "pragma_c_code", since
> that hasn't been a pragma for a long time now.

Will do in a separate change.

> That looks fine otherwise.

Thanks.

Zoltan.