Re: [PATCH v2 04/20] t0040-parse-options: test parse_options() with various 'parse_opt_flags'

From: SZEDER Gábor <szeder.dev@gmail.com>
Date: Aug 20, 2022, 11:14 UTC
Message-ID: <20220820111422.GE3373722@szeder.dev>
In-Reply-To: <220819.86k0746t7v.gmgdl@evledraar.gmail.com>

On Fri, Aug 19, 2022 at 07:23:44PM +0200, Ævar Arnfjörð Bjarmason wrote:

> > +		error("'cmd' is mandatory");
> > +		usage_with_options(usage, test_flag_options);
> 
> I think you want usage_msg_opt() instead.

As I responded to a similar remark in the previous round, parse-options uses the "error:" prefix in its error messages:

  $ ./t/helper/test-tool parse-options -U
  error: unknown switch `U'
  $ ./t/helper/test-tool parse-options --unknown
  error: unknown option `unknown'
  $ ./t/helper/test-tool parse-options -i
  error: switch `i' requires a value
  $ ./t/helper/test-tool parse-options --int
  error: option `integer' requires a value
  $ ./t/helper/test-tool parse-options -i foo
  error: switch `i' expects a numerical value
  $ ./t/helper/test-tool parse-options --int foo
  error: option `integer' expects a numerical value
  $ ./t/helper/test-tool parse-options --quiet=42
  error: option `quiet' takes no value
  $ ./t/helper/test-tool parse-options --mode1 --mode2
  error: option `mode2' is incompatible with --mode1

Subcommand-related error messages should be consistent with these, and use the "error:" prefix as well. Unfortunately, both usage_msg_opt() and usage_msg_optf() use the "fatal:" prefix instead, so I will not use those functions anywhere in this patch series.

Show 5 quoted lines

> > +test_expect_success 'NO_INTERNAL_HELP works for -h' '
> > +	test_expect_code 129 test-tool parse-options-flags --no-internal-help cmd -h 2>err &&
> > +	cat err &&
> 
> Stray "cat", presumably in error..

Leftover debugging :(