bug#33468: A bug with yes and --help

[Bernhard Voelker]

On 2/18/19 11:20 AM, Assaf Gordon wrote:
> [...] what do you think ?

Thanks for driving this.

To Eric's suggestion, I'd remove the RESET_OPTIND function argument,
because it's never used.
Re. OPTIND: what about resetting the values of all involved externals
to their previous value?

+  int saved_opterr = opterr;
+  int saved_optind = optind;
+  int saved_optopt = optopt;
+  char *saved_optarg = optarg;

...

+  /* Restore previous values.  */
+  opterr = saved_opterr;
+  optind = saved_optind;
+  optopt = saved_optopt;
+  optarg = saved_optarg;


In the attached, I've also changed the test a bit:
- remove unused _cleanup().
- add 'print_ver_ yes' to satisfy sc_env_test_dependencies,
- use case/esac to avoid spawning expr for the matching,
- and fix a missing "|| fail=1" in the final 'yes' test.

*BUT*: for 'yes', we'd introduce a regression due to option reordering:

  $ /usr/bin/yes a -- --help | head -n1
  a -- --help

  $ src/yes a -- --help | head -n1
  -- a --help

Any idea?

Thanks & have a nice day,
Berny

0001-all-detect-help-and-version-more-consistently.patch
Description: Text Data

gnulib-0001-long-options-add-parse_gnu_standard_options_only.patch
Description: Text Data