Make true return false less frequently

[197g]

This acknowledges a quirk in GNU true and makes true and false behave consistent to it.

Now treats recognized command line options and ignores unrecognized
command line options instead of returning a special exit status for
them.

true may return a non-true exit status (in particular EXIT_FAIL)
if writing the diagnostics of a GNU specific option fails. For example,
true --version > /dev/full would fail and have exit status 1. See commits
for more details.

This now passes tests/misc/false-status.sh.

Closes: #3007, #2996, #2989

[tertsdiepraam]

First of all, that title is amazing 😄 And the docs look great too! I think this makes a lot of sense. I wanted to suggest using AppSettings::IgnoreErrors at first, but it turns out that suppresses the help and version messages too, so the solution you found is the best we can do for now.

Could you add some tests for this behaviour?

[197g]

While adding more exhaustive tests, I noticed that -h and -V are auto-generated by clap as well but these options do not work in GNU and are ignored. In the spirit of mimicking GNU even more closely I've modified App to hide these shorthands and tests verify that the output is indeed empty.

[tertsdiepraam]

With that change, would it make sense to do it the other way around? I.e. use AppSettings::DisableHelpFlag and then create an arg that calls App::print_help if it is present? That way also only --help shows up in the documentation. The current solution is also fine, but feels more hacky.

[197g]

Using dedicated options looks cleaner. At least the similarities, as well as differences, between true and false look more obvious to me now, with less nesting. In the whole process I found yet another edge case when combining both flags (true --version --help) which is treated like invalid command line options in GNU. A custom handler for these options was seemingly necessary in any case, help is now exclusive to trigger the empty error path in combination.

[tertsdiepraam]

Who thought true and false could be so complicated :)

Nice work!

[tertsdiepraam]

Awesome! Only one nit about the formatting of the about

[tertsdiepraam]

Suggested change

	static ABOUT: &str = "
	Returns false, an unsuccessful exit status.
	Immediately returns with the exit status `1`. When invoked with one of the recognized options it
	will try to write the help or version text. Any IO error during this operation is diagnosed, yet
	the program will also return `1`.
	";
	static ABOUT: &str = "\
	Returns false, an unsuccessful exit status.
	Immediately returns with the exit status `1`. When invoked with one of the recognized options it
	will try to write the help or version text. Any IO error during this operation is diagnosed, yet
	the program will also return `1`.
	";

[tertsdiepraam]

Suggested change

	static ABOUT: &str = "
	Returns true, a successful exit status.
	Immediately returns with the exit status `0`, except when invoked with one of the recognized
	options. In those cases it will try to write the help or version text. Any IO error during this
	operation causes the program to return `1` instead.
	";
	static ABOUT: &str = "\
	Returns true, a successful exit status.
	Immediately returns with the exit status `0`, except when invoked with one of the recognized
	options. In those cases it will try to write the help or version text. Any IO error during this
	operation causes the program to return `1` instead.
	";

[197g]

Only one nit about the formatting of the about.

Good to know! Formatting was borrowed from base32, the first utility with all the options I was looking for as a reference. Might want to promote this comment to an issue for base32, base64, basenc as they share similar formatting in their About text.

[tertsdiepraam]

Might want to promote this comment to an issue for base32, base64, basenc as they share similar formatting in their About text.

Oh I see. That's a good idea!

[rivy]

Looks good!
I like the revised logic.
👍🏻 for merge.