run: preserve the error type from stdlib/flag

[marco-m]

Hello Lea,

I was surprised that -h was not considered as a short for -help:

go run . -h
Error: flag: help requested
See 'foo --help' for usage
exit status 1

Looking at the clir code, I realized that if we were to "simply" add -h as shortcut to the -help flag, due to the stdlib/flag limitation, this would make appear two lines:

Flags:

  -help
    	Get help on the 'foo' command.
  -h
    	Get help on the 'foo' command.

Instead, I modified the return fmt.Errorf("Error: %s\nSee '%s --help' for usage", err, c.commandPath) to use the %w verb. This allows user code to do:

cli := clir.NewCli("foo", "desc", "ver")

err := cli.Run()
if errors.Is(err, flag.ErrHelp) {
	cli.PrintHelp()
	return nil
}
return err

and, going back to the first example, to obtain:

go run . -h
foo - bar

Available commands:

   ...

Flags:

  -help
    	Get help on the 'foo' command.

[marco-m]

Ah. This doesn't work for subcommands: it prints the top level help. Need to think a bit more...

[marco-m]

@leaanthony I am actually quite satisfied with this second attempt.

I am introducing a dependency, testscript from https://github.com/rogpeppe/go-internal, BUT it is a test-only dependency :-)

If you are not familiar with it, I think you will like it! It is perfect to test libraries such as this one that prints directly to stdout/stderr.

(It would also allow to transform many of the existing tests in "real" tests, since I noticed that many of them do not assert anything ?)

[marco-m]

only for tests :-)

[marco-m]

I suggest to start from here

[marco-m]

change in behavior

[marco-m]

change in behavior