stop_on_unknown doesn't work

[NathanZook]

Explicit raises are now inside the block. The entire mechanism seems to have been neglected, as it's application is uneven.

[kbrock]

@Student Any tests/examples showing how you would like it to work?

[NathanZook]

I'll see what I can come up with.

[Smarre]

@Student Is ignore_invalid_options the you are searching for?

I don’t know what stop_on_unknown() was for, it seems to be readonly on public API anyway, IMHO at least should be renamed.

[NathanZook]

Sorry, I've been sick. (Still am.)

Suppose we have something like this:

  opt :a, :name => 'a', :type => :int
  opt :b, :name => 'b', :type => :int

And an args string:

-a 2 -c 3 -b 5

With stop_on_unknown, we should have a return {a: 2} and no error. With ignore_invalid_options, we should have a return {a: 2, b: 5}.

As for my use case, consider the following:

I have a tool which operates in several different modes. The enactment of any particular mode is complex enough that it might be necessary to debug it. So I might define the type and debug options and process. Part of processing the type is to find out what options are available under the type mode. I am then ready to process the rest of the command line. But what I do not want is for a debug setting that applies to running in some mode to be applied while I'm setting up that mode. It's a lot of noise that the users should never have to endure.

I want stop_on_unknown.

[NathanZook]

cat bad_stop.rb:

require 'trollop'

opts = Trollop::options do
  opt :alpha, "The biggest, most important opt"
  stop_on_unknown
end

~/work/trollop ruby bad_stop.rb --beta
Error: unknown argument '--beta'.
Try --help for help.

This looks like a fail somewhere...

[kbrock]

@Student thanks - I'll look at this once the tests start passing again.

Something odd happening in the heads