Signatures

No problem to determine incorrect node signatures at the call, process, and process_block levels. The output should notify the node, and ideally the join causing the problem. Once a trace is in place, the join could be found by searching back in the trace for the specified inputs:

trace.reverse_each
  if node.joins has the present task
  and the inputs come from the join (match object, or find object in an array)
  then consider the join found (or search everything... maybe the issue is upstream?)

And once you have the join, help can be provided to suggest what needs to be done to fix the error.

Replay

A breakpoint could be added to step through the workflow. In this case, or perhaps generally, the trace should be saved to disk allowing quick replay. Will not work with objects that cannot be serialized, and that's ok. Simply encourages people not to pass around IO or blocks, or to fix YAML.

This will cause an error because a single default const_name is not determined for the overlapping load paths:

TAP_LOAD_PATHS='[lib, lib/tap]' ruby -I../lazydoc/lib -I../configurable/lib bin/tap run --help

The error message needs to be cleaned up.

It would be helpful if apps allowed middleware to have before/after hooks for a run:

def run
  synchronize do
    return self unless state == State::READY
    @state = State::RUN
  end

  middleware.each do |middleware|
    if middleware.respond_to?(:before_run)
      middleware.before_run
    end
  end

  begin
    while state == State::RUN
      break unless entry = queue.deq
      dispatch(*entry)
    end
  rescue(TerminateError)
    # gracefully fail for termination errors
    queue.unshift(*entry)
  ensure
    synchronize { @state = State::READY }
  end

  middleware.each do |middleware|
    if middleware.respond_to?(:after_run)
      middleware.after_run
    end
  end

  self
end

Currently the app behavior is reversed. Empty joins should be respected, while nil joins should invoke the defaults. Easy fix, but will require some trickery for reassigning joins.

Also be sure to fix Null in this case; afterwards the null task should provide a sensible message for:

% tap run -- null --: null

A preview config that can prints sh commands rather than running them. In this way a task can generate a script on the fly.

Or maybe have an advanced Rap::Utils that requires highline and net-ssh to perform this functionality.

In principle it should be able to activate the wrong gem from Env when multiple versions of the gem are present on a box. This is because gems are activated by the require for a Constant, and all the Constant has to go off of is the relative path.

Config for how long to sleep.

http://bka-bonn.de/wordpress/index.php/2009/06/01/determine-os-with-ruby/ (see comments)

require ‘rbconfig’
puts RbConfig::CONFIG[’target_os’]
=> “linux”

This is where the update needs to happen.

def match_platform?(*platforms)
      platforms.each do |platform|
        platform.to_s =~ /^(non_)?(.*)/

        non = true if $1
        match_platform = !RUBY_PLATFORM.index($2).nil?
        return false unless (non && !match_platform) || (!non && match_platform)
      end

      true
    end