rspec users - Oct 2007 - Specing exit codes

I am writing a small ruby script that will be accepting input from
postfix''s pipe command (ie, not running via the shell, directly
executing).

One of the things I need to do it spec the exit codes to make sure I
am returing the correct exit codes for each condition as Postfix will
then return SMTP errors as appropriate.

I have two files that concern this bit of the program, init.rb and init_spec.rb.

init.rb right now looks like this:
------------------------
class Init
  exit 1
end
------------------------

init_spec.rb looks like this:
------------------------
require ''spec''
require ''systemu''
require ''init''

describe Init do
  it "should exit on status code 1 without parameters"
    command = "ruby mail_dump/init.rb" # not portable
    status, stdout, stderr = systemu command
    status.should == # what do I put here?
  end
end
------------------------

I have tried a number of things, from trying to stub exit to aliasing
kernel.exit to something else and replacing it... all without joy.

The spec runs and hits the "exit 1" in init.rb and does what it is
mean to do... exit.  But that also exits RSpec and so the test is
never run!

The only thing I found DID work is if I alias Kernel.exit inside the
init.rb file to "real_exit" and then redefine Kernel exit like so:

class Object
  module Kernel
    alias real_exit exit
    def exit(arg)
      return true if arg == 1
    end
  end
end

and then test exit by mocking it and making sure it returns true...
but a spec that has to modify the test code isn''t going to scale too
well... and this doesn''t seem right.

Has anyone else had this problem?  How did you solve it?

thanks..

Mikel

On Oct 6, 2007, at 11:31 PM, Mikel Lindsaar wrote:
> I am writing a small ruby script that will be accepting input from
> postfix''s pipe command (ie, not running via the shell, directly
> executing).
>
> One of the things I need to do it spec the exit codes to make sure I
> am returing the correct exit codes for each condition as Postfix will
> then return SMTP errors as appropriate.
>
> I have two files that concern this bit of the program, init.rb and  
> init_spec.rb.
>
> init.rb right now looks like this:
> ------------------------
> class Init
>   exit 1
> end
> ------------------------
>
> init_spec.rb looks like this:
> ------------------------
> require ''spec''
> require ''systemu''
> require ''init''
>
> describe Init do
>   it "should exit on status code 1 without parameters"
>     command = "ruby mail_dump/init.rb" # not portable
>     status, stdout, stderr = systemu command
>     status.should == # what do I put here?
>   end
> end
> ------------------------
>
> I have tried a number of things, from trying to stub exit to aliasing
> kernel.exit to something else and replacing it... all without joy.
>
> The spec runs and hits the "exit 1" in init.rb and does what it
is
> mean to do... exit.  But that also exits RSpec and so the test is
> never run!
>
> The only thing I found DID work is if I alias Kernel.exit inside the
> init.rb file to "real_exit" and then redefine Kernel exit like
so:
>
> class Object
>   module Kernel
>     alias real_exit exit
>     def exit(arg)
>       return true if arg == 1
>     end
>   end
> end
>
> and then test exit by mocking it and making sure it returns true...
> but a spec that has to modify the test code isn''t going to scale
too
> well... and this doesn''t seem right.
>
> Has anyone else had this problem?  How did you solve it?
I had a similar problem a while back.  Aslak tipped me off to  
dependency injection.  Something like this:

describe Init do
   before :each do
     @kernel = mock(Kernel)
     @kernel.stub!(:exit).and_return 1
   end

   it "should exit on status code 1 without parameters"
     init = Init.new(@kernel)
     init.start
     init.status.should == 1
   end
end

module Init
   def start(kernel=Kernel)
     kernel.start
   end
end


Scott

Hey Scott,

Thanks for the speedy answer.... dependency injection is new to me, so
I am getting my head around it all.

I put your solution in and now I am getting:

1)
SystemExit in ''Init handling exit codes before(:all)''
exit
./init.rb:8:in `exit''
./init.rb:8:in `initialize''

2)
SystemExit in ''Init handling exit codes after(:all)''
exit
./init.rb:8:in `exit''
./init.rb:8:in `initialize''

Finished in 0.020305 seconds

26 examples, 2 failures


Which seems on the right path... how do I catch SystemExit?

I don''t have a before :all or after :all defined explicitly...

Regards

Mikel


On 10/7/07, Scott Taylor <mailing_lists at railsnewbie.com>
wrote:
>
> On Oct 6, 2007, at 11:31 PM, Mikel Lindsaar wrote:
>
> > I am writing a small ruby script that will be accepting input from
> > postfix''s pipe command (ie, not running via the shell,
directly
> > executing).
> >
> > One of the things I need to do it spec the exit codes to make sure I
> > am returing the correct exit codes for each condition as Postfix will
> > then return SMTP errors as appropriate.
> >
> > I have two files that concern this bit of the program, init.rb and
> > init_spec.rb.
> >
> > init.rb right now looks like this:
> > ------------------------
> > class Init
> >   exit 1
> > end
> > ------------------------
> >
> > init_spec.rb looks like this:
> > ------------------------
> > require ''spec''
> > require ''systemu''
> > require ''init''
> >
> > describe Init do
> >   it "should exit on status code 1 without parameters"
> >     command = "ruby mail_dump/init.rb" # not portable
> >     status, stdout, stderr = systemu command
> >     status.should == # what do I put here?
> >   end
> > end
> > ------------------------
> >
> > I have tried a number of things, from trying to stub exit to aliasing
> > kernel.exit to something else and replacing it... all without joy.
> >
> > The spec runs and hits the "exit 1" in init.rb and does what
it is
> > mean to do... exit.  But that also exits RSpec and so the test is
> > never run!
> >
> > The only thing I found DID work is if I alias Kernel.exit inside the
> > init.rb file to "real_exit" and then redefine Kernel exit
like so:
> >
> > class Object
> >   module Kernel
> >     alias real_exit exit
> >     def exit(arg)
> >       return true if arg == 1
> >     end
> >   end
> > end
> >
> > and then test exit by mocking it and making sure it returns true...
> > but a spec that has to modify the test code isn''t going to
scale too
> > well... and this doesn''t seem right.
> >
> > Has anyone else had this problem?  How did you solve it?
>
> I had a similar problem a while back.  Aslak tipped me off to
> dependency injection.  Something like this:
>
> describe Init do
>    before :each do
>      @kernel = mock(Kernel)
>      @kernel.stub!(:exit).and_return 1
>    end
>
>    it "should exit on status code 1 without parameters"
>      init = Init.new(@kernel)
>      init.start
>      init.status.should == 1
>    end
> end
>
> module Init
>    def start(kernel=Kernel)
>      kernel.start
>    end
> end
>
>
> Scott
>
> _______________________________________________
> rspec-users mailing list
> rspec-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/rspec-users
>