# When test just can't be automated

**URL:** <https://www.ruby-forum.com/t/when-test-just-cant-be-automated/193548>\
**Category:** Ruby\
**Created:** [August 19, 2010, 2:46pm UTC](https://www.ruby-forum.com/t/when-test-just-cant-be-automated/193548 "2010-08-19T14:46:39Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![fourchette](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/f/ac91a4/32.png) [@fourchette](https://www.ruby-forum.com/u/fourchette)\
**Post date:** [August 19, 2010, 2:46pm UTC](https://www.ruby-forum.com/t/when-test-just-cant-be-automated/193548/1 "2010-08-19T14:46:39Z")

</div>

hi there,

i do hope i’m posting at the right place.

I am supposed to test some sort of a network device let’s say. i want to  
do it with ruby unit testing framework.

Basically you can send it commands using TCP/IP in a custom language  
made for the device. It also sends back results.

so basically pretty much a lot of commands can be tested.

But some can not be completly automated.

for instance, this little devices beeps. There is a command to disable  
sound. You can check that the command reported something like “sound  
successfully disabled”

but the ruby script can’t actually “hear” that the device has stopped  
emitting sound after the command has been issued.

So i was thinking of having the test script do something like

init device  
send a command that should make the device beep  
–\> at this point have the script stop and ask tester for confirmation  
that a sound has been heard (ex: by hitting SPACE the test confirms that  
she heard the sound)  
send a command to disable sound  
send a command that should normally have the device beeping  
–\> again ask the test to confirm no sound was heard

what’s the best way to implement this ? do you have a suggestion ? (the  
testers using the scripts aren’t techies… :())

thanks  
fourchette

---

<div class="post-metadata">

**Author:** ![fourchette](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/f/ac91a4/32.png) [@fourchette](https://www.ruby-forum.com/u/fourchette)\
**Post date:** [August 19, 2010, 3:53pm UTC](https://www.ruby-forum.com/t/when-test-just-cant-be-automated/193548/2 "2010-08-19T15:53:59Z")

</div>

Best way:  
Automate the tests that are automatable, have them generate a report  
that  
the tester reviews/signs off on.

Then have the remaining tests performed manually, or semi-assisted with  
scripts (in the manner that  
you describe below). Its rarely useful to have tests that can be fully  
automated, interrupted by manual  
prompts (it will prevent other tests from running).

regards,  
Richard.

On Thu, Aug 19, 2010 at 1:46 PM, Fourchette Fourchette \<

---

<div class="post-metadata">

**Author:** ![fourchette](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/f/ac91a4/32.png) [@fourchette](https://www.ruby-forum.com/u/fourchette)\
**Post date:** [August 19, 2010, 4:19pm UTC](https://www.ruby-forum.com/t/when-test-just-cant-be-automated/193548/3 "2010-08-19T16:19:18Z")

</div>

> what’s the best way to implement this ? do you have a suggestion ? (the  
> testers using the scripts aren’t techies… :())

If they can’t cope with a text script with Y/N prompts, maybe you could  
pop up a dialog box (e.g. with Tk). Google “ruby tk examples” or  
similar.

Maybe you’re worried about CLI interaction because you’re using a test  
framework like Test::Unit. Tk should work fine there too. Otherwise you  
can write a standalone test program which doesn’t use the test runner,  
so you’re in control of the console at all times. (You could use the  
framework’s assertion methods still)

---

<div class="post-metadata">

**Author:** ![fourchette](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/f/ac91a4/32.png) [@fourchette](https://www.ruby-forum.com/u/fourchette)\
**Post date:** [August 23, 2010, 11:01am UTC](https://www.ruby-forum.com/t/when-test-just-cant-be-automated/193548/4 "2010-08-23T11:01:06Z")

</div>

Thank you for the feedback. the trouble is i think most tests are 10%  
manual + 90% automatic. Basically it’s a complicated setup but  
quick&easy tests (“setup” as in “setup VS tearDown”). I think setup can  
be nearly totally automated, most of the tests too, but the expected  
behavior (assertions) can only be observed by a human person.

i’m not very familiar with tk and i’m affraid it’s going to be hard to  
have it adopted internally (heavy history of Microsft Visual Basic  
GUIs).

Can the GUI be easily web-based? or such an approach would just be  
overkill ? (compared to a tk UI i mean)

thanks

alex