Author Topic: Tests - they seem to get a bit lost! (Read 7895 times)

ronnie · « **on:** January 07, 2006, 10:56:22 am »

What are the guidelines for maintaining tests in EA?
I have a problem that they seem to get a bit lost - it's not obvious when an object has some tests - it seems that we have to select each object with the "Test Cases" pane open and see if there are some there. This is hampered by a missing aggregated 'All Tests' tab - so you have to flick between all of the test types as well.
So I resorted to an advanced search - that's better - I've found all of the objects within a section of the tree that have test objects but as I use the keyboard goig up/down with the arrow keys in the search results list doesn't update the test pane (clicking does!).

Ok - So finally I want to run a tracability matrix to see if all of my requirement objects have tests and stuff like that - but they don't seem to appear in there

.

I am resorting to build test-cases packaged in use-cases where the use-case has a dependency stereotyped as "tests" (don't feel that is the right link) to the requirement and then putting a bunch of the tests into the use-case.

I think this is going to help a little but I think there is still a load of useful stuff that could be done with the tests (maybe there already is and I am missing it

)

Eve · « **Reply #1 on:** January 08, 2006, 01:08:44 pm »

One tip. Open a diagram properties and check 'Show Testing'.

This will show all of your tests for each element on the diagram.

Bruno.Cossi · « **Reply #2 on:** January 08, 2006, 03:20:58 pm »

Hi Simon,

that is very good, but it does not work for Use Cases (unless one uses the Rectangle notation).

Thanks,
Bruno

Quote

One tip. Open a diagram properties and check 'Show Testing'.

This will show all of your tests for each element on the diagram.

Graham_Moir · « **Reply #3 on:** January 12, 2006, 09:29:35 am »

What am I doing wrong ? I can't get the test details to appear for an element on the diagram.

I have "Show testing" selected, a class with associated tests in the test pane, but no indication on the diagram that these exist.

thanks

Graham_Moir · « **Reply #4 on:** January 12, 2006, 10:27:26 am »

I take it back - they just appeared. However there was a significant lag. I saved the diagram, toggled the "Show testing" property all to no effect, then about half an hour later magically the test details appeared. Looks as though it could be a bug in the diagram refresh logic.

Graham_Moir · « **Reply #5 on:** January 13, 2006, 07:38:09 am »

Would it be a sensible suggestion to consider a "shape script" that would display a small icon on the element in the diagram if tests exist for the element ?

Or is this a silly suggestion ?

thanks

Bruno.Cossi · « **Reply #6 on:** January 13, 2006, 09:59:36 am »

Actually what I would like is for the tests to appear in the Element Browser window. The same for Element Defects and Changes as well.

What do you think?

Bruno

ronnie · « **Reply #7 on:** January 13, 2006, 10:46:11 am »

I agree - if they ended up as regular elements then they could be dependancy traced and I could control the reports much better!
Also then I could maybe point multiple requirements and other stuff at the same test - but isn't that what a TestCase is? And how would the testing pane then work?

There seems to be very little in the literature I have regarding how tests and testing fits into the UML. If I am going to put requirements in and trace things through use cases to code and deployment surely I want to link all that back so that I can test the final system against the initial requirements and verify with a measurable test-requirement coverage that we are 'done'. I just can't see how that can be acheived with EA at the moment.

I am starting to think that an external system that serves the early and late part (requirements gathering and testing against them) which feeds into EA through the DB or automation interface might be the best shot - has anyone done silimar?

Bruno.Cossi · « **Reply #8 on:** January 13, 2006, 11:29:26 am »

We have connected EA with TestDirector, that works just fine for us.

On a different note, the way I have sometimes use EA is to use the TestCase element (stereotyped Use Case). I would connect the TestCase to whichever elements it would cover, then I would create the actual tests under the TestCase. Not the most elegant solution, but it worked.

Bruno

Quote

I agree - if they ended up as regular elements then they could be dependancy traced and I could control the reports much better!
Also then I could maybe point multiple requirements and other stuff at the same test - but isn't that what a TestCase is? And how would the testing pane then work?

There seems to be very little in the literature I have regarding how tests and testing fits into the UML. If I am going to put requirements in and trace things through use cases to code and deployment surely I want to link all that back so that I can test the final system against the initial requirements and verify with a measurable test-requirement coverage that we are 'done'. I just can't see how that can be acheived with EA at the moment.

I am starting to think that an external system that serves the early and late part (requirements gathering and testing against them) which feeds into EA through the DB or automation interface might be the best shot - has anyone done silimar?

dmantilla · « **Reply #9 on:** January 19, 2006, 10:19:28 am »

Other alternative is to query the EA Repository (MS Access, Oracle or whatever you are using) directly by using a SQL Statement joining t_package, t_object and t_objecttests tables.

Hope this helps.

Daniel

ronnie · « **Reply #10 on:** January 20, 2006, 03:43:51 am »

Been looking in there recently - I have a test that failed and now passes - just wondering how that can be captured.

Firstly I think there needs to either be a way of storing multiple runs of a test or at least a 'duplicate test' option...

Anyway - now I have 2 test results - one that failed on Tuesday and one that passed today - but I also print a report of all failed tests as a quick heads-up on what's not working. Was immagining adding a new status type like "Used to Fail" which I could filter out but can't find where the test result list comes from

Alternatively I'm thinking of adding a type but it doesn't feel right (Standard, Load, Regression, Obsolete) - nah!

What are other people doing in these cases?

Ronnie

mcoletti · « **Reply #11 on:** March 30, 2006, 03:18:51 am »

I am using EA for the first time to model a test plan. The system to be tested is an electronic surveillance system.

I have found this discussion useful, and followed this path:

I create a "test case" element for each test to be performed;
I create association having the "test case" as a source, and the element to be tested as the destination. This association is stereotyped as "testing device" if the destination element is not the object of testing but is required to perform the test.
I have customised the testing RTF template, adding diagrams images and details about the connectors.

In this way I can code only once the test for similar devices; the RTF report provides a "testing" guide for the tester, with a visual indication of the devices to be tested.

The same EA model holds the topology of the systems, so it is useful also as a reference.

If a test has a "partial succes" result (i.e. if I am testing 20 cameras, 2 of which aren't working) I will import the test on the element that failed, changing the classifier status to "Failed".

If I want to be very formal, I can copy the test from the test case to each tested component. The tester will sign the report, certifying that all the planned components were tested.

I will provide later the results of this "usage profile".

Sparx Systems Forum

News: