Page 1 of 1

What gives? Did the SalesForce application regress?

PostPosted: Thu May 16, 2013 2:36 pm
by pGustavo
We've got a SalesForce application and we have a beautiful 300+ line eValid script that has been fully converted into a structural-test format. It's been working great, but now it seems to fail. What gives? Did the SalesForce application regress? I've sent you the script! Help!

Re: What gives? Did the SalesForce application regress?

PostPosted: Thu May 16, 2013 3:10 pm
by eValid
We looked at the script you sent and indeed, your previously converted pure-structural script does indeed go into "non-responding" at a particular point in the playback.

The conversion to pure-structural is done very well, and usually that assures very reliable playback.

Except if the application changes in an essential way, which is what's happened here.

We looked at the script in "slow motion" -- using the single step feature and studying the EventLog we found that a script passage that looks like this:

Code: Select all
        IndexSet 0
        IndexFindElement 0 DOWN "name" "newContact" ""
        IndexSubmitClick 0 ""


throws an error because eValid can't find an element with the string "newContact" in a property called "name".

So we dug around and found that in this case the SalesForce application had changed. The "New Contact" button, which was implemented in straight HTML in the old version, is now implemented in JavaScript inside a frame. The new passage (which works fine) is this:

Code: Select all
        IndexSet 0
        IndexFindElement 0 DOWN "innerHTML" "New Contact" "number:1"
        IndexFollowLink 0 "number:1"


This kind of change in a well-tested product like SalesForce is unusual -- but obviously not impossible. The good news is that eValid did its job: it found the application regression.

_______________________________
eValid Structural Testing Team