looking for leads as to why Infusion Builder tests are failing.

Colin Clark colinbdclark at gmail.com
Sun Nov 1 17:54:05 UTC 2009

Hi Laurel,

On 27-Oct-09, at 11:44 AM, Laurel A. Williams wrote:

> I'm trying to determine why I have some JS tests failing in the  
> Infusion Builder and wondering if anyone has any leads to point me  
> to the solution. In FireFox 3.5.3 on XP when the tests are run, a  
> bunch of them (specifically 4, 5, 6, 12, 13, 14)  fail, but only on  
> every second reload of the test page. In IE 8 on XP the tests fail  
> (4, 5, 6, 12, 13, 14 and also 8, 16) all the time. This is peculiar  
> enough to make me wonder in anyone else has seen anything similar.
> If you want to try the tests yourself, check out https://source.fluidproject.org/svn/incubator/custom-build/trunk/ 
>  and run infusion-builder/tests/html/customBuild-tests.html
> Thanks in advance for your thoughts.

This didn't ring any bells for me, but I did take some time to trace  
through your tests to learn more about what is going on. It looks  
pretty clear to me that a call to jQuery UI's simulate() is failing  
periodically.  The element you are trying to click on is definitely  
there, it's just not correctly receiving the event. I'm seeing this,  
for example, at line 404 of customBuild-tests.js.

I'd like to hear more from others who know simulate() better than I  
do. If necessary, an alternative approach is to programmatically  
adjust the "checked" attribute on your checkboxes instead of  
simulating click events.

An aside, I found your tests really difficult to debug, due to the  
strange way you've structured them. It looks like you committed a big  
refactoring to these tests back at r7673, and I'm not sure it was  
entirely for the best. By moving your test functions out into these  
"generateXyz" functions on a "pseudo-that" called "testingFunctions,"  
you make it harder to set breakpoints in a specific test, rather than  
for the general case of all similar tests. It also makes readability  
more difficult. This feels to me like an attempt to remove repetitive  
code, but at the wrong level of abstraction. I really like how  
thorough all of these tests are, but I find it hard to quickly see the  
difference between utility code, setup code, and the test bodies.

At some point in the future, let's consider what it would take to  
split up these tests into smaller units based. That said, I don't  
think we want to invest the time in reworking these tests yet again  
before we get the Builder released. It's something you'll probably  
encounter yourself as you use Firebug to track down this issue, but  
something that can wait for a bit.

Hope this helps,


Colin Clark
Technical Lead, Fluid Project

More information about the fluid-work mailing list