Add additional test harness error handling to test_controller.

Add additional test harness error handling to test_controller.

We are seeing messed up output on firefox and the existing
eventhandler listinging for error is not sufficient for getting full
error output on firefox.
Committed: https://code.google.com/p/dart/source/detail?r=14253

[Mads Ager (google), 2012-10-30 09:06:47]

lgtm

However, could you loop in the authors of this code. I don't know the framework
this is running in and therefore cannot check the details.

[ricow1, 2012-10-30 09:08:30]

Adding Seattle people

[ricow1, 2012-10-30 09:37:39]

On 2012/10/30 09:08:30, ricow1 wrote:
> Adding Seattle people

Added console clearing on every invocation as well, PTAL

[ahe, 2012-10-30 09:48:51]

LGTM

https://codereview.chromium.org/11339039/diff/4001/pkg/unittest/test_controll...
File pkg/unittest/test_controller.js (right):

https://codereview.chromium.org/11339039/diff/4001/pkg/unittest/test_controll...
pkg/unittest/test_controller.js:12: console.clear();
Unfortunately, I only think this works in Firebug. But it is worth a try.

[ricow1, 2012-10-30 10:01:55]

On 2012/10/30 09:48:51, ahe wrote:
> LGTM
> 
>
https://codereview.chromium.org/11339039/diff/4001/pkg/unittest/test_controll...
> File pkg/unittest/test_controller.js (right):
> 
>
https://codereview.chromium.org/11339039/diff/4001/pkg/unittest/test_controll...
> pkg/unittest/test_controller.js:12: console.clear();
> Unfortunately, I only think this works in Firebug. But it is worth a try.
Added comment that this is Firebug only functionality.

[Siggi Cherem (dart-lang), 2012-10-30 17:36:52]

https://codereview.chromium.org/11339039/diff/6001/pkg/unittest/test_controll...
File pkg/unittest/test_controller.js (right):

https://codereview.chromium.org/11339039/diff/6001/pkg/unittest/test_controll...
pkg/unittest/test_controller.js:16: window.onerror = function (message, url,
lineNumber) {
could this be problematic in non-firefox browsers?

I feel some tension between this line and line 82 below. Should we have some
dynamic test to do one or the other based on the runtime browser, rather than
try to do both always?

[ricow1, 2012-10-31 10:17:53]

Siggi, thanks for the input, I still feel that we should make sure to bring all
possible errors that forward so that we can track down flaky issues.

https://codereview.chromium.org/11339039/diff/6001/pkg/unittest/test_controll...
File pkg/unittest/test_controller.js (right):

https://codereview.chromium.org/11339039/diff/6001/pkg/unittest/test_controll...
pkg/unittest/test_controller.js:16: window.onerror = function (message, url,
lineNumber) {
On 2012/10/30 17:36:53, Siggi Cherem (dart-lang) wrote:
> could this be problematic in non-firefox browsers?
> 
> I feel some tension between this line and line 82 below. Should we have some
> dynamic test to do one or the other based on the runtime browser, rather than
> try to do both always?
This is not firefox only:
http://www.w3.org/wiki/DOM/window.onerror
There seems to be some issues with some browsers where all errors will not be
caught by this. Always catching all errors is important in debugging issues with
out surrounding infrastructure (I would rather have the error twice than not at
all)

[Siggi Cherem (dart-lang), 2012-10-31 15:57:00]

On 2012/10/31 10:17:53, ricow1 wrote:
> Siggi, thanks for the input, I still feel that we should make sure to bring
all
> possible errors that forward so that we can track down flaky issues.
> 
>
https://codereview.chromium.org/11339039/diff/6001/pkg/unittest/test_controll...
> File pkg/unittest/test_controller.js (right):
> 
>
https://codereview.chromium.org/11339039/diff/6001/pkg/unittest/test_controll...
> pkg/unittest/test_controller.js:16: window.onerror = function (message, url,
> lineNumber) {
> On 2012/10/30 17:36:53, Siggi Cherem (dart-lang) wrote:
> > could this be problematic in non-firefox browsers?
> > 
> > I feel some tension between this line and line 82 below. Should we have some
> > dynamic test to do one or the other based on the runtime browser, rather
than
> > try to do both always?
> This is not firefox only:
> http://www.w3.org/wiki/DOM/window.onerror
> There seems to be some issues with some browsers where all errors will not be
> caught by this. Always catching all errors is important in debugging issues
with
> out surrounding infrastructure (I would rather have the error twice than not
at
> all)

I used to think that the signature of the callback (e.g. 1 vs 3 arguments) was
different on FF than Chrome, that was the source of my concern... I was just
playing around with it and it seems to work in both, good to know :)

Thanks Rico!