cumulativeOffset bug and failing IE tests

[Brad Greenlee]

I spend the morning tracking down an IE bug we were seeing only to
find it in Trac:
http://dev.rubyonrails.org/ticket/5515

I applied the suggested fix and everything works great now.

I was disappointed to see that the patch was submitted a year ago, but
the ticket was closed because no unit tests were written. So I set out
to write the tests myself. However, I've discovered that a number of
tests are already failing on IE (testPositionedOffset,
testViewportOffset, and testOffsetParent in dom, and testWithin in
position). Before I blow more time on this, I thought I'd check to see
what the deal was. Is this a known issue? Is someone already working
on this? Are these just things that IE doesn't support?

In the end, I would love to just see that patch applied, if only to
save some other poor soul from banging his head on the wall for hours
like I did. I'm happy to do what I can to make that happen.

Cheers,

Brad

[Thomas Fuchs]

In general, unit tests shouldn't fail. Please help us by fixing them. :)

Best,
Thomas

[Grant Hollingworth]

* Thomas Fuchs <t.f...@wollzelle.com> [2007-05-25 03:06]:

>In general, unit tests shouldn't fail. Please help us by fixing them. :)

I've run into some failing tests on older versions of Safari.

* testFormActivating fails on 2.0.2.

* the current Prototype.ScriptFragment crashes versions of Safari before 2.0.2 (see #8332).

* testElementReplace errors on 2.0.

I saw a mention in a list post that Prototype should work on Safari 1.3.2. Is there a list of supported browsers?

Does Prototype have a continuous integration server?

[Mislav Marohnić]

It should work, yeah. If you know how to fix it, please do. Otherwise, provide as much info as you can on the tracker.

Does Prototype have a continuous integration server?

Haha :)

In theory it could be set up, yeah - single OS X with VMware or Parallels. Until someone does, we do it the old-fashioned way :)

[Brad Greenlee]

At the risk of veering completely off-topic, is it possible to run IE
tests directly from OS X w/ Parallels/Windows installed? I assumed no,
and went ahead and set up a testing environment in Windows.

-b

On May 25, 9:35 am, "Mislav Marohnić" <mislav.maroh...@gmail.com>
wrote:
> On 5/25/07, Grant Hollingworth <g...@antiflux.org> wrote:
>
>
>
>
>
> > * Thomas Fuchs <t.fu...@wollzelle.com> [2007-05-25 03:06]:

[Mislav Marohnić]

On 5/25/07, Brad Greenlee <bgre...@gmail.com> wrote:

At the risk of veering completely off-topic, is it possible to run IE
tests directly from OS X w/ Parallels/Windows installed? I assumed no,
and went ahead and set up a testing environment in Windows.

Everything is possible. For instance, I'm testing in IE on Linux without virtualization.

[Brad Greenlee]

Is there documentation somewhere on how to set this up? (OS X/
Parallels/Windows, although I'm curious as to how you're testing IE on
Linux).

Cheers,

-b

On May 25, 9:59 am, "Mislav Marohnić" <mislav.maroh...@gmail.com>
wrote:

[Andrew Sumin]

I have the same problem.
Please see the save Dojo function it uses VBS object for IE and it
works perfect.

[Thomas Fuchs]

With WINE? Note that IE sometimes will misbehave, that is work
differently from when installed on Windows/run in virtualization
(i've seen this with my own eyes, no urban rumors here).

Anyway, we've also had a case where IE was working when running
natively; but had a bug when running on Parallels (probably due to
some kind of issue with the virtualization _or_ with a driver thats
provided by Parallels).

Note that these cases should be _very_rare, so in general you can
rely on the virtualization. I'm just unsure about a WINE-based
environment, as some things will differ (like font rendering, etc);
although there are most likely no issues with JavaScript unit testing.

Best,
Thomas

[Mislav Marohnić]

On 5/27/07, Thomas Fuchs <t.f...@wollzelle.com> wrote:

With WINE? Note that IE sometimes will misbehave, that is work
differently from when installed on Windows/run in virtualization
(i've seen this with my own eyes, no urban rumors here).

Yeah, heard about that too. So far there have been no differences in unit tests results when I do it with Wine or on native Windows.

[Thomas Fuchs]

Good to hear. Seems to be a sign of how mature WINE is. :)

That being said, continous integration could also be done with more
than one machine; all machines listen to test requests (webrick?) and
start stuff accordingly; then report back to the server machine.

Shouldn't be to difficult to implement; and we could even have Safari
1.3 and 2.0 testing going; implementing a commit hook or somesuch in
svn shouldn't be that difficult too. Error reporting can send mails
into prototype-core. I'm all for it. :)

Best,
Thomas

[Michael Peters]

Thomas Fuchs wrote:

> Shouldn't be to difficult to implement; and we could even have Safari
> 1.3 and 2.0 testing going; implementing a commit hook or somesuch in
> svn shouldn't be that difficult too. Error reporting can send mails
> into prototype-core. I'm all for it. :)

I don't have a completely un-biased viewpoint, but you might want to check out
Smolder - http://sourceforge.net/projects/smolder (btw, Smolder uses Prototype
and Scriptaculous) as a way to provide email notifications, rss feeds and nice
reports of test failures.

I'm currently doing some work on it as part of a Perl Foundation grant that
should be finished in a month or so. After that I'd be more than willing to help
set this up.

We use this at my place of employment and for other projects I've worked on with
some automated testing machines and it works really well. I haven't done it for
a JS based project yet, but it shouldn't be too hard. We just need to get the
testing results into TAP (Test Anything Protocol) -
http://search.cpan.org/~petdance/TAP-1.00/TAP.pm

--
Michael Peters
Developer
Plus Three, LP