Hi there Harris - I did want to mention how much the long hours you had put into looking into this issue
were appreciated and I'm sorry if my mail came across as dismissive. I hope we can keep a dialogue going
either here or in the channel so that we can deal with subtle issues like this in smaller units of work and
keep track of the directions we are all taking :)
Thanks to Justin for his helpful review and comments on this issue which I'll
deal with this evening.
Antranig.
On 30/05/2011 00:58, Antranig Basman wrote:
Hi Harris - thanks for looking into this bug. I suggest the following two
options:
4. Find where such bugs in IE9 are reported and make sure that the erroneous
behaviour of
element.getAttribute() is reported there (even though the historical rate of
resolving such bugs is
extremely low)
5. Switch to the use of $(el).prop("id") rather than the $(el).attr("id") that
we currently use.
It appears that jQuery's clear distinction since 1.6 of element attributes and
element properties allows
this bug to be resolved, since IE9's corruption only extends to namespaced
attribute values.
Please make sure when you address bug reports, that you take care to find the
root cause of the issue,
rather than just taking actions to make sure that the test case passes again.
Although Fluid now deprecates
the use of rsf:id and expects that users will not follow this strategy, it's
not acceptable for the
framework to fail in such a serious way when such namespaced attributes are
found in the markup (even if
they are not present for our own renderer, there are numerous other rendering
frameworks, for example,
Wicket, which may present just the same kind of namespaced id attribute in the
markup).
It's important to distinguish this kind of test case failure, example, from the
ones we see for the cases of
focus and blur on a few browsers. Having investigated these failures, we know
that these are benign, since
they do not cause failures which will be visible to users in real operation of
the framework, but are simply
artefacts of the limitations on fidelity of event simulation in test cases.
However, the failures in this test case are not benign, and correspond to real
fatal errors which will be
seen by users when operating the component on the actual platform. It's crucial
to make sure that these
errors can never occur in any circumstances, which is why I don't think the
three resolutions you have
proposed are sufficient.
I've pushed a branch which resolves our use of attr("id") and
getAttribute("id") framework-wide to the
appropriate use of properties to
https://github.com/amb26/infusion/commits/FLUID-3953 along with a pull
request - justino, could you review - I've tested on FF3, IE9 and Chrome so far,
Cheers,
Antranig.
_______________________________________________________
fluid-work mailing list - [email protected]
To unsubscribe, change settings or access archives,
see http://fluidproject.org/mailman/listinfo/fluid-work
