114304 - dom html table testcase fails

Status: VERIFIED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Bernd]

go to http://mozilla.org/quality/ngdriver/suites/dom1/htbl050.html

Comment 1

[Fabian Guisset]

We should note that it's because we changed the attributes to be reported as
lower case instead of having the first letter upper-cased. The testcases should
use toLowerCase() or some such when comparing strings.

Comment 2

[Boris Zbarsky [:bzbarsky]]

All the DOM testcases need to be updated.... we now no longer case-fold
attribute values when returning them.

Comment 3

[Sivakiran Tummala]

looks like this bug is invalid.

Comment 4

[Bernd]

Huuuh, invalid how can this bug be invalid? Could you be more specific? Are the
testcases working for you? Didn't janc require to file against him? May be we
should reassign the bug. I looked trough bugzilla entries and could not find
another matching component. The testfailure has been introduced by changing the
dom behaviour and not running the necessary tests (fabian, bz: here I go with my
beat the dom people to run automated regression tests). The tests are public
available and should be also linked from the DOM documentation. As these tests
have a internal netscape counterpart (marvin), this should be fixed on both
sides of the firewall.

Comment 5

[Jan Carpenter]

I'm not sure why this bug is assigned to me, unless it's because I long ago
wrote the test case... I'm not on that team and am not sure who is in charge of
the DOM test cases now.  Reassigning to team manager to figure out who should
get this.

Comment 6

[Boris Zbarsky [:bzbarsky]]

I have to admit this was the first I had heard of DOM regression tests....  I
looked around, but I cannot find an automated driver for the tests; am I missing
something?

Is there a way new tests can be added, by the way?  If so, I may write some at
some point...

Comment 7

[Fabian Guisset]

Bernd, it makes no sense to build regression tests upon broken behavior (i.e.
returning a case-folded value instead of an all-lowercase value) in the
beginning. The tests mentionned in the URL are not regression tests, they are
compliance tests. The tests you pointed out incorrectly assumed that the correct
behavior was the one exhibited by Mozilla at the time the tests were written,
i.e. return a case-folded value. The bug fix did not cause a regression, it
caused the expected effects.

I can't see any obvious regression tests for the DOM. The most likely test suite
that should be run is the official W3C NIST DOM test suite, that should be
released soon.

Comment 8

[Bernd]

http://bugzilla.mozilla.org/show_bug.cgi?id=113174#c12 states that all responses
should be lower case. I filed bug 114842 to get this done.

Comment 9

[Bernd]

Attachment: patch

Fabian could you please r?

Comment 10

[Fabian Guisset]

Comment on attachment 61894 [details] [diff] [review]
patch

gosh! magnificient cleanup!
minor nits:

>+///*********************** BEGIN MAIN TEST BODY **************************
Why the triple slash?

>-	aValues = [ "Left", "Right", "Top", "Bottom", "Center", "Middle" ];
>+	aValues = [ "left", "right", "Top", "Bottom", "center", "Middle" ];
> 
Shouldn't everything be lowercase?

>-        var aBugNums = new Array("Bug #46954 - still open 10-10-2000");
>+        var aBugNums = new Array("Bug <a href='http://bugzilla.mozilla.org/show_bug.cgi?id=46954'>46954</a> - still open 10-10-2000");
> 
Please modify the date

Excellent!! r=fabian

Comment 11

[Bernd]

fix checked in 

Comment 12

[Sivakiran Tummala]

Bernd i said it is invalid because we do not case-fold the return value (Boris
comments) and the testcase needs to be corrected and this is not a bug in the
product.
Marking Verified.

Comment 13

[Sivakiran Tummala]

Bernd & Johnny, I will publish new testcases to the mozilla suites, the existing
testcases are according to DOM specs. I think the code changes are done to
reflect the same behaviour as in IE. I would like to have two copies of testcases 
1. To test our implementation
2. To test DOM Specs (standards).
The DOM Spec's say "The return value of an attribute that has a data type that
is a value list is always capitalized, independent of the case of the value in
the source document. For example, if the value of the align attribute on a P
element is "left" then it is returned as "Left". For attributes with the CDATA
data type, the case of the return value is that given in the source document." 
http://www.w3.org/TR/2000/WD-DOM-Level-2-HTML-20001113/html.html#ID-642250288.

Comment 14

[Boris Zbarsky [:bzbarsky]]

sivarikan, take a look at 
http://bugzilla.mozilla.org/show_bug.cgi?id=113174#c12, please.  The spec is 
apparently wrong in some respects and will be changed...