474984 - Move name calculation for GridCell element to new namerules infrastructure.

Status: RESOLVED FIXED

(Core :: Disability Access APIs, defect)

Description

[Marco Zehe (:MarcoZ)]

Spun off from bug 474281.

Comment 1

[Marco Zehe (:MarcoZ)]

Attachment: Patch

I have a feeling the test for the regular table cell isn't run, but am not certain. Alex, is this generally the right approach?

Comment 2

[alexander :surkov (:asurkov)]

Comment on attachment 358395 [details] [diff] [review]
Patch

Yes, generally this approach is correct.

>+    <ruleset id="tablecell">
>+      <rule fromsubtree="false"/>
>+    </ruleset>
>+
>+    <ruleset id="gridcell">
>+      <rule fromsubtree="true"/>
>+    </ruleset>

You don't need to introduce new rules for this case, I guess it's correct to use "htmlelm" ruleset, because I believe html:td will get the name from ARIA attributes and from related html:label.

Comment 3

[alexander :surkov (:asurkov)]

(In reply to comment #1)
> Created an attachment (id=358395) [details]
> Patch
> 
> I have a feeling the test for the regular table cell isn't run, but am not
> certain.

If they are run then they should be listed in passed tests

Comment 4

[Marco Zehe (:MarcoZ)]

(In reply to comment #2)
> >+    <ruleset id="tablecell">
> >+      <rule fromsubtree="false"/>
> >+    </ruleset>
> >+
> >+    <ruleset id="gridcell">
> >+      <rule fromsubtree="true"/>
> >+    </ruleset>
> 
> You don't need to introduce new rules for this case, I guess it's correct to
> use "htmlelm" ruleset, because I believe html:td will get the name from ARIA
> attributes and from related html:label.

I tried that, and when I do the regular table with the htmlelm ruleset, I get this output:

305 INFO TEST-PASS | chrome://mochikit/content/a11y/accessible/test_name_markup.html | Attribute 'aria-label' test. Wrong name of the accessible for [object HTMLTableCellElement]
306 ERROR TEST-UNEXPECTED-FAIL | chrome://mochikit/content/a11y/accessible/test_name_markup.html | Error thrown during test: attrValue is null - got 0, expected 1

And for the gridcell, I definitely need something to check the rule fromsubtree, and htmlelm doesn't include that.

Comment 5

[alexander :surkov (:asurkov)]

> I tried that, and when I do the regular table with the htmlelm ruleset, I get
> this output:
> 
> 305 INFO TEST-PASS |

I guess it's because you didn't add aria-label attirubte and other stuffs :)

> And for the gridcell, I definitely need something to check the rule
> fromsubtree, and htmlelm doesn't include that.

htmlctrl should work for gridcell

Comment 6

[Marco Zehe (:MarcoZ)]

Attachment: Patch2

Working tests.

Comment 7

[alexander :surkov (:asurkov)]

Comment on attachment 358418 [details] [diff] [review]
Patch2


>+
>+    <ruleset id="gridcell">
>+      <ruleset ref="htmlelm"/>
>+      <rule fromsubtree="true"/>
>+    </ruleset>

honestly, it's a bit strange it's working because order of rule elements is important and algorithm should check name from @title before subtree name. I'll try to look here deeply. But any way it's correct to use "htmlctrl" ruleset here.

r=me with the change.

Comment 8

[Marco Zehe (:MarcoZ)]

Pushed in changeset:
http://hg.mozilla.org/mozilla-central/rev/537af42cc65a

Comment 9

[alexander :surkov (:asurkov)]

(In reply to comment #7)

> honestly, it's a bit strange it's working because order of rule elements is
> important and algorithm should check name from @title before subtree name. I'll
> try to look here deeply. But any way it's correct to use "htmlctrl" ruleset
> here.

All is ok, with gridcell ruleset I get an error

not ok - Attribute 'title' test. Wrong name of the accessible for [object HTMLTableCellElement @ 0x61145f8 (native @ 0x4a88278)] got " This is a paragraph This is a link This is a list", expected "This is a paragraph This is a link This is a list"

Comment 10

[Marco Zehe (:MarcoZ)]

Yes, I ran into that test failure as well when I swapped things around a bit there. But the way I checked it in finally turned out to not cause any oranges.

Comment 11

[alexander :surkov (:asurkov)]

(In reply to comment #10)
>  But the way I checked it in finally turned out to not cause any oranges.

Sure because you used "htmlctrl" ruleset :). Just comment #6 confused me and made me think there is error in logic of name calculation test.