RE: camvas accessibility test results

Thanks for the report and update.

Do you think any of the reported failures should block our progress to Proposed Recommendation?

Can you update the test results at https://www.w3.org/wiki/HTML/Canvas_Task_Force/CR-Test?

/paulc

Paul Cotton, Microsoft Canada
17 Eleanor Drive, Ottawa, Ontario K2E 6A3
Tel: (425) 705-9596 Fax: (425) 936-7329

From: Richard Schwerdtfeger [mailto:schwer@us.ibm.com]
Sent: Wednesday, August 05, 2015 6:36 PM
To: Canvas; HTML Accessibility Task Force
Cc: Dominic Mazzoni; Rik Cabanier
Subject: camvas accessibility test results


I ran all the remaining canvas tests that indicated the test information needed to be filled in. There were a total of 5.

Dominique you should take a look at test case 5. Let me know if you want me to open a bug with Chrome.

Net, net: Chrome passes all cases except the ability to remove the hit region in test 5 below (I can still mouse over it and get an accessible from the point). This did not result in the first hit region being removed and the second hit region (bounds) were off in the accessibility API mapping.

Rik, A general bug I found with Firefox is that the paths drawn for the rectangles are consistently 4 pixels short on both width and height and is then reflected in the hit region and accessibility API mappings. This bug is a general canvas problem as it impacts drawing and not just hit testing. There may be a transform error.

The last two tests did not print out the results of the mouse overs to reflect the ids associated on a mouse "hit". The test cases are incomplete. I saw timeouts being generated.

I tested with:

  *   Version 46.0.2473.0 canary (64-bit)
  *   Firefox 39


1. TODO addHitRegion receives a dictionary object whose id is null, and a previous region has a null id

https://www.w3.org/wiki/HTML/Canvas_Task_Force/CR-Test#TODO_addHitRegion_receives_a_dictionary_object_whose_id_is_not_null.2C_and_has_a_non-null_control.2C_and_a_previous_region_having_the_same_control_exists


Passes Chrome
Passes Firefox

2. TODO addHitRegion receives a dictionary object whose id is null, and a previous region has a null id
https://www.w3.org/wiki/HTML/Canvas_Task_Force/CR-Test#TODO_addHitRegion_receives_a_dictionary_object_whose_id_is_null.2C_and_a_previous_region_has_a_null_id


Passes Chrome and Firefox

Also shows incomplete mapping in accessibility API as it should because the associated button is outside the canvas element region.


3. TODO addHitRegion receives a dictionary object whose id is not null, with a non-null region, and a previous region for the id is not null
https://www.w3.org/wiki/HTML/Canvas_Task_Force/CR-Test#TODO_addHitRegion_receives_a_dictionary_object_whose_id_is_not_null.2C_with_a_non-null_region.2C_and_a_previous_region_for_the_id_is_not_null

Passes in Chrome

Firefox has a hit region on the first region that is not right but then again there are two regions having the same id and the second overrides the first. This is really an error condition as you have two hit regions having the same id so you would not be able to tell which object the hit test applied to if you had an event handler on canvas. I would consider this an author error although it is necessary that this be tested.

Firefox has a width and height of the fallback button 4 pixels short of the 40 pixels.
When the second, green rectangle (button 2), is drawn it also is 4 pixels short on the width and height and the first red rectangle (button 1).


4. addHitRegion receives a dictionary object whose id is not null, with a non-null region, and a non-null control id where a previous non-null region, associated with this control, does exist in the hit region list but has a different id

https://www.w3.org/wiki/HTML/Canvas_Task_Force/CR-Test#addHitRegion_receives_a_dictionary_object_whose_id_is_not_null.2C_with_a_non-null_region.2C_and_a_non-null_control_id_where_a_previous_non-null_region.2C_associated_with_this_control.2C_does_exist_in_the_hit_region_list_but_has_a_different_id

This passed both Chrome and Firefox in terms of the accessibilitiy API mappings and locations on the screen. For some reason the test harness indicated a timeout although I am not sure why

5. addHitRegion is called twice. The first call receives a dictionary object whose id is not null, with a non-null region, and a non-null control. The control is associated with a fallback <div> having a WAI-ARIA role="button" and descendant text. The second call receives a dictionary object whose id is not null, with a non-null region, and a non-null control. Both hit regions have distinctly different ids and controls. The control is associated with a fallback <a> having and descendant text. Each hit region is at distinctly different locations within the canvas. The path used to create each hit region is drawn on the canvas. Call removeHitRegion() for the ID of the first region.

http://cptvitamin.github.io/wpt-canvas/2dcontext/hit-regions/removeHitRegion.TwoRegions-manual.html

Chrome:

- FAIL: the hit region start location of the red rectangle (top left is off) in the accessibility API. The hit regions should have been removed. It was not.
- the hit region start location of the green rectangle (top left is off) in the accessibility API

- The test file did not implement the mouse handler to print the id's of the hit when you mouse over the squares.

Firefox:

PASS except for the consistent 4 pixel errors in widths and heights.

The test harness indicated a time out. I am not sure why.

Rich Schwerdtfeger