1261377 - With current child caching reorganization, NVDA is no longer able to render some content

Status: VERIFIED WORKSFORME

(Core :: Disability Access APIs, defect)

Description

[Marco Zehe (:MarcoZ)]

Seen in the 2016-03-31 build.

STR:
1. With NVDA running, go to https://bugzilla.mozilla.org, and click the "My Bugs" link.
2. A new page appears that has a table of all bugs one filed or is assigned to, a standard Bugzilla query. Press t when NVDA's virtual buffer is active to go to that table.

Expected: NVDA should see the table that contains all the found bugs. In my case over 130. It should pull the virtual cursor there.
Actual: A few days ago, NVDA was still seeing that table, but in the 2016-03-31 build, it is not. NVDA will say "No more tables found".

3. Press NVDA+F5 to refresh the virtual buffer.

Result: Still, the table is not found.

4. Inspect the accessible tree. After a text node saying "xxx bugs found.", you'll find said table and its children. But for some reason, its properties seem to have changed in a way that NVDA can no longer render it.

Comment 1

[alexander :surkov (:asurkov)]

Attachment: patch

Comment 2

[alexander :surkov (:asurkov)]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=2cc2f27a7845

Comment 3

[Marco Zehe (:MarcoZ)]

Comment on attachment 8737268 [details] [diff] [review]
patch

This doesn't fix the bug. The table gets inserted into the tree, but NVDA still doesn't add it to its virtual buffer, even after a v buffer refresh. NVDA's developer info for the table accessible, as navigated to via NVDA object navigator, looks like this:

INFO - globalCommands.GlobalCommands.script_navigatorObject_devInfo (09:44:34):
Developer info for navigator object:
name: None
role: ROLE_TABLE
states: STATE_READONLY
isFocusable: False
hasFocus: False
Python object: <NVDAObjects.IAccessible.mozilla.BrokenFocusedState object at 0x0F29C830>
Python class mro: (<class 'NVDAObjects.IAccessible.mozilla.BrokenFocusedState'>, <class 'NVDAObjects.IAccessible.mozilla.Mozilla'>, <class 'NVDAObjects.IAccessible.ia2Web.Ia2Web'>, <class 'NVDAObjects.IAccessible.IAccessible'>, <class 'NVDAObjects.window.Window'>, <class 'NVDAObjects.NVDAObject'>, <class 'baseObject.ScriptableObject'>, <class 'baseObject.AutoPropertyObject'>, <type 'object'>)
description: u''
location: (474, -6996, 2449, 8794)
value: None
appModule: <'firefox' (appName u'firefox', process ID 596) at address 5674cd0>
appModule.productName: u'Nightly'
appModule.productVersion: u'48.0a1'
TextInfo: <class 'NVDAObjects.NVDAObjectTextInfo'>
windowHandle: 3671358L
windowClassName: u'MozillaWindowClass'
windowControlID: 0
windowStyle: 399441920
windowThreadID: 5588
windowText: u'Bug List - Nightly'
displayText: u''
IAccessibleObject: <POINTER(IAccessible2) ptr=0x80a4f94 at 542c3a0>
IAccessibleChildID: 0
IAccessible event parameters: windowHandle=3671358L, objectID=-4, childID=-1291
IAccessible accName: None
IAccessible accRole: ROLE_SYSTEM_TABLE
IAccessible accState: STATE_SYSTEM_READONLY, STATE_SYSTEM_VALID (64)
IAccessible accDescription: u''
IAccessible accValue: None
IAccessible2 windowHandle: 3671358
IAccessible2 uniqueID: -1291
IAccessible2 role: ROLE_SYSTEM_TABLE
IAccessible2 states: IA2_STATE_OPAQUE (1024)
IAccessible2 attributes: u'margin-left:0px;text-align:start;text-indent:0px;margin-right:0px;tag:table;class:bz_buglist sortable;margin-top:0px;margin-bottom:0px;display:table;'

Jamie, see anything suspicious why you don't add it to your v buffers?

Comment 4

[Marco Zehe (:MarcoZ)]

FYI, the above is from a try build made in comment #2. By comparison, using the regular nightly build from 2016-04-02, the info looks as follows:

Developer info for navigator object:
name: None
role: ROLE_TABLE
states: STATE_READONLY, STATE_OFFSCREEN
isFocusable: False
hasFocus: False
Python object: <NVDAObjects.IAccessible.mozilla.BrokenFocusedState object at 0x059FFA70>
Python class mro: (<class 'NVDAObjects.IAccessible.mozilla.BrokenFocusedState'>, <class 'NVDAObjects.IAccessible.mozilla.Mozilla'>, <class 'NVDAObjects.IAccessible.ia2Web.Ia2Web'>, <class 'NVDAObjects.IAccessible.IAccessible'>, <class 'NVDAObjects.window.Window'>, <class 'NVDAObjects.NVDAObject'>, <class 'baseObject.ScriptableObject'>, <class 'baseObject.AutoPropertyObject'>, <type 'object'>)
description: u''
location: (474, 1951, 1738, 88)
value: None
appModule: <'firefox' (appName u'firefox', process ID 1900) at address 5679670>
appModule.productName: u'Nightly'
appModule.productVersion: u'48.0a1'
TextInfo: <class 'NVDAObjects.NVDAObjectTextInfo'>
windowHandle: 1966846L
windowClassName: u'MozillaWindowClass'
windowControlID: 0
windowStyle: 399441920
windowThreadID: 6272
windowText: u'Bug List - Nightly'
displayText: u''
IAccessibleObject: <POINTER(IAccessible2) ptr=0xc23900c at 5425f80>
IAccessibleChildID: 0
IAccessible event parameters: windowHandle=1966846L, objectID=-4, childID=-899
IAccessible accName: None
IAccessible accRole: ROLE_SYSTEM_TABLE
IAccessible accState: STATE_SYSTEM_OFFSCREEN, STATE_SYSTEM_READONLY, STATE_SYSTEM_VALID (65600)
IAccessible accDescription: u''
IAccessible accValue: None
IAccessible2 windowHandle: 1966846
IAccessible2 uniqueID: -899
IAccessible2 role: ROLE_SYSTEM_TABLE
IAccessible2 states: IA2_STATE_OPAQUE (1024)
IAccessible2 attributes: u'margin-left:0px;text-align:start;layout-guess:true;text-indent:0px;margin-right:0px;tag:table;margin-top:0px;margin-bottom:0px;display:table;'

So I suspect it has something to do with the table becoming visible, and us not noticing it.

Comment 5

[alexander :surkov (:asurkov)]

(In reply to Marco Zehe (:MarcoZ) from comment #3)
> Comment on attachment 8737268 [details] [diff] [review]
> patch
> 
> This doesn't fix the bug.

We have one more issue on this bug I think. I don't guarantee but I think this bug cannot be fixed without taking this patch, refer to a last test in the mochitest which fails.

Comment 6

[Marco Zehe (:MarcoZ)]

Comment on attachment 8737268 [details] [diff] [review]
patch

Very well then, r=me for this one.

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/954444d01cd1

Comment 8

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/954444d01cd1

Comment 9

[Marco Zehe (:MarcoZ)]

Status as of the 2016-04-06 nightly build: The very first time after the Nightly update I brought up the "My Bugs" page, it was complete, with a table of 133 found bugs. Subsequent calls of that page later during the day again showed this particular bug, and even refreshing the page two or three times (not the buffer, but the page) didn't produce any changes. So this bug is still present.

Comment 10

[alexander :surkov (:asurkov)]

Marco, can you check if bug 1268677 improves it at all?

Comment 11

[Marco Zehe (:MarcoZ)]

Yes, as far as I can tell, the try server build in bug 1268677 fixes this problem and also some other inconsistencies in IRCCloud and elsewhere I've been seeing. \-O-/ Marking this bug as a dependency, so when that other one lands, this one can be marked WFM or so.

Comment 12

[alexander :surkov (:asurkov)]

canceling Jamie request per comment #11

Comment 13

[Marco Zehe (:MarcoZ)]

Fixed by bug 1268677. Verified in 49.0a1 nightly 2016-05-04.

Comment 14

[Firefox Bug Husbandry Bot]

Removing leave-open keyword from resolved bugs, per :sylvestre.