mozilla :: #accessibility

20 Mar 2017
00:02Jamietbsaunde: Right. I'm just pointing out that it doesn't solve the problem system wide, so I'm not sure there's much point implementing the workaround for us; if it's broken, some of the user's apps aren't going to work
00:18tbsaundeJamie: yeah, you not trying to handle it seems fair
19:17yzentbsaunde: hi
19:20yzentbsaunde: any ideas we might have the following sequence when opening a url in content tab: receiving pdoc accessible constructor for top level doc, then deallocating it, then receiving another pdoc accessible constructor?
19:20yzens/we/why we
19:27yzenmaybe aklotz ^
19:28aklotzyzen: No idea, but sounds... interesting...
19:29yzenaklotz: ok ill get more data and file a bug if necessary
19:30aklotzyzen: Please cc :handyman if you file that bug. This might be relevant to his interests as well.
19:30yzenok
19:32aklotzhandyman: So yzen just asked this a couple minutes ago:
19:32aklotzany ideas we might have the following sequence when opening a url in content tab: receiving pdoc accessible constructor for top level doc, then deallocating it, then receiving another pdoc accessible constructor?
19:32aklotzhandyman: He's going to look into it a bit more, but I thought you might want to know about that
19:32aklotzhandyman: I've asked him to cc you if he files a bug
19:42handymanaklotz: thanks. we pass the parent by IPDL managed object identity -- (ie parameter DocAccessibleParent* aParentDoc). I wonder if IPDL will return the deleted object pointer in that case...
19:42tbsaundeyzen: I assume delocation is coming from ActorDestroy or RecvShutdown?
19:42tbsaundeyzen: that seems kind of weird I guess, but not particularly wrong
19:43aklotztbsaunde, yzen, handyman: I just wonder if there's any kind of fallout from that rapid construct/destroy sequence that could be related to our lifetime issues
19:44tbsaundeaklotz: depends how the deallocation happened, I'd like to see the stack for that
19:45handymanI take it the idea is that the parent deallocates without the child's knowledge... in which case no more messages from the child should be handled, no?
19:45handymanBut the IPDL shutdown stuff is complex enough that there might be edge cases
19:46tbsaundehandyman: I would assume ipdl should never try and send a message to a deallocated actor, and I'm not really sure how that could happen
19:47yzenso I was looking into it because JAWS developer reported this: "I am still getting the node for the document that goes away while I am trying to access it.
19:47yzenThe element has a role of document but no children. I am seeing a whole bunch of events for element under the document node but the document node has no descendants."
19:47yzenwas thinking it might be related
19:48handymanyzen: ...sounds that way to me, too
19:48handymanbut I dont see how thats happening
19:48tbsaundeyzen: well, if they are getting events for objects that don't exist clearly something is wrong there
19:49yzenhandyman, tbsaunde ill get the stacks for the things i see if that would help
19:50tbsaundeyzen: only the deallocation one is interesting the others will just be some ipc stuff
19:50yzenk
20:38firebotNew Core - Disability Access APIs bug 1348965 filed by gwimberly@softvision.com.
20:38firebothttps://bugzil.la/1348965 NEW, nobody@mozilla.org NVDA Screenreader does not read drop down items when hovered over or selected with Surface Pro pen
21 Mar 2017
No messages
   
Last message: 63 days and 18 hours ago