mozilla :: #accessibility

6 Jan 2017
21:43tbsaundeaklotz: want to review my latest version of the diagnostic patch in bug 1184217 or shall I just go with davidb's old r+
21:43firebothttps://bugzil.la/1184217 REOPENED, tbsaunde+mozbugs@tbsaunde.org crash in mozilla::a11y::DocAccessibleParent::Unbind()
22:57yzentbsaunde hi
22:57tbsaundeyzen: pong
22:59yzentbsaunde im looking at the igeckocustom stuff and was planning on landing the DOMNodeID patch for tests, but i noticed that querying that interface in a proxy never succeeds
22:59yzentbsaunde (when running one of our browser tests)
23:01tbsaundeyzen: hrm, I suspect we aren't loading the typelib for that code since its different from the ia2/ISimpleDom/msaa ones
23:01tbsaundeaklotz: ideas on^?
23:02yzentbsaunde yes that's a different problem
23:03tbsaundeyzen: how do you know its different?
23:04yzentbsaunde what you'er describing results in an assert in https://dxr.mozilla.org/mozilla-central/source/ipc/mscom/Interceptor.cpp#119
23:04yzentbsaunde i worked around this locally by adding bits for IGeckoCustom.tlb similar to how it's done for other libs
23:05yzenbut not sure if it was done correctly
23:06tbsaundeyzen: ok then, I don't really have ideas, so just debug it I guess
23:07yzentbsaunde ok
23:07tbsaundedoes a QueryInterface get called at all in the child process?
23:09tbsaundeyzen: ^?
23:10yzentbsaunde yes, in my case when i want to get an id of the accessiblewrap's queryinterface is called and the GeckoCustom is created and HRESULT looks good but when proxy's QI method resumes the result is a fail
23:10yzentbsaunde id for a doc accessible's proxy i mean
23:14tbsaundehuh
23:15tbsaundeyzen: any kind of interesting error code or is it just E_FAIL?
23:18yzentbsaunde ill take a look
23:18yzenanyways ill comment on the bug
23:18yzengotta run now
23:21tbsaundeok see ya
23:57aklotztbsaunde: just go ahead with davidb's review
23:59aklotztbsaunde, yzen: re: the QueryInterface problem, it's a separate but related problem. In content the typelib needs to be registered or else that assertion will fire, yes. But it also needs to be registered in the chrome process where the interceptor code isn't used
7 Jan 2017
No messages
   
Last message: 46 days and 8 hours ago