GNOME Bugzilla – Bug 601552
Need a way to denote out-of-process accessibles
Last modified: 2010-12-31 03:26:16 UTC
Some applications, such as Moonlight and perhaps Gecko in the future, will have a UI that spans multiple processes, so we need a way to denote that an accessible in another process should be represented as a child of a given accessible. See http://lists.linux-foundation.org/pipermail/accessibility-atspi/2009-September/000586.html for the original proposal discussing this.
Created attachment 147478 [details] [review] Proposed patch.
Which module should the patch be applied to?
Created attachment 148368 [details] [review] Proposed patch--the correct one, this time Sorry--I accidentally uploaded some other diff I had lying around on my system, rather than the patch for this bug.
I noticed there is printf in the code for debug. Do we still need that? I think there should be patches for at-spi2 also, can I have the link?
Thanks for the catch; I have removed the printf. The at-spi2-atk change is in commit 9e77ce18c391f574e4cda97b553eef172c2e276f
Thanks for pointing me to the patch. I have one question: how does a plug (say GtkPlug) embeds its children as the children of a socket (say GtkSocket)?
The plug would create an AtkPlug object and call atk_plug_get_id (which is implemented by at-spi). It would need to somehow pass the string returned by atk_plug_get_id() to the socket, which should create an AtkSocket and call atk_socket_embed with the plug id.
Then I am OK with the patch and will release it with 1.29.3 today.
Created attachment 148720 [details] [review] patch Remove the printf.
With /desktop/gnome/interface/accessibility set to true on my OpenSolaris b129 box with the verm-130-x86 package from http://pkg.ireland/verm-x86, I get the following errors in my .xsession-errors file when I attempt to login and my login fails: ld.so.1: gnome-settings-daemon: fatal: relocation error: file /usr/lib/gtk-2.0/m odules/libatk-bridge.so: symbol atk_socket_get_type: referenced symbol not found ld.so.1: metacity: fatal: relocation error: file /usr/lib/gtk-2.0/modules/libatk -bridge.so: symbol atk_socket_get_type: referenced symbol not found ld.so.1: gnome-session: fatal: relocation error: file /usr/lib/gtk-2.0/modules/l ibatk-bridge.so: symbol atk_socket_get_type: referenced symbol not found It seems to be related to the patch for this RFE. :-(
What version of atk do you have installed? I'm wondering if you've built at-spi2 against atk 2.29.3 but have an older version installed somewhere that is being loaded instead. It isn't clear to me how else you could get that error; atk_socket_get_type should be generated by the G_DEFINE_TYPE call in atksocket.c (I tried manually adding an atk_socket_get_type, but it would conflict with the one generated by the macro).
(In reply to comment #11) > What version of atk do you have installed? I'm wondering if you've built > at-spi2 against atk 2.29.3 but have an older version installed somewhere that > is being loaded instead. It isn't clear to me how else you could get that > error; atk_socket_get_type should be generated by the G_DEFINE_TYPE call in > atksocket.c (I tried manually adding an atk_socket_get_type, but it would > conflict with the one generated by the macro). Something is definitely very strange -- this might be an integration/build problem specific to the OpenSolaris product. I'm hoping Li might be able to lend some insight -- all I did was take a vanilla OpenSolaris b129 machine and install the verm-130-x86 package from http://pkg.ireland/verm-x86 (a Sun internal site). The verm-130-x86 package *seems* to be GNOME 2.28, so I'm very very confused at how I ended up with the atk version that has atk_socket_get_type in it. I may have done something wrong, but I'm pretty sure I'm just using simple bits from b129 and verm-130 and have not build/installed anything from source on this machine yet.
Our QE have run the accessibility test cases on both vermillion stable build 130 and vermillion development build 130 and didn't see your errors. The only place can cause the error is atk-bridge in at-spi2 which only exists in vermillion development build (which is gnome 2.29). And I checked the SUNWlibatk package released with vermillion development build 130, atk_socket_get_type is included in the library. So I am guessing you had a incomplete upgrade to vermillion development build 130, probably SUNWlibatk is not upgraded.
(In reply to comment #13) > Our QE have run the accessibility test cases on both vermillion stable > build 130 and vermillion development build 130 and didn't see your > errors. > > The only place can cause the error is atk-bridge in at-spi2 which only > exists in vermillion development build (which is gnome 2.29). And I > checked the SUNWlibatk package released with vermillion development > build 130, atk_socket_get_type is included in the library. So I am > guessing you had a incomplete upgrade to vermillion development build > 130, probably SUNWlibatk is not upgraded. I'm going to guess the IPS I was working with was somewhat useless.
(In reply to comment #14) > I'm going to guess the IPS I was working with was somewhat useless. Taking into account this comment, and the fact that all the patches were already applied, I guess that this bug could be closed as FIXED. Right?