Fix accessibility in Windows x64 build.

Fix accessibility in Windows x64 build.

The only fix needed was to cast the LPARAM to a DWORD when
handling WM_GETOBJECT in two places.

Also fixes and re-enables a bunch of Windows accessibility tests,
and deletes a couple of old disabled tests that aren't worth fixing.

BUG=179717, 383240
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=277457

[dmazzoni, 2014-06-13 20:17:05]

jschuh: please review the DWORD cast in the WM_GETOBJECT handlers to fix 64-bit
support.

jam: please review content/ including the fixes to the browser test.

sky: OWNERS review for ui/views/

[jschuh, 2014-06-13 23:49:27]

https://codereview.chromium.org/329863003/diff/1/ui/views/win/hwnd_message_ha...
File ui/views/win/hwnd_message_handler.cc (right):

https://codereview.chromium.org/329863003/diff/1/ui/views/win/hwnd_message_ha...
ui/views/win/hwnd_message_handler.cc:1419: DWORD obj_id =
static_cast<DWORD>(l_param);
Sorry to be pedantic, but this is an undefined conversion, so it would blow up
on e.g. lexan. You should actually just leave l_param as it is and expand
OBJID_CLIENT as follows:

static const LPARAM kObjIdClient = static_cast<ULONG>(OBJID_CLIENT);

https://codereview.chromium.org/329863003/diff/1/ui/views/win/hwnd_message_ha...
ui/views/win/hwnd_message_handler.cc:1422: if (OBJID_CLIENT == obj_id) {
This would become:

if (kObjIdClient == obj_id) {

[dmazzoni, 2014-06-15 07:56:03]

https://codereview.chromium.org/329863003/diff/1/ui/views/win/hwnd_message_ha...
File ui/views/win/hwnd_message_handler.cc (right):

https://codereview.chromium.org/329863003/diff/1/ui/views/win/hwnd_message_ha...
ui/views/win/hwnd_message_handler.cc:1419: DWORD obj_id =
static_cast<DWORD>(l_param);
On 2014/06/13 23:49:27, Justin Schuh wrote:
> Sorry to be pedantic, but this is an undefined conversion, so it would blow up
> on e.g. lexan. You should actually just leave l_param as it is and expand
> OBJID_CLIENT as follows:
> 
> static const LPARAM kObjIdClient = static_cast<ULONG>(OBJID_CLIENT);

OK.

https://codereview.chromium.org/329863003/diff/1/ui/views/win/hwnd_message_ha...
ui/views/win/hwnd_message_handler.cc:1422: if (OBJID_CLIENT == obj_id) {
On 2014/06/13 23:49:27, Justin Schuh wrote:
> This would become:
> 
> if (kObjIdClient == obj_id) {

Done.

[jam, 2014-06-16 03:27:27]

lgtm

[jschuh, 2014-06-16 14:39:19]

lgtm

[sky, 2014-06-16 14:39:52]

LGTM

[dmazzoni, 2014-06-16 15:48:09]

The CQ bit was checked by dmazzoni@chromium.org

[commit-bot: I haz the power, 2014-06-16 15:49:50]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/dmazzoni@chromium.org/329863003/20001

[commit-bot: I haz the power, 2014-06-16 16:09:09]

Change committed as 277457

[dmazzoni, 2014-06-17 23:46:16]

OK, it turns out this patch doens't fix it, it fixes some clients that were
previously broken but just breaks others. If I revert to my original patch, it
fixes all clients.

On Windows x64, LPARAM is a 64-bit signed integer, and DWORD is a 32-bit
unsigned integer.

The problem appears to be that when a client passes OBJID_CLIENT (which is -4),
it sometimes gets extended to 0x00000000fffffffc or 4294967292, a positive
integer - rather than being properly sign-extended to 0xfffffffffffffffc.
However, other clients will have the value sign-extended properly.

So basically, we need to handle both 0x00000000fffffffc and 0xfffffffffffffffc
and treat them both as OBJID_CLIENT.

Justin, what syntax would you prefer?