Re: [3/3] imm32,user32: Create and destroy the default IME window implicitly. (try 9; resend)

[Sebastian Lackner]

On 23.08.2016 15:13, Akihiro Sagawa wrote:
> Try 9:
>  - Withdraw try 8 changes as we have ImmDisableIME now.
>  - Rebased and resolve conflicts against ImmDisableIME patch.
>
> Signed-off-by: Akihiro Sagawa <sagawa.aki(a)gmail.com>
> ---
>  dlls/imm32/imm.c           | 99 +++++++++++++++++++++++++++++++++++-----------
>  dlls/imm32/imm32.spec      |  1 +
>  dlls/imm32/tests/imm32.c   | 12 +++---
>  dlls/user32/misc.c         |  4 +-
>  dlls/user32/user_private.h |  1 +
>  dlls/user32/win.c          | 26 ++++++++++++
>  dlls/user32/win.h          |  1 +
>  7 files changed, 113 insertions(+), 31 deletions(-)

This patch broke various user32 tests in virtual desktop mode, for example:

$ (cd dlls/user32/tests/; make broadcast.ok)
../../../tools/runtest -q -P wine -T ../../.. -M user32.dll -p user32_test.exe.so broadcast && touch broadcast.ok
broadcast.c:174: Test failed: Asynchronous message sent instead
broadcast.c:152: Test failed: Returned: 0
broadcast.c:153: Test failed: Asynchronous message sent instead
broadcast.c:174: Test failed: Asynchronous message sent instead
broadcast.c:238: Test failed: Returned: 0
broadcast.c:239: Test failed: Asynchronous message sent instead
broadcast.c:257: Test failed: Asynchronous message sent instead
broadcast.c:238: Test failed: Returned: 0
broadcast.c:239: Test failed: Asynchronous message sent instead
broadcast.c:257: Test failed: Asynchronous message sent instead
broadcast.c:298: Test failed: Returned: 0 error 1460
broadcast.c:299: Test failed: Asynchronous message sent instead
broadcast.c:309: Test failed: Returned: 0 error 1460
broadcast.c:310: Test failed: Asynchronous message sent instead
broadcast.c:320: Test failed: Returned: 0 error 1460
broadcast.c:321: Test failed: Asynchronous message sent instead
broadcast.c:332: Test failed: Asynchronous message sent instead
make: *** [Makefile:197: broadcast.ok] Fehler 17

Could you please take a look?

[3/3] imm32,user32: Create and destroy the default IME window implicitly. (try 9; resend)

[Akihiro Sagawa]

On Tue, 30 Aug 2016 21:48:49 +0200, Sebastian Lackner wrote:
> This patch broke various user32 tests in virtual desktop mode, for example:
>
> $ (cd dlls/user32/tests/; make broadcast.ok)
> ../../../tools/runtest -q -P wine -T ../../.. -M user32.dll -p user32_test.exe.so broadcast && touch broadcast.ok
> (snip)
> make: *** [Makefile:197: broadcast.ok] Fehler 17
>
> Could you please take a look?

Sure. I'll take a look.
Thanks for report.

Akihiro Sagawa

[3/3] imm32,user32: Create and destroy the default IME window implicitly. (try 9; resend)

[Sebastian Lackner]

On 31.08.2016 12:49, Akihiro Sagawa wrote:
> On Tue, 30 Aug 2016 21:48:49 +0200, Sebastian Lackner wrote:
> > This patch broke various user32 tests in virtual desktop mode, for example:
> >
> > $ (cd dlls/user32/tests/; make broadcast.ok)
> > ../../../tools/runtest -q -P wine -T ../../.. -M user32.dll -p user32_test.exe.so broadcast && touch broadcast.ok
> > (snip)
> > make: *** [Makefile:197: broadcast.ok] Fehler 17
> >
> > Could you please take a look?
>
> Sure. I'll take a look.
> Thanks for report.
>
> Akihiro Sagawa

Just to let you know, in the meantime I have tracked down what is causing the test failure
(but I am not yet sure how to fix it). Starting with your change, we are also creating IME
windows for "OleMainThreadWndClass" windows. In this particular case, the watch_desktop_folders
thread created by explorer.exe is causing the trouble. The thread does not have any message
loop which is blocking broadcast messages.

Should we filter this case in imm32.needs_ime_window?

[3/3] imm32,user32: Create and destroy the default IME window implicitly. (try 9; resend)

[Sebastian Lackner]

On 02.09.2016 05:35, Sebastian Lackner wrote:
> On 31.08.2016 12:49, Akihiro Sagawa wrote:
> > On Tue, 30 Aug 2016 21:48:49 +0200, Sebastian Lackner wrote:
> > > This patch broke various user32 tests in virtual desktop mode, for example:
> > >
> > > $ (cd dlls/user32/tests/; make broadcast.ok)
> > > ../../../tools/runtest -q -P wine -T ../../.. -M user32.dll -p user32_test.exe.so broadcast && touch broadcast.ok
> > > (snip)
> > > make: *** [Makefile:197: broadcast.ok] Fehler 17
> > >
> > > Could you please take a look?
> >
> > Sure. I'll take a look.
> > Thanks for report.
> >
> > Akihiro Sagawa
>
> Just to let you know, in the meantime I have tracked down what is causing the test failure
> (but I am not yet sure how to fix it). Starting with your change, we are also creating IME
> windows for "OleMainThreadWndClass" windows. In this particular case, the watch_desktop_folders
> thread created by explorer.exe is causing the trouble. The thread does not have any message
> loop which is blocking broadcast messages.
>
> Should we filter this case in imm32.needs_ime_window?

Actually we should probably skip creation for all windows with HWND_MESSAGE parent.
Nevertheless, I will leave a fix to someone more familiar. ;)

Regards,
Sebastian

[3/3] imm32,user32: Create and destroy the default IME window implicitly. (try 9; resend)

[Akihiro Sagawa]

On Fri, 2 Sep 2016 05:41:48 +0200, Sebastian Lackner wrote:
> > Just to let you know, in the meantime I have tracked down what is causing the test failure
> > (but I am not yet sure how to fix it). Starting with your change, we are also creating IME
> > windows for "OleMainThreadWndClass" windows. In this particular case, the watch_desktop_folders
> > thread created by explorer.exe is causing the trouble. The thread does not have any message
> > loop which is blocking broadcast messages.
> >
> > Should we filter this case in imm32.needs_ime_window?
>
> Actually we should probably skip creation for all windows with HWND_MESSAGE parent.
> Nevertheless, I will leave a fix to someone more familiar. ;)

Great! I verified that the default IME window isn't created for
message-only window (= a window created with HWND_MESSAGE parent).
I'm preparing tests now.

Thanks a lot,
Akihiro Sagawa