Based on a patch by Mark Jansen.
Signed-off-by: Alistair Leslie-Hughes <leslie_alistair@hotmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
There's no point to have it lingering around, since it's always recreated
before being shown again.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
When the listbox is visible, ESC should hide it. Only when it's not visible
should it be forwarded to the edit control. This matches Windows behavior.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
When selecting an item from the AutoComplete's listbox, the Return key
should act the same as a left click on it (place the text, select it,
and hide the listbox).
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Clean the CreateWindowEx code of CW_USEDEFAULT since it's not valid for
a child window and ends up being 0 anyway. Even more, it's useless as the
actual size is given when the window is shown.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Windows' AutoComplete seems to bypass hijacking in some cases. However,
WM_GETTEXT seems to be able to get hijacked.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Send some of the messages directly to the edit control's window procedure
to match Windows behavior and to be able to process WM_SETTEXT.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Most control characters sent via some CTRL+key combination should not
autocomplete at all. ^C is one example, where just copying some text should
not show the auto-suggestion box (if not visible). ^V is another example,
where it is already handled in WM_PASTE, so it has to be a no-op here,
else auto-append from WM_PASTE would complete the text and then the ^V
autocompletion would remove every other suggestion in the listbox.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
The user can right-click on the editbox control and choose one of these
operations, so they need to be handled separately regardless.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
When going up past the topmost item in the listbox, go through txtbackup
first before wrapping around, just like when going down. This matches
Windows behavior.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
There's no need to send a WM_KEYUP anymore since it now matches Windows
behavior.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
We must not auto-append on control characters, most notably Backspace,
but also ^X (cut) and so on.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
AutoComplete currently shows up when the user releases a key, which is
wrong. Windows does it when the user presses a key, so use both WM_KEYDOWN
and WM_CHAR and redesign it so that it matches Windows behavior.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
There's no need to have filled, since cpt can already provide the same
information.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
We can retrieve the length of the string from the SendMessage calls already.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
The quickComplete format can have more than one % argument, or stuff like
%*.* or %1234s, which can be exploited since the format string can be read
from the registry, so handle it manually instead of using sprintf.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Handle heap_alloc failure, reg strings without a \ character at all,
try harder to find the reg path (if only value fails the lookup), and
read the registry value with any size.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Gabriel Ivăncescu