Avoid using existing parameter names for local variables.
Avoid using variable names from upper level for nested loops.
Avoid local variables hiding existing global variables.
Apple go on making it harder and harder to support multiple
versions of their pathetic operating system in one code base.
kIOMainPortDefault is the "non-deprecated" replacement for
kIOMasterPortDefault. However, it is now marked as being available only
on macOS 12 and later so it refuses to build in a codebase designed to
run on earlier versions of macOS. Le bubbling sigh.
According to https://developer.apple.com/documentation/iokit/kiomasterportdefault?language=objc
passing NULL has the same effect and avoids the stupid warning.
Hopefully this will work till Apple decides to spend their trillion
dollars doing more stupid shit.
This fixes an issue where restoring from traditional fullscreen to a
maximized state was resulting in an incorrectly sized frame buffer. Now
all framebuffer and window size calculation happens in the same way in
all the various glfw codepaths that do it.
Let the input system do it for us. That wy we can hopefully use calls to
unmarkText as a signal to commit the pre-edit text.
Hopefully not clearing the marked text on key down doesnt break anything
else.
Without this, kitty fails to compile:
```
glfw/cocoa_init.m:462:14: error: case value not in enumerated type
'NSEventModifierFlags' (aka 'enum NSEventModifierFlags')
[-Werror,-Wswitch]
case (NSEventModifierFlagShift | NSEventModifierFlagOption):
^
glfw/cocoa_init.m:465:14: error: case value not in enumerated type
'NSEventModifierFlags' (aka 'enum NSEventModifierFlags')
[-Werror,-Wswitch]
case (NSEventModifierFlagShift | NSEventModifierFlagCommand):
^
```
I thought about changing the type of the `modifierFlags` function parameter from `NSEventModifierFlags` to `NSUInteger` since `NSEventModifierFlags` is defined as an enum according to https://developer.apple.com/documentation/appkit/nseventmodifierflags?language=objc and is technically not the proper type for a bit field. But since Apple themselves define `modifierFlags` as `NSEventModifierFlags` according to https://developer.apple.com/documentation/appkit/nsevent/1534405-modifierflags?language=objc, I think the fix in this commit is better.
This warning was introduced with commit f9944e6140.
The ForwardKeyEvent ibus signal is used to maintain ordering between CommitText events and non-IME key events.
The ordering is important for Korean IMEs, where it automatically commits text whenever a character is fully composed. This is different from Chinese/Japanese IMEs, where user manually presses a key to commit each word.
Without this patch, kitty can't handle non-IME key events when used with ibus-hangul, which utilizes ForwardKeyEvent signal.
Without ForwardKeyEvent:
```
key | sequence of events from application's view
r | UpdatePreeditText 'ㄱ'
k | UpdatePreeditText '가'
1 | (receives the reply to ProcessKeyEvent call: "native_key: 0x31 release: 0 handled: 0")
| -> UpdatePreeditText ''
| -> CommitText '가'
```
Application receives "1가", which is incorrect.
With ForwardKeyEvent:
```
key | sequence of events from application's view
r | UpdatePreeditText 'ㄱ'
k | UpdatePreeditText '가'
1 | (receives the reply to ProcessKeyEvent call: "native_key: 0x31 release: 0 handled: 1", and kitty discards the event)
| -> UpdatePreeditText ''
| -> CommitText '가'
| -> ForwardKeyEvent keysym=0x31("1")
```
Application receives "가1", which is correct.
Relevant ibus-hangul github issue to implement ForwardKeyEvent: https://github.com/libhangul/ibus-hangul/issues/42
Relevant changes in Qt to handle ForwardKeyEvent: https://codereview.qt-project.org/c/qt/qtbase/+/212179, https://codereview.qt-project.org/c/qt/qtbase/+/255587
fcitx5 works fine with kitty, except that when you restart fcitx5, kitty stop working with fcitx5, because the dbus_connection_get_is_connected(ibus->conn) in #check_connection is always true when you use fcitx5.
this patch add a dbus name owner changed detection to detect the online status of ibus, in case the ibus backend uses the session bus, like fcitx5.
Try to ensure we have a functioning displaylink always.
GLFW skips over sleeping monitors during a poll and also
had a bug where the display link was not re-created for a monitor
that already had a glfw monitor entry.
Also add a bunch more debug reporting
Sway will send the next configure event (for example when focus changes)
with titlebar height added, so to avoid a sudden resize at that time,
bump the height by the titlebar.
This is caused, as far as I can tell, by CVDisplayLink getting stuck.
Apple apparently are incapable of writing a simple timer robustly.
So if it remains stuck after a second delete and recreate it to force it
to restart.
Fixes#2016
Apparently mutter needs window geometry set after the surface commit not
before, otherwise it doesnt "stick". When decorations are not hidden
there were other commits caused by the decorations, masking the issue.
Fixes#3507
The constants `NSControlStateValueOn`, `NSControlStateValueOff` and `NSControlStateValueMixed` are only available for macOS 10.13 or above according to https://developer.apple.com/documentation/appkit/nscontrolstatevalueon?language=objc.
Without this commit, compilation fails with this message:
```
glfw/cocoa_window.m:1537:28: error: use of undeclared identifier 'NSControlStateValueOn'
item.state = NSControlStateValueOn;
^
glfw/cocoa_window.m:1539:28: error: use of undeclared identifier 'NSControlStateValueMixed'
item.state = NSControlStateValueMixed;
^
glfw/cocoa_window.m:1542:47: error: use of undeclared identifier 'NSControlStateValueOn'
item.state = controller.isDesired ? NSControlStateValueOn : NSControlStateValueOff;
^
glfw/cocoa_window.m:1542:71: error: use of undeclared identifier 'NSControlStateValueOff'
item.state = controller.isDesired ? NSControlStateValueOn : NSControlStateValueOff;
^
```
To fix this, simply redefine the constants to use the old and now deprecated constants on older macOS versions.
The code that causes this was introduced in 98519bf326.
The current version of GNOME has a regression that causes the compositor
to send incorrect geometry with the next configure event if the CSD
buffers are created after a call to set_window_geometry.
So call set_window_geometry last. And shake head sadly.
We make use of the fact that wayland tells us when the window is being
resized to track actual user requested sizes as opposed to compositor
requested ones.
The co-ordinates are in the surface co-ordinate system. When the
compositor specifies sizes it apparently adds one subsurface size
in both x and y directions, not both. Possibly because the origin is
placed after the top and left subsurfaces. So modify it by adding
and extra subsurface size to that.
The numbers dont make logical sense, but they do what is expected on
GNOME and since only GNOME is stupid enough to insist on CSD, that's
all we care about.
Apparently on X11 the maximum icon size is 128x128. 256x256 is too
large for the X11 protocol because the X server unserializes the icons
using "unsigned long" which is 64 bits on Linux. So we have to use
64bits per pixel instead of 32, with 32bits padded to 0.
While there I also got rid of the kitty.rgba file replacing it with a
128x128 PNG file.
Fixes#3260
