Passing True for owner_events in the XGrabPointer call makes all
XI_RawMotion events appear in the queue twice, with the only difference
between them being the value of XGenericEventCookie::cookie. These have
always been filtered out by a check in the XI_RawMotion handler,
however with a mouse that polls at more than 1 kHz frequency, there
also exist legitimate events that appear indistinguishable from these
duplicated events. These must not be filtered out, otherwise the
pointer may move at an inconsistent speed, appearing like a bad pointer
acceleration implementation.
Change owner_events to False in the XGrabPointer and remove the
duplicate event detection code to fix this.
Signed-off-by: Torge Matthies <openglfreak@googlemail.com>
(cherry picked from commit f18b5656f6f859e4d4e096d290afd9fae884a5b8)
Setting the cursor clip area to a single pixel prevents the relative mouse motion remote desktop warping from working, so the mouse is never recentered.
(cherry picked from commit daffe02b117ccd484763eadc716e4d4453c89868)
It's legitimate to have multiple contexts.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
(cherry picked from commit 93861e1de00a76080fa6ed756a6bd9c764b3ec07)
`EM_ASM_` and `EM_ASM_INT_V` are calls that have been deprecated
for a long time.
Since the return value isn't used for the call to `EM_ASM_`, it
can be replaced with `EM_ASM`.
`EM_ASM_INT_V` is now (for the last few years) `EM_ASM_INT`.
(cherry picked from commit a8e89f2567b9069c919f9b21996e0a46cd4bb679)
MessageBoxes attached to a window in macOS should use modal APIs and not
use a poll/sleep pattern on the main thread. Sleeping the main thread
makes the NSWindow message loop sluggish and interferes with external
applications that need to send messages to that window, such as
VoiceOver.
libdecor plugins can change the min/max window size values internally to enforce a minimum window size, and errors and crashes can result if the window size is below the internal limit.
On versions of libdecor >= 0.1.1, the minimum width and height can be queried and the minimum required window size will be enforced. The application requested window size is still respected, however, the actual window may be slightly larger than the drawable area to accommodate the required libdecor minimum size.
On version 0.1.0 of libdecor, which lacks the function to retrieve the minimum size, the internal limits are overridden before committing a frame, so that the internal limits always match the window size as a workaround, even if the window is technically smaller than the plugin would normally allow.
(cherry picked from commit 423a82cd4b65cf72668551093bfdf58d49bce9ce)
Libdecor can crash if attempting to un-maximize a window that was not set as resizable.
(cherry picked from commit e35c3872dc6a8f7741baba8b786b202cef7503ac)
Perform a round trip when maximizing and restoring windows so the changes have already taken effect when the associated functions return.
(cherry picked from commit 50f2eb7d41fbba7c82630e55ead4ecf9c32f4c69)
A previous cleanup commit inverted a statement that checked the return value of
a KMSDRM_CreateSurfaces call during KMSDRM_CreateWindow, which causes the video
backend to always fail despite success.
This commit restores the intended behavior.
Fixes: 3c501b963dd8 ("Clang-Tidy fixes (#6725)").
(cherry picked from commit 0187209f461b84d22fc49a03369506e06448f47f)
This happens when using Win+Alt+Left/Right on a resizable fullscreen desktop window on Windows
(cherry picked from commit 650e16a8245f3fb5056fc7f6b398fb88244b3477)
It was previously thought that these function calls were unnecessary as the initial bug and reproduction case that necessitated their addition seemed to be fixed, but apparently there are still cases where this needs to be set explicitly. Set the xdg surface geometry at creation time and when the window size changes.
Partially reverts #6361. This is not needed in the libdecor path, as libdecor calls this for the content surface internally.
(cherry picked from commit 90a964f132c63b07cf85e5687041d214fdea60f6)
Num Lock and Caps Lock always need to be explicitly handled by the modifier handler function, or they won't be correctly set if active at application startup, or if the lock state is changed while the application lacks focus since a key press for these keys will never be received. In these cases, the internal SDL modifier state can end up the inverse of the actual modifier state.
(cherry picked from commit 653e484da17d773f40bff94121a23a3b38f47a0a)
I updated .clang-format and ran clang-format 14 over the src and test directories to standardize the code base.
In general I let clang-format have it's way, and added markup to prevent formatting of code that would break or be completely unreadable if formatted.
The script I ran for the src directory is added as build-scripts/clang-format-src.sh
This fixes:
#6592#6593#6594
(cherry picked from commit 5750bcb174300011b91d1de20edb288fcca70f8c)
The new protocol adds support for more native communication of
fractional scaling.
Everything in the wayland backend already existed only our fractional
scale was calculated implicitly through a combination of output size
guesswork for fullscreen windows.
This new protocol makes that explicit, providing a more robust solution
and a solution for non-fullscreen surfaces. The fallback code is still
left in place for now whilst compositors gain support.
- do not overwrite error message set by SDL_InitFormat (SDL_AllocFormat)
- set proper error message (Cocoa_Metal_CreateView)
- protect against allocation failure (UIKit_Metal_CreateView)
(cherry picked from commit cf0cb44df88a4293805fdc926880155d58a46bea)
* Add braces after if conditions
* More add braces after if conditions
* Add braces after while() conditions
* Fix compilation because of macro being modified
* Add braces to for loop
* Add braces after if/goto
* Move comments up
* Remove extra () in the 'return ...;' statements
* More remove extra () in the 'return ...;' statements
* More remove extra () in the 'return ...;' statements after merge
* Fix inconsistent patterns are xxx == NULL vs !xxx
* More "{}" for "if() break;" and "if() continue;"
* More "{}" after if() short statement
* More "{}" after "if () return;" statement
* More fix inconsistent patterns are xxx == NULL vs !xxx
* Revert some modificaion on SDL_RLEaccel.c
* SDL_RLEaccel: no short statement
* Cleanup 'if' where the bracket is in a new line
* Cleanup 'while' where the bracket is in a new line
* Cleanup 'for' where the bracket is in a new line
* Cleanup 'else' where the bracket is in a new line
(cherry picked from commit 6a2200823c66e53bd3cda4a25f0206b834392652 to reduce conflicts merging between SDL2 and SDL3)
Explicitly include the Wayland protocol headers when statically linking against the Wayland libraries or older system headers might be used instead of the local versions.
(cherry picked from commit 836eb224428aca3bdab2a6bf56d347262e475b15)
The typedef seems to be pulled in coincidentally with newer SDKs, but
older ones need to import the header explicitly.
(cherry picked from commit d2910904fb4062c313636c7595e971f1bf248075)
When changing the window state from non-floating to floating (e.g. leaving fullscreen), libdecor can send bogus content sizes that are +/- the height of the window title bar and start 'walking' the window height in one direction or the other with every transition.
The floating window size is known, so use the cached value instead of the size reported by libdecor when restoring the floating state.
Some applications (and embarrassingly, testime is one of them) call
SDL_StartTextInput() or SDL_SetTextInputRect() every frame. On KDE/KWin
with fcitx5, this causes there to be several preedit events every frame
(particularly given some of the workarounds in Wayland_StartTextInput),
which slows testime down to an unusable crawl.
Instead, make SDL_StartTextInput() a no-op if text input is already
enabled, and cache the input rect, only changing it when the new rect is
actually different.
With these changes, we only get preedit events (and hence
SDL_TEXTEDITING events) when the preedit string actually changes. This
matches the behaviour under XWayland, and works very smoothly.
Most of these are probably harmless, but the changes to SDL_immdevice.c and SDL_pixels.c appear to have fixed genuine bugs.
SDL_audiocvt.c: By separating the calculation of the divisor, I got rid of the suspicion that dividing a double by an integer led to loss of precision.
SDL_immdevice.c: Added a missing test, one that could have otherwise led to dereferencing a null pointer.
SDL_events.c, SDL_gamecontroller.c, SDL_joystick.c, SDL_malloc.c, SDL_video.c: Made it clear the return values weren't used.
SDL_hidapi_shield.c: The size is zero, so nothing bad would have happened, but the SDL_memset() was still being given an address outside of the array's range.
SDL_dinputjoystick.c: Initialize local data, just in case IDirectInputDevice8_GetProperty() isn't guaranteed to write to it.
SDL_render_sw.c: drawstate.viewport could be null (as seen on line 691).
SDL.c: SDL_MostSignificantBitIndex32() could return -1, though I don't know if you want to cope with that (what I did) or SDL_assert() that it can't happen.
SDL_hints.c: Replaced boolean tests on pointer values with comparisons to NULL.
SDL_pixels.c: Looks like the switch is genuinely missing a break!
SDL_rect_impl.h: The MacOS static checker pointed out issues with the X comparisons that were handled by assertions; I added assertions for the Y comparisons.
SDL_yuv.c, SDL_windowskeyboard.c, SDL_windowswindow.c: Checked error-result returns.
This reverts commit 04b50f6c6b.
It turns out OpenGL vsync has broken again in macOS 12, so
we're reintroducing our CVDisplayLink code to deal with it,
again.
Reference Issue #4918.
SDL is built around the concept of keyboards having a fixed layout with scancodes that correspond to physical keys no matter what linguistic layout is used. Virtual keyboards don't have this concept and can present an arbitrary layout of keys with arbitrary scancodes and names, which don't fit the SDL model. When one of these keyboards is encountered, it requires special handling: use the keysym of the pressed keys to derive their ANSI keyboard scancode equivalents for control keys and ASCII characters. All other characters are passed through as text events only.
It turns out that we can safely create a Metal view on an existing window, and that avoids issues with the window being recreated with the wrong orientation in iOS 16.
Fixes https://github.com/libsdl-org/SDL/issues/6289
Clamp the wl_seat max version to 5 if being built against a version of libwayland below 1.21.0, or containers that bundle newer versions of SDL with older versions of libwayland can break if the compositor advertises support for a protocol version above 5.
Most of this code is disabled out for now.
- For mouse cursors we have a wl_surface for both system and custom
cursors which needs recreating.
- The other patch is about nullification after deletions
If relative mouse mode is explicitly enabled, don't modify the capture flag on button events or the window might report having lost mouse focus if a button is pressed while moving the cursor.
This is a functional state for some window managers (tested using stock Ubuntu 22.04.1), and removing that state, e.g. using SDL_RestoreWindow(), results in a window centered and floating, and not visually covering the rest of the desktop.
This uses a newer browser API to get physical scancodes, but still
uses the (deprecated) event field that we were already using for
scancodes, but for keycodes instead now, which appears to be more
accurate.
Since keyboard layout isn't (generally) available to web apps, this
adds an internal interface to send key events with both scancode
and keycode to SDL's internals, instead of sending just scancodes and
expecting SDL to use its own keymap to generate keycodes.
Future work in this area would be to use the keyboard layout APIs
on browsers that support them, which would allow us to use SDL's
usual keymap code and not rely on a deprecated browser API, but
until we get there, this patch gives significantly more correct
results than we would have before.
Fixes#2098.
Exit the fullscreen sequence sooner if it is requested that a popup window be fullscreen.
The surface commit formerly in this path is irrelevant and can be removed as previous changes made it so that SetFullscreen() is no longer called from anywhere except Wayland_SetWindowFullscreen().
Clear the window to black on the initial window draw, to avoid a really
obnoxious white flash. This doesn't always eliminate it, but it
definitely minimizes it.
This makes the colorspace match across different graphics APIs. By
default, OpenGL was getting a much more saturated colorspace (maybe
Display P3?) and it was looking very different from the rendering done
by Metal or MoltenVK.
Replace instances of fprintf(stderr, ...) with SDL_SetError(), replace C++ comments with C style, use a uniform format for multi-line comments, and remove unused headers as poll and select aren't used in this file (the SDL function which calls them is used instead).
libdecor_dispatch() needs to be called, as libdecor plugins might do some required internal processing within, however care must be taken to avoid double-blocking in the case of a timeout, which can occur if libdecor_dispatch() and the SDL event processing both work on the main Wayland queue. Additionally, assumptions that libdecor will always dispatch the main queue or not process zero-length queues (can occur if a wait is interrupted by the application queueing an event) should not be made, as this behavior is outside the control of SDL and can change.
SDL handles polling for Wayland events and then calls libdecor to do its internal processing and dispatch. If libdecor operates on the main queue, it will dispatch the queued events and the additional wl_display_dispatch_pending() call will be a NOP. If a libdecor plugin uses its own, separate queue, then the wl_display_dispatch_pending() call will ensure that the main queue is always dispatched.
In all cases they were using SDL_SCANCODE_TABLE_XFREE86_2 with some keycodes remapped or fewer than expected keycodes. This adds a sanity check that catches all of them and gives them the right scancode table.
If we can't find the X11 keysym, it's likely that either the keysym is NoSymbol, in which case we won't hit it anyway, or it's been mapped to a character, in which case the existing mapping is correct for the scancode and the character will be reflected in the keycode mapping.
* Consolidated scancode mapping tables into a single location for all backends
* Verified that the xfree86_scancode_table2 is largely identical to the Linux scancode table
* Updated the Linux scancode table with the latest kernel keycodes (still unmapped)
* Route X11 keysym -> scancode mapping through the linux scancode table (which a few hand-written exceptions), which will allow mappings to automatically get picked up as they are added in the Linux scancode table
* Disabled verbose reporting of missing keysym mappings, we have enough data for now
