Adam M.
Okay, here is the problem, I think.
During the first blit, RLEAlphaSurface is called to do RLE conversion of the RGBA source into a format allowing it "to be quickly alpha-blittable onto dest". Since the destination is the screen, it has no alpha channel. RLEAlphaSurface calls copy_opaque(dst, src + runstart, len, sf, df) (where copy_opaque is copy_32), which has this code:
SDL_RLEaccel.c:984:
RGBA_FROM_8888(*src, sfmt, r, g, b, a);
PIXEL_FROM_RGBA(*d, dfmt, r, g, b, a);
On the first line, it reads the source pixel 0xFFFFFFFF. The second line drops the alpha value (because dfmt for the screen has no alpha channel) and writes 0x00FFFFFF. Later, when the RLE conversion is being undone, uncopy_32 is called, which has the following code:
SDL_RLEaccel.c:1001:
Uint32 pixel = *s++;
RGB_FROM_PIXEL(pixel, sfmt, r, g, b);
a = pixel >> 24;
PIXEL_FROM_RGBA(*dst, dfmt, r, g, b, a);
However, the the alpha channel has already been dropped by copy_opaque (= copy_32), so pixel = 0x00FFFFFF and 'a' becomes 0. Thus, all opaque pixels lose their alpha channel when being unRLE'd. (I don't know what happens to pixels with alpha from 1-254, but they should be checked too.)
So, that seems to be the problem, but I'm not sure what the solution should be. Since opaque pixels have alpha == 255, I'm thinking to create another uncopy function for opaque pixels that simply uses 255 for alpha.
However, there may be other problems here. For translucent pixels, uncopy_32 assumes the alpha channel is stored in the upper 8 bits, but copy_32 doesn't store it there. Instead, it stores it in whatever location is appropriate for the destination surface. Isn't one of their behaviors incorrect, given the other? I'm not sure which to change, however.
For translucent pixels, it seems that the blit function uses do_blend, which is the BLIT_TRANSL_888 macro, which also assumes alpha is in top 8 bits. It has the comment "we have made sure the alpha is stored in the top 8 bits...", but it seems that's not true (copy_32 doesn't make sure the alpha goes there).
Perhaps the correct fix is to make copy_32 put the alpha there, but then that seems to require that RLE conversion be limited to destination surfaces that don't use the upper 8 bits. However, looking further, it seems that has already been done: if (masksum != 0x00ffffff) return -1; /* requires unused high byte */
Martin Gerhardy
Not sure - but I think there might be a logic flaw in SDL_SetWindowGrab.
The problem here is that this modifies the window flags and e.g. sets
SDL_WINDOW_INPUT_GRABBED - but the _this->grabbed_window pointer is not
yet set.
Then in SDL_UpdateWindowGrab the _this->grabbed_window pointer is only
set if the function pointer _this->SetWindowGrab is not NULL. But if
this is NULL, the _this->grabbed_window pointer is never set, but every
future call to any of the grab functions include an assert for:
SDL_assert(!_this->grabbed_window || ((_this->grabbed_window->flags &
SDL_WINDOW_INPUT_GRABBED) != 0));
That means the first call works, the second always fails and triggers
the assert.
This happens when the pointer is grabbed, but it's not clear if this is a
bug in x.org or my misunderstanding of the XGrabPointer() documentation.
At any rate, we'll want to revisit this later for a better solution.
Fixes Bugzilla #2963.
This was the only thing that made SDL_config.h generate differently between
32 and 64-bit versions of Linux, so instead we force a function cast in our
X11 code to match our dynamic loader version, which removes the compile error
on some machines that prompted this test in the first place.
Xlib never wrote to this data, so if you're on an older Xlib where this param
wasn't const, your data should still be intact when we force the caller to
think it was actually const after all.
Fixes Bugzilla #1893.
Fixes Bugzilla #2895.
His notes:
The following trivial changes make SDL2 tree (mostly) compatible with Visual
Studio 2005:
* SDL_stdlib.c: Similar to VS2010 and newer, VS2005 also generates memcpy(),
(it also generates memset(), see below), so propagate the #if condition to
cover VS2005.
* SDL_pixels.c (SDL_CalculateGammaRamp): VS2005 generates a memset() call for
gamma==0 case, so replace the if loop with SDL_memset().
* SDL_windowsvideo.h: Include msctf.h only with VS2008 and newer, otherwise
include SDL_msctf.h
* SDL_windowskeyboard.c: Adjust the #ifdefs so that SDL_msctf.h inclusion is
always recognized correctly.
Fixes Google Chrome, etc, freezing up when SDL owns the clipboard selection
and actually sends thems the correct text for pasting. Confirmed working with
Unicode strings in UTF-8 format.
There were a few tweaks in this patch, but the specific fix is that
event.xselection.target in the SelectionNotify event we send back in reply
must be set to the same atom as the request ("TARGETS" in this case), and
we failed to do that in this special case. Things that don't ask for a target,
like the Gnome Terminal app, worked fine because they don't ask for TARGETS
and just go right to asking for a UTF8_STRING, and Mozilla apparently just
was more liberal in what they accepted in reply.
Chrome would reject our wrong reply and freeze up waiting for a valid one.
Someone should fix that in Chrome, too. :)
Fixes Bugzilla #2926.
The window needs to catch ClientMessage events for one specific window, but
XNextEvent() catches everything, and XWindowEvent doesn't catch ClientMessage,
so we need predicate procedure and XIfEvent() here.
Fixes Bugzilla #2980.
These events accidentally slipping in sometimes appears to be a bug (or
maybe new behavior) in 10.10, as previous versions of Mac OS X don't appear
to ever trigger this.
Thanks to Paulo Marques for pointing out the fix on the SDL mailing list!
Fixes Bugzilla #2842 (again).
Zack Middleton
The change to the keymap to use SDL_SCANCODE_TO_KEYCODE in SDL_x11keyboard.c causes all SDL scancodes without a Usc4 character to be XOR'd with SDLK_SCANCODE_MASK, but not all key code are suppose to be (as seen in include/SDL_keycodes.h). SDLK_BACKSPACE is not 0x4000002A.
I think the full list of keys affected are return, escape, backspace, tab, and delete.
Volumetric
The "Unknown touch device" message appears because the initial touch device setup loop uses SDL_GetTouch() as a guard for calling SDL_AddTouch(). SDL_GetTouch() will always report "Unknown touch device" since the device hasn't been added yet. The SDL_GetTouch() call is unnecessary since SDL_AddTouch() calls SDL_GetTouchIndex() to verify that the device hasn't been added yet, and SDL_GetTouchIndex() has the benefit of not reporting an error for a device it can't find.
Jacob Lee
If a user has a non-standard keyboard mapping -- say, their caps lock key has been mapped to Ctrl -- then SDL_GetModState() is no longer accurate: it only considers the unmapped keys. This is a regression from SDL 1.2.
I think there are two parts to this bug: first, GetModState should use keycodes, rather than scancodes, which is easy enough.
Unfortunately, on my system, SDL considers Caps Lock, even when mapped as Control, to be both SDL_SCANCODE_CAPSLOCK and SDLK_CAPSLOCK. The output from checkkeys for it is:
INFO: Key pressed : scancode 57 = CapsLock, keycode 0x40000039 = CapsLock modifiers: CAPS
Whereas the output for xev is:
KeyPress event, serial 41, synthetic NO, window 0x4a00001,
root 0x9a, subw 0x0, time 40218333, (144,177), root:(1458,222),
state 0x10, keycode 66 (keysym 0xffe3, Control_L), same_screen YES,
XKeysymToKeycode returns keycode: 37
XLookupString gives 0 bytes:
XmbLookupString gives 0 bytes:
XFilterEvent returns: False
I think the problem is that X11_UpdateKeymap in SDL_x11keyboard.c only builds a mapping for keycodes associated with a Unicode character (anything where X11_KeyCodeToUcs returns a value). In the case of caps lock, SDL scancode 57 becomes x11 keycode 66, which becomes x11 keysym 65507(Control_L), which does not have a unicode value.
To fix this, I suspect that SDL needs a mapping of the rest of the x11 keysyms to their corresponding SDL key codes.
Jason Wyatt
Currently the keymapnotify event handling is commented out as FIXME in SDL_x11events.c (It looks like this may have functioned SDL1.2).
Not handling this event means that if a window manager shortcut such as ALT+SPACE is used, SDL will send an ALT key down signal, but not an up signal. Also querying SDL about the key state, it believes the ALT key remains pressed.
X passes the events keypress (alt), ?focusout?, ?focusin?, keymapnotify.
This is a Win32-specific fix for bug 2726. A WinRT fix for this bug was applied
separately, via https://hg.libsdl.org/SDL/rev/91f56dcad879
This fix applies SDL_TOUCH_MOUSEID to 'mouse' events coming from touch devices,
but only when relative-mouse-mode is turned OFF. This bug is still present
when relative-mouse-mode is ON, however Microsoft does not provide documentation
on whether or not those input events (which come from WM_INPUT) can be
identified as touch-specific or not. Unofficially, that data might be available
(via GetMessageExtraInfo()), however this patch only uses MS-documented APIs.
Zack Middleton
Using X11 (on Debian Wheezy), SDL_GetModState() & KMOD_NUM and KMOD_CAPS are not set to system state (numlock/capslock LEDs). Pressing numlock or capslock toggles the mod state, though if num/caps lock is enabled before starting the program it's still reversed from system state. This makes getting KMOD_NUM and KMOD_CAPS in programs unreliable. This can be seen using the checkkeys test program.
The function that appears to have handle this in SDL 1.2 is X11_SetKeyboardState. The function call is commented out with "FIXME:" in SDL 2.
Using Windows backend through WINE; on first key press if numlock and/or capslock is enabled on system, numlock/capslock SDL_SendKeyboardKey is run and toggles KMOD_NUM/KMOD_CAPS to the correct state. On X11 this does not happen.
The attached patch makes X11 backend set keyboard state on window focus if no window was previously focused. It sets all keys to system up/down state and toggles KMOD_NUM/KMOD_CAPS via SDL_SendKeyboardKey to match system if needed. The patch is based on SDL 1.2's X11_SetKeyboardState.
Coriiander
Upon creating a window, a window property is added to it through the Win32-function "SetProp". This is done in the SDL-function "SetupWindowData" in file "src\video\windows\SDL_windowswindow.c".
Whenever you call "SetProp" to add a property to a Win32-window, you should also call the Win32-function "RemoveProp" to remove it before destroying that Win32-window.
While you might think that it's ok and that Windows will clean up nicely itself, it is not ok. It is against all Win32-API guidelines and is mostlikely a leak. Especially on external windows (CreateWindowFrom) you want to have things done right, not messy and leaky, affecting some other module. Even if SDL gets shutdown entirely that external window will now forever still have the "SDL_WindowData" prop attached to it.
The backing scale factor can change when the window moves between retina and non-retina displays.
The only other way to detect such a change is to compare the output of SDL_GL_GetDrawableSize or SDL_GetRendererOutputSize every frame, which is less than desirable, especially since the necessary app logic is likely already being executed when a window resize event is received.
