Before the initialization finishes, AdjustPosition could return Inf
values and in this case AdjustPosition's returning values don't make
sense. Let's return NaN in this case.
Closes#1545
This change doesn't mean we won't support Nintendo Switch. However,
until we finish preparing the developing environment, let's not
mention about Nintendo Switch yet.
Updates #744
QuadVertices or verticesBackend.slice reused its backend slice.
This caused a race condition. QuadVertices can be accessed from
multiple goroutines, and resetting the head and copying the data
at internal/graphicscommand might not be synced.
This change fixes this issue by basically reverting
9cb631e30f.
Closes#1546
Apparently 'uinput-fpc' is a fingerprint reader and is recognized as
a joystick unexpectedly. This was often a primary gamepad on Ebiten
and confused Ebiten applications.
This change fixes this issue by skipping 'uinput-fpc' at the joystick
detection.
Closes#1542
When an image was used in both ways source and destination, Ebiten
worked ineffectively since Ebiten tried to make the image on a texture
atlas after the image was used as a source for a while, which tried to
create a new internal texture every time.
This fix mitigates this issue by using an exponential way to determine
whether the image should be on a texture atlas again or not, so that
an image often used as a destination will require much longer time to
become on a texture atlas again.
Updates #1464
createWindow created a window and also registered the SetSize callback.
This was problematic in setWindowSize, since the callback was invoked
in setWindowSize unexpectedly.
This change separates createWindow and registring the callback so that
the created window in setWindowSize doesn't invoke the callback until
setWindowSize finishes.
Closes#1505