Initialised relevant variables before stating thread,
to prevent a branch on uninitialised data.
Fixed race condition in polling that could miss a device change.
(cherry picked from commit fe4265ad46)
When the viewport's size.y becomes lower than 2, the storage->frame.current_rt->effects.mip_maps[0].sizes Vector during rendering becomes empty, resulting in crashes in at least GLES3. This is a temporary fix to stop rendering a viewport when its size is below 2 rather than below 1.
(cherry picked from commit 892a4b175a)
KEY_MASK_CMD is automatically replaced by KEY_MASK_CTRL on non-OSX
and KEY_MASK_META (Command key) on OSX, so it should be used for all
Ctrl/Cmd + key shortcuts.
Also de-hacked the macOS shortcut replacements with proper conditional
definition. Not tested on macOS, cannot judge if they are good shortcuts.
Fixes#10761.
(cherry picked from commit 3f09cac267)
The code had a subtle signed/unsigned bug -
```cpp
if( signed - unsigned < 0)
// signed - unsigned is unsigned in c++, so
if( unsigned < 0)
// and thus the if block will never be executed
```
Thus all the following code would be ran, including unnecessary retries
of compacting the pool.
(cherry picked from commit 2bbe6144ff)
Physics2DDirectSpaceStateSW was applying the result limit to broadphase
collision detection instead of narrow. This is inconsistent with its 3D
variant, as well as the rest of the 2D direct space state functions.
Broadphase is now limited by INTERSECTION_QUERY_MAX like everything else,
and narrow phase is exited early when the result limit has been reached.
(cherry picked from commit 1ba106a71e)
Documents CollisionObject2D mouse_entered, mouse_exited and input_event requiring at least one collision_layer to be set.
(cherry picked from commit da73bcca6f)
We want to add the individual strings to the list
and not add a list object to the list.
Without this patch, sorting failed because "str < list"
is not a valid operation in python.
(cherry picked from commit f312582326)
Otherwise we run into situations where commits to stable branches
induce very long build times, as they have to basically build from
scratch but also invalidate the cache for future commits on the
master branch.
This commit also makes the cache folder branch-specific, but since
it's still limited to 1 GB of total cache size, we don't enable it
for non-master, as we would still run into issues with non-master
build invalidating the master cache.
(cherry picked from commit b021bdbf1f)