Since we have explicit .gitignore rules for QT creator and several other
editor-specific swap/utility files I'd like to add Eclipse CDT projects
also as this is my editor of choice.
(cherry picked from commit 90ab7162e7)
The 'pos' variable passed to get_color() and get_offset() can be
negative if the color ramp itself is empty. This causes a lookup in an
empty position in the color Vector which leads to a crash.
We add a check so we never do a lookup in the color Vector if the
gradient is empty.
This fixes#10501
(cherry picked from commit 67b9d6eef2)
Fixes#10539 in a diplomatic way (as opposed to telling Gentoo users
to just change their distro or rebuild zlib from upstream...).
(cherry picked from commit 67e6c1b735)
Godot-specific changes are undone in this commit, they will be readded
and properly documented in the following commit.
(cherry picked from commit 5681b7b60a)
This is due to a Godot bug which is now fixed in the master branch, but would require
too much work for a short-lived branch such as 2.1. Building release binaries with GCC 6+
will trigger segfaults due to UB when dereferencing pointers with Object::cast_to.
In theory this check should also be made for Windows cross-compilation with MinGW,
but I am more concerned about Linux distros shipping broken binaries than adventurous
cross-compilers hitting a wall.
- Change get_rect_region to return UV region instead of pixel rect
- Fixed atlas texture Sprite3D Y offsets being applied in the wrong direction and fixed Flip not adjusting the geometry.
- Changed get_rect_region name to get_rect_region_uv_rect
The change in #10524 subtly changes the behavior of set_motion_scale()
and set_motion_offset() if the ParallaxLayer does not have a parent
node. Previously it would still set the corresponding property, but
after this change the property change would be discarded.
I'm not entirely sure if this actually matters as there doesn't appear
to be any code that picks up this change if the ParallaxLayer gets
re-parented later, but it's better to not change behavior regardless.
This backports the work in #7815 and the subsequent fixes in #8393
The following program now works as expected in this branch in both
release_debug and debug mode:
```gdscript
print(sqrt(-1))
print(sqrt(-1))
var simple1=asin(10.0)
var simple2=acos(10.0)
print(simple1)
print(simple2)
```
And successfully prints -nan 4 times
This fixes#9580 and fixes#8925