In #7839 I see the same error that was fixed in #7833 occuring on the
Windows platform. This moves the audio driver closing to the same place
in OS_Windows::finalize() as it is in OS_X11::finalize()
This fixes#7839
During runtime godot calls chdir() several times. This doesn't really
matter normally but when using tools such as gprof the location of the
profiling data is kind of hard to intuit.
With this PR we simply store the current working directory at start and
restore it once we're almost done exiting.
This doesn't use the OS abstractions as when we need to get the current
workdir we haven't yet initialized it (by necessity). This would break
if we tried to build X11 for windows, but since the X11 target is
hardcoded to use the UNIX abstractions I don't think it matters.
The audio driver cleanup needs to happen at the start of finish
otherwise a race still seems to exist with the destruction of the
audioserver. I think that destroying the X resoures before has something
to do with it.
The audiodrivers loaded by OS_X11 are not destroyed before the
audioserver is. This causes a segfault on exit.
The code is taken from os_windows.cpp which did have the cleanup code.
This fixes HashMap where a key or part of a key is a floating point
number. To fix this the following has been done:
* HashMap now takes an extra template argument Comparator. This class
gets used to compare keys. The default Comperator now works correctly
for common types and floating point numbets.
* Variant implements ::hash_compare() now. This function implements
nan-safe comparison for all types with components that contain floating
point numbers.
* Variant now has a VariantComparator which uses Variant::hash_compare()
safely compare floating point components of variant's types.
* The hash functions for floating point numbers will now normalize NaN
values so that all floating point numbers that are NaN hash to the same
value.
C++ module writers that want to use HashMap internally in their modules
can now also safeguard against this crash by defining their on
Comperator class that safely compares their types.
GDScript users, or writers of modules that don't use HashMap internally
in their modules don't need to do anything.
This fixes#7354 and fixes#6947.
In preload() parsing this code will lookup the identifier in the local
constant database. If the identifier corresponds to a string constant
it is used as the path for preload().
Currently this does not work for global constants, only constants
declared in the same class as the preload is happening. We can implement
a full fix too. Maybe we can use this PR to discuss the possibilities.
This (partially) fixes#6798
Adds an additional flag to SceneTree::create_timer() that tells it whether or not to process when the game is paused.
Defaults to false in order to not break existing functionality.
After discussing this with Reduz this seemed like the best way to
fix#7354. This will make composite values that contain NaN in the same
places as well as the same other values compare as the same.
Additionally non-composite values now also compare equal if they are
both NaN. This breaks IEEE specifications but this is probably what most
users expect. There is a GDScript function check for NaN if the user
needs this information.
This fixes#7354 and probably also fixes#6947
This is a continuation of an on-going work for 64-bit floating point builds, started in PR #7528. Covers physics, physics/joints and physics_2d code.
Also removed matrixToEulerXYZ function in favor of Basis::get_euler.
The code would get a pointer to the beginning of the call_args by using
operator[] at the stack Vector. This does bound checking. When there are
no call_args this bound check fails and the error mentioned in #7796
gets triggered.
This bound check is actually not necessary as call_args just gets set to
NULL and never dereferenced. This new code will just unconditionally set
the pointer to the place where the call_args are if there are any. There
is no NULL check for call_args anywhere so this is safe.
Fixes#7796
