Short diff, long explaination:
This will hopefully clarify whether or not issues are about the dubgger
like the case of the never ending thread about MacOS "Remote debugger
fails", which started as a real issue, and ended up being referenced for
every strange OSX issue where the window was not visible or the engine
crashed without appearing on screen.
Unified BSD and Winsock sockets into a single implementation of a
generic NetSocket interface.
This is some ground work for few network improvements:
- Reuse as much code as possible between Posix and Windows.
- Provide a single point of implementation for exotic sdks (consoles).
- Provide platform agnostic StreamPeerTCP and PacketPeerUDP in core.
- Implement connect for UDP allowing for DTLS implementation.
This allows more consistency in the manner we include core headers,
where previously there would be a mix of absolute, relative and
include path-dependent includes.
Fixed duplicating AnimationPlayers leaving empty tracks.
I don't think duplicating arrays worked before. There was even this weird
//does not make a long of sense but should work
comment at the changed line. Fixes#21616
A* now exits when next node from open set with least cost happens to be end_point,
not when node with least cost has end_point as a neigbour.
Added two tests for astar:
* ABC tests case where start and end node are
neigbours
* ABCX tests case with intermediate nodes
This average is not a proper approximation of a grayscale value,
get_v() is better suited for that.
If we want a real to_grayscale() conversion, it's somewhat more
involved: https://en.wikipedia.org/wiki/Grayscale
Remove the deprecated Gray() from C# bindings as it conflicts
with new named color constants.
This is to prevent crashes for code like:
...
void Material::set_next_pass(const Ref<Material> &p_pass) {
ERR_FAIL_COND(p_pass == this);
...
that's been fixed in 031f763d4f
They were introduced in #14704 but need more discussion IMO,
they don't strike me as core features that would have to be
registered in Variant directly.
Moreover, they currently break the documentation XML as string
constants end up encoded as e.g. `value=""..""`.
Use a Microsoft recommended way of process termination for the project
process run from the editor. This allows loaded DLLs to receive and handle
DLL_PROCESS_DETACH notification and cleanup any global state before the
process actually exits.
We delete the faces for consideration in this loop but we can still
sometimes find an edge that connects to this face. We now interate over
all edges and disconnect edges connecting to this face.
This fixes#16560 and fixes#17569
- Count and panel per script.
- Ability to disable warnings per script using special comments.
- Ability to disable warnings globally using Project Settings.
- Option to treat enabled warnings as errors.
A couple of null appends got typoed to append a '0' character instead.
Removed them here since String already takes care of the null terminator for us.
Fixes exported property modified values lost when creating a placeholder script instance with a failed script compilation
- Object set/get will call PlaceHolderScriptInstance's new fallback set/get methods as a last resort. This way, placeholder script instances can keep the values for storage or until the script is compiled successfuly.
- Script::can_instance() will only return true if a real script instance can be created. Otherwise, in the case of placeholder script instances, it will return false.
- Object::set_script(script) is now in charge of requesting the creation of placeholder script instances. It's no longer Script::instance_create(owner)'s duty.
- PlaceHolderScriptInstance has a new method set_build_failed(bool) to determine whether it should call into its script methods or not.
- Fixed a few problems during reloading of C# scripts.
When a Vector of Vectors gets resized the 'this' pointer of the Vectors
change. This means that the VectorWriteProxy _parent references get
invalidated. Thanks a lot to @ibrahn for finding the root cause of this.
To fix this we now create a pointer to CowData in Vector (which won't
change when the vectors move) and pass that to the write proxy also.
This fixes#20475
Yesterday, when playing around with my network code, I realized there is
a security issue in decode_variant, at least when decoding PoolArrays.
Basically, the size of the PoolArray is encoded in a uint32_t, when
decoding it, that value is cast to int when comparing if the packet is
actually that size causing numbers with MSB=1 to be interpreted as
negative thus always passing the check. That same value though, is used
as uint32_t again to resize the output vector. For this reason, sending
a malformed packet with declared type PoolByteArray and size of 2^31(+x)
causes the engine to try to allocate 2+GB of pool memory, causing the
engine to crash.
(cherry picked from commit 5262d1bbcc)
- Refactored all builder (make_*) functions into separate Python modules along to the build tree
- Introduced utility function to wrap all invocations on Windows, but does not change it elsewhere
- Introduced stub to use the builders module as a stand alone script and invoke a selected function
There is a problem with file handles related to writing generated content (*.gen.h and *.gen.cpp)
on Windows, which randomly causes a SHARING VIOLATION error to the compiler resulting in flaky
builds. Running all such content generators in a new subprocess instead of directly inside the
build script works around the issue.
Yes, I tried the multiprocessing module. It did not work due to conflict with SCons on cPickle.
Suggested workaround did not fully work either.
Using the run_in_subprocess wrapper on osx and x11 platforms as well for consistency. In case of
running a cross-compilation on Windows they would still be used, but likely it will not happen
in practice. What counts is that the build itself is running on which platform, not the target
platform.
Some generated files are written directly in an SConstruct or SCsub file, before the parallel build starts. They don't need to be written in a subprocess, apparently, so I left them as is.
This commit makes operator[] on Vector const and adds a write proxy to it. From
now on writes to Vectors need to happen through the .write proxy. So for
instance:
Vector<int> vec;
vec.push_back(10);
std::cout << vec[0] << std::endl;
vec.write[0] = 20;
Failing to use the .write proxy will cause a compilation error.
In addition COWable datatypes can now embed a CowData pointer to their data.
This means that String, CharString, and VMap no longer use or derive from
Vector.
_ALWAYS_INLINE_ and _FORCE_INLINE_ are now equivalent for debug and non-debug
builds. This is a lot faster for Vector in the editor and while running tests.
The reason why this difference used to exist is because force-inlined methods
used to give a bad debugging experience. After extensive testing with modern
compilers this is no longer the case.
-Project/Editor settings now show tooltips properly
-Settings thar require restart now will show a restart warning
-Video driver is now visible all the time, can be changed easily
-Added function to request current video driver
Same as erase, but it returns a boolean value indicating whether the pair was erased or not.
This method should be removed during the next compatibility breakage, and 'Dictionary::erase(key)' should be changed to return a boolean.
Previous problems include failure to remove the splitter from the last
element found and a possible infinite loop dealing with multi-character
splitter strings.
Before this change, missing User-Agent and Accept headers were automatically
added on all platforms. Setting the User-Agent header forces the browser to
do a CORS preflight (see 1) which fails if the HTTP endpoint is not
configured appropriate. It's not neccesary to set either header as the
browser sets them and so this commit disables that functionality on the JS
target.
1: https://developer.mozilla.org/en-US/docs/Web/HTTP/CORS#Simple_requests
Initialized the PID to -2, which will be the value returns in blocking-
mode where the PID is not available. (-1 was already taken to signify an
execution failure).
OS::execute will now properly return a non-OK error code when it fails
to execute the target file.
The documentation was rewritten to be very clear about the differences
between blocking and non-blocking mode.
Fixes#19056.
Implements the same heuristic as Curl (and web browsers): if no `Content-Length`, no `Connection: keep-alive` and no chunked transfer encoding, assume th rest of the data until EOF is the body, gracefully setting the HTTP client back to the disconnected state.
Theoretically, this is not compliant with HTTP 1.1, by which `keep-alive` is the default, but in practice, an explicit header is sent by servers.
Adds following functions to the Engine singleton:
get_author_info - names of Godot authors
get_copyright_info - detailed source copyright get_license_info
get_donor_info - donor names
get_license_info - full text of licenses used, indexed by license names
get_license_text - the text of the Godot Expat license
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.
- Adds q/quit option to console debugging
- Adds options (variable_prefix)
- Breaks into debugger with Ctrl-C in local debug mode (Unix/Windows)
- Added option to list all breakpoints
- Fixes add/remove breakpoint bug (invalid path parsing)
- Minor cleanup
- Tool scripts will be executed and can be accessed by plugins.
- Other script languages can implement add/remove_named_global_constant
to make use of this functionality.
If at the time of the _GLOBAL_DEF call a setting itself was unknown,
the function would always return the supplied default value instead of
checking for overrides. This commit changes that, lookup now always
happens which correctly takes overrides into account.
I had a situation coming from godot-python where the caller of
Variant::get_call_error_text() passed null for `p_argptrs`. In
addition to fixing that in the caller, seems like good practice to
defend against that situation in the callee to prevent a crash.
So this patch just substitutes some semi-useful text for the source
type name and keeps going so the user's actual error gets emitted.
Now generating mouse events from touch is optional (on by default) and it's performed by `InputDefault` instead of having each OS abstraction doing it. (*)
The translation algorithm waits for a touch index to be pressed and tracks it translating its events to mouse events until it is raised, while ignoring other pointers.
Furthermore, to avoid an stuck "touch mouse", since not all platforms may report touches raised when the window is unfocused, it checks if touches are still down by the time it's focused again and if so it resets the state of the emulated mouse.
*: In the case of Windows, since it already provides touch-to-mouse translation by itself, "echo" mouse events are filtered out to have it working like the rest.
On X11 a little hack has been needed to avoid a case of a spurious mouse motion event that is generated during touch interaction.
Plus: Improve/fix tracking of current mouse position.
** Summary of changes to settings: **
- `display/window/handheld/emulate_touchscreen` becomes `input/pointing_devices/emulate_touch_from_mouse`
- New setting: `input/pointing_devices/emulate_mouse_from_touch`
Fixes reported logically dead codes by Coverity
* image.cpp: Doesn't really need any modification. But to remove the bug
report then we have to move the MAX call away from the for loop
statement.
* rasterizer_gles3.cpp: Removes unnecessary elif condition since it is
checked earlier in the function
* collada.cpp: If stamement never reached due to macro ERR_CONTINUE does
the same.
* navigation_mesh.cpp: Variables should always be null - however, also
checked for the very same condition in their function call. Leaving this
for review (whether the function call is necessary or not)
* path_editor_plugin.cpp: If cancel is true, then it should restore the
edited value to the original provided.
http://docs.godotengine.org/en/3.0/classes/class_editorspatialgizmo.html#class-editorspatialgizmo-commit-handle
* spatial_editor_gizmos.cpp: the very condition of i >= 3 is
predetermined in the if case right before it. Thus case 1 is always '1'
and case 2 is always '-1'
* grid_map_editor.cpp: Same as above in spatial_editor_gizmos.cpp
* voxel_light_baker.cpp: Same as above in spatial_editor_gizmos.cpp
* visual_server.cpp: Same as above in spatial_editor_gizmos.cpp
* visual_script_expression.cpp: char '-' is already true in the switch
case mechanism. Thus it can never reach to default case.
* particles.cpp: Case 'PARAM_MAX' is unreachable due to index checking
right before the switch execution.
* shader_language.cpp: Invalid index is handled in switch default case.
`type < TYPE_FLOAT && type > TYPE_VEC4` -> `(type < TYPE_FLOAT || type > TYPE_VEC4`)
Fixes the "always false problem" in TODO comment.
Fixes most current reports on Coverity Scan of uninitialized scalar
variable (CWE-457): https://cwe.mitre.org/data/definitions/457.html
These happen most of the time (in our code) when instanciating structs
without a constructor (or with an incomplete one), and later returning
the instance. This is sometimes intended though, as some parameters are
only used in some situations and should not be double-initialized for
performance reasons (e.g. `constant` in ShaderLanguage::Token).
- StringName::StringName(const StringName &p_name)
Non-static class member _data is not initialized in this constructor nor in any functions that it calls.
- StringName::_Data()
Non-static class member idx is not initialized in this constructor nor in any functions that it calls.
- String::num_uint64(...)
This less-than-zero comparison of an unsigned value is never true. n % base < 0UL.
- String::hex_to_int(...) and String::hex_to_int64(...)
Execution cannot reach this statement (deadcode)
Also ensure that get_scale doesn't arbitrarlity change the signs of scales, ensuring that the combination of get_rotation and get_scale gives the correct basis.
Added various missing functions and constructors.
Should close#17968.
Starting from April 2018 Apple no longer accepts apps that do not
support iPhone X. For games this mainly means respecting the safe area,
unobstructed by notch and virtual home button. UI controls must be
placed within the safe area so that users can interact with them.
This commit:
- Adds OS::get_window_safe_area method that returns unobscured area of
the window, where interactive controls should be rendered.
- Reorganizes how launch screens are exported - the previous way was
incorrect and modern iPhones did not pick up the correct screens and
because of that used a non-native resolution to render the game.
- Adds launch screen options for iPhone X.
- Makes launch screens optional in the export template. If not
specified, a white screen will be used.
- Adds App Store icon (1024x1024) export option as it now has to be
bundled with the app instead of being provided in iTunes Connect.
- Fixes crash when launching games in iOS Simulator. It happened because
controllerWasConnected callback came before the engine was
initialized. Now in such case the controllers will be queued up and
registered after initialization is done.
- Fixes issue with the virtual keyboard where for some reason
autocorrection panel would intersect with the keyboard itself and not
allow you to use the top row of the keyboard. This is fixed by
disabling autocorrection altogether.
Closes#17358. Fixes#17428. Fixes#17331.
Add new class _TimerSync to manage timestep calculations.
The new class handles the decisions about simulation progression
previously handled by main::iteration(). It is fed the current timer
ticks and determines how many physics updates are to be run and what
the delta argument to the _process() functions should be.
The new class tries to keep the number of physics updates per frame as
constant as possible from frame to frame. Ideally, it would be N steps
every render frame, but even with perfectly regular rendering, the
general case is that N or N+1 steps are required per frame, for some
fixed N. The best guess for N is stored in typical_physics_steps.
When determining the number of steps to take, no restrictions are
imposed between the choice of typical_physics_steps and
typical_physics_steps+1 steps. Should more or less steps than that be
required, the accumulated remaining time (as before, stored in
time_accum) needs to surpass its boundaries by some minimal threshold.
Once surpassed, typical_physics_steps is updated to allow the new step
count for future updates.
Care is taken that the modified calculation of the number of physics
steps is not observable from game code that only checks the delta
parameters to the _process and _physics_process functions; in addition
to modifying the number of steps, the _process argument is modified as
well to stay in expected bounds. Extra care is taken that the accumulated
steps still sum up to roughly the real elapsed time, up to a maximum
tolerated difference.
To allow the hysteresis code to work correctly on higher refresh
monitors, the number of typical physics steps is not only recorded and
kept consistent for single render frames, but for groups of them.
Currently, up to 12 frames are grouped that way.
The engine parameter physics_jitter_fix controls both the maximum
tolerated difference between wall clock time and summed up _process
arguments and the threshold for changing typical_physics_steps. It is
given in units of the real physics frame slice 1/physics_fps. Set
physics_jitter_fix to 0 to disable the effects of the new code here.
It starts to be effective against the random physics jitter at around
0.02 to 0.05. at values greater than 1 it starts having ill effects on
the engine's ability to react sensibly to dropped frames and framerate
changes.
This is accomplished by setting a special value (-1) to the device variable
in the InputEvent that's being used to compare with the one received from the OS.
This special value is invalid for a regular input, so it should be safe.
Implements #17942
There was a percent-prefixed version, which was exposed, and a http-prefixed version which was not (only to GDNative).
This commit keeps the percent-prefixed versions, but with the http-prefixed implementations.
This commit fixes issue #17585: renaming/moving resources can corrupt
scene files. The corruption was caused by
'ResourceFormatLoaderBinary::rename_dependencies' updating the file
format version field of the affected scene file to the latest version
without actually updating the content of the file to that version,
resulting in a file whose content does not match its file format version
field. The fix preserves the file format version field and the engine
version fields when renaming dependencies.
When adding a directory path to the inventory of the pack, an empty file name was being added to the file list. That made `Directory.get_ntext()` signal end-of-list too early so that files in a subdirectory were missed.
Fixes#15801.
Helps with #16798.
Now the action name is quoted if it contains spaces. Also, quotation
mark (") is added to the forbidden character list for action names, as
it was also a bug.
Fix#17322
The previous logic with VERSION_MKSTRING was a bit unwieldy, so there were
several places hardcoding their own variant of the version string, potentially
with bugs (e.g. forgetting the patch number when defined).
The new logic defines:
- VERSION_BRANCH, the main 'major.minor' version (e.g. 3.1)
- VERSION_NUMBER, which can be 'major.minor' or 'major.minor.patch',
depending on whether the latter is defined (e.g. 3.1.4)
- VERSION_FULL_CONFIG, which contains the version status (e.g. stable)
and the module-specific suffix (e.g. mono)
- VERSION_FULL_BUILD, same as above but with build/reference name
(e.g. official, custom_build, mageia, etc.)
Note: Slight change here, as the previous format had the build name
*before* the module-specific suffix; now it's after
- VERSION_FULL_NAME, same as before, so VERSION_FULL_BUILD prefixed
with "Godot v" for readability
Bugs fixed thanks to that:
- Export templates version matching now properly takes VERSION_PATCH
into account by relying on VERSION_FULL_CONFIG.
- ClassDB hash no longer takes the build name into account, but limits
itself to VERSION_FULL_CONFIG (build name is cosmetic, not relevant
for the API hash).
- Docs XML no longer hardcode the VERSION_STATUS, this was annoying.
- Small cleanup in Windows .rc file thanks to new macros.
Found via `codespell -q 3 --skip="./thirdparty,./editor/translations" -I ../godot-word-whitelist.txt`
Whitelist consists of:
```
ang
doubleclick
lod
nd
que
te
unselect
```
Regression introduced in #16825.
My logic was correct, but not the error code I was expecting.
The error reporting in FileAccess likely needs a review too.
Windows does not fully respect ISO 639-1 like other systems,
so we have to override its locale values for those languages.
Also added comments to document the locale provenance.
The ear clipping algorithm used to triangulate polygons has a slightly too conservative point-in-triangle test which can, in some configurations prevent it from finding a possible tessellation. Relaxing the test by considering that points exactly on edges don't belong the triangle fixes the issue. Changing the semantic of the test is safe because no other code makes use of it. A more detailed explanation can be found in issue #16395.
Fixes#16395.
When running the engine with -d we get a message on the command-line for
each control being clicked. After discussing with @reduz it seems that
this is old and should be removed. Commented out as requested.
Dictionaires did not use the VariantHasher and VariantComparator making
them unsafe for use with NaN values as keys. This PR uses the
appropriate Variant implementations for these functions.
var d = {}
d[Vector2(NAN, NAN)] = 0
d[Vector2(NAN, NAN)] = 0
print(d.size())
will now output '1' and not '2'
This fixes#16031
