godot/core/io/resource_format_binary.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

2545 lines
70 KiB
C++
Raw Normal View History

2014-02-09 22:10:30 -03:00
/**************************************************************************/
/* resource_format_binary.cpp */
/**************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* https://godotengine.org */
/**************************************************************************/
/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */
/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */
/* */
/* Permission is hereby granted, free of charge, to any person obtaining */
/* a copy of this software and associated documentation files (the */
/* "Software"), to deal in the Software without restriction, including */
/* without limitation the rights to use, copy, modify, merge, publish, */
/* distribute, sublicense, and/or sell copies of the Software, and to */
/* permit persons to whom the Software is furnished to do so, subject to */
/* the following conditions: */
/* */
/* The above copyright notice and this permission notice shall be */
/* included in all copies or substantial portions of the Software. */
/* */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
/**************************************************************************/
2014-02-09 22:10:30 -03:00
#include "resource_format_binary.h"
#include "core/config/project_settings.h"
#include "core/io/dir_access.h"
#include "core/io/file_access_compressed.h"
#include "core/io/missing_resource.h"
#include "core/object/script_language.h"
#include "core/version.h"
#include "scene/property_utils.h"
#include "scene/resources/packed_scene.h"
2014-02-09 22:10:30 -03:00
//#define print_bl(m_what) print_line(m_what)
#define print_bl(m_what) (void)(m_what)
2014-02-09 22:10:30 -03:00
enum {
//numbering must be different from variant, in case new variant types are added (variant must be always contiguous for jumptable optimization)
VARIANT_NIL = 1,
VARIANT_BOOL = 2,
VARIANT_INT = 3,
VARIANT_FLOAT = 4,
2014-02-09 22:10:30 -03:00
VARIANT_STRING = 5,
VARIANT_VECTOR2 = 10,
VARIANT_RECT2 = 11,
VARIANT_VECTOR3 = 12,
VARIANT_PLANE = 13,
2021-01-20 07:02:02 +00:00
VARIANT_QUATERNION = 14,
2017-11-16 21:09:00 -05:00
VARIANT_AABB = 15,
VARIANT_BASIS = 16,
VARIANT_TRANSFORM3D = 17,
VARIANT_TRANSFORM2D = 18,
2014-02-09 22:10:30 -03:00
VARIANT_COLOR = 20,
VARIANT_NODE_PATH = 22,
VARIANT_RID = 23,
VARIANT_OBJECT = 24,
VARIANT_INPUT_EVENT = 25,
VARIANT_DICTIONARY = 26,
VARIANT_ARRAY = 30,
VARIANT_PACKED_BYTE_ARRAY = 31,
VARIANT_PACKED_INT32_ARRAY = 32,
VARIANT_PACKED_FLOAT32_ARRAY = 33,
VARIANT_PACKED_STRING_ARRAY = 34,
VARIANT_PACKED_VECTOR3_ARRAY = 35,
VARIANT_PACKED_COLOR_ARRAY = 36,
VARIANT_PACKED_VECTOR2_ARRAY = 37,
VARIANT_INT64 = 40,
VARIANT_DOUBLE = 41,
VARIANT_CALLABLE = 42,
VARIANT_SIGNAL = 43,
VARIANT_STRING_NAME = 44,
VARIANT_VECTOR2I = 45,
VARIANT_RECT2I = 46,
VARIANT_VECTOR3I = 47,
VARIANT_PACKED_INT64_ARRAY = 48,
VARIANT_PACKED_FLOAT64_ARRAY = 49,
VARIANT_VECTOR4 = 50,
VARIANT_VECTOR4I = 51,
VARIANT_PROJECTION = 52,
VARIANT_PACKED_VECTOR4_ARRAY = 53,
2014-02-09 22:10:30 -03:00
OBJECT_EMPTY = 0,
OBJECT_EXTERNAL_RESOURCE = 1,
OBJECT_INTERNAL_RESOURCE = 2,
OBJECT_EXTERNAL_RESOURCE_INDEX = 3,
// Version 2: Added 64-bit support for float and int.
// Version 3: Changed NodePath encoding.
// Version 4: New string ID for ext/subresources, breaks forward compat.
// Version 5: Ability to store script class in the header.
// Version 6: Added PackedVector4Array Variant type.
FORMAT_VERSION = 6,
FORMAT_VERSION_CAN_RENAME_DEPS = 1,
FORMAT_VERSION_NO_NODEPATH_PROPERTY = 3,
2014-02-09 22:10:30 -03:00
};
void ResourceLoaderBinary::_advance_padding(uint32_t p_len) {
2014-02-09 22:10:30 -03:00
uint32_t extra = 4 - (p_len % 4);
if (extra < 4) {
for (uint32_t i = 0; i < extra; i++) {
2014-02-09 22:10:30 -03:00
f->get_8(); //pad to 32
}
2014-02-09 22:10:30 -03:00
}
}
static Error read_reals(real_t *dst, Ref<FileAccess> &f, size_t count) {
if (f->real_is_double) {
if constexpr (sizeof(real_t) == 8) {
// Ideal case with double-precision
f->get_buffer((uint8_t *)dst, count * sizeof(double));
#ifdef BIG_ENDIAN_ENABLED
{
uint64_t *dst = (uint64_t *)dst;
for (size_t i = 0; i < count; i++) {
dst[i] = BSWAP64(dst[i]);
}
}
#endif
} else if constexpr (sizeof(real_t) == 4) {
// May be slower, but this is for compatibility. Eventually the data should be converted.
for (size_t i = 0; i < count; ++i) {
dst[i] = f->get_double();
}
} else {
ERR_FAIL_V_MSG(ERR_UNAVAILABLE, "real_t size is neither 4 nor 8!");
}
} else {
if constexpr (sizeof(real_t) == 4) {
// Ideal case with float-precision
f->get_buffer((uint8_t *)dst, count * sizeof(float));
#ifdef BIG_ENDIAN_ENABLED
{
uint32_t *dst = (uint32_t *)dst;
for (size_t i = 0; i < count; i++) {
dst[i] = BSWAP32(dst[i]);
}
}
#endif
} else if constexpr (sizeof(real_t) == 8) {
for (size_t i = 0; i < count; ++i) {
dst[i] = f->get_float();
}
} else {
ERR_FAIL_V_MSG(ERR_UNAVAILABLE, "real_t size is neither 4 nor 8!");
}
}
return OK;
}
StringName ResourceLoaderBinary::_get_string() {
uint32_t id = f->get_32();
if (id & 0x80000000) {
2017-12-15 09:04:57 -03:00
uint32_t len = id & 0x7FFFFFFF;
if ((int)len > str_buf.size()) {
str_buf.resize(len);
}
if (len == 0) {
return StringName();
}
f->get_buffer((uint8_t *)&str_buf[0], len);
return String::utf8(&str_buf[0], len);
}
return string_map[id];
}
Error ResourceLoaderBinary::parse_variant(Variant &r_v) {
uint32_t prop_type = f->get_32();
print_bl("find property of type: " + itos(prop_type));
2014-02-09 22:10:30 -03:00
switch (prop_type) {
2014-02-09 22:10:30 -03:00
case VARIANT_NIL: {
r_v = Variant();
} break;
case VARIANT_BOOL: {
r_v = bool(f->get_32());
} break;
case VARIANT_INT: {
r_v = int(f->get_32());
} break;
case VARIANT_INT64: {
r_v = int64_t(f->get_64());
} break;
case VARIANT_FLOAT: {
2014-02-09 22:10:30 -03:00
r_v = f->get_real();
} break;
case VARIANT_DOUBLE: {
r_v = f->get_double();
} break;
2014-02-09 22:10:30 -03:00
case VARIANT_STRING: {
r_v = get_unicode_string();
} break;
case VARIANT_VECTOR2: {
Vector2 v;
v.x = f->get_real();
v.y = f->get_real();
r_v = v;
} break;
case VARIANT_VECTOR2I: {
Vector2i v;
v.x = f->get_32();
v.y = f->get_32();
r_v = v;
2014-02-09 22:10:30 -03:00
} break;
case VARIANT_RECT2: {
Rect2 v;
v.position.x = f->get_real();
v.position.y = f->get_real();
2014-02-09 22:10:30 -03:00
v.size.x = f->get_real();
v.size.y = f->get_real();
r_v = v;
} break;
case VARIANT_RECT2I: {
Rect2i v;
v.position.x = f->get_32();
v.position.y = f->get_32();
v.size.x = f->get_32();
v.size.y = f->get_32();
r_v = v;
2014-02-09 22:10:30 -03:00
} break;
case VARIANT_VECTOR3: {
Vector3 v;
v.x = f->get_real();
v.y = f->get_real();
v.z = f->get_real();
r_v = v;
} break;
case VARIANT_VECTOR3I: {
Vector3i v;
v.x = f->get_32();
v.y = f->get_32();
v.z = f->get_32();
r_v = v;
} break;
case VARIANT_VECTOR4: {
Vector4 v;
v.x = f->get_real();
v.y = f->get_real();
v.z = f->get_real();
v.w = f->get_real();
r_v = v;
} break;
case VARIANT_VECTOR4I: {
Vector4i v;
v.x = f->get_32();
v.y = f->get_32();
v.z = f->get_32();
v.w = f->get_32();
r_v = v;
} break;
2014-02-09 22:10:30 -03:00
case VARIANT_PLANE: {
Plane v;
v.normal.x = f->get_real();
v.normal.y = f->get_real();
v.normal.z = f->get_real();
v.d = f->get_real();
2014-02-09 22:10:30 -03:00
r_v = v;
} break;
2021-01-20 07:02:02 +00:00
case VARIANT_QUATERNION: {
Quaternion v;
2014-02-09 22:10:30 -03:00
v.x = f->get_real();
v.y = f->get_real();
v.z = f->get_real();
v.w = f->get_real();
r_v = v;
} break;
2017-11-16 21:09:00 -05:00
case VARIANT_AABB: {
AABB v;
v.position.x = f->get_real();
v.position.y = f->get_real();
v.position.z = f->get_real();
2014-02-09 22:10:30 -03:00
v.size.x = f->get_real();
v.size.y = f->get_real();
v.size.z = f->get_real();
r_v = v;
} break;
case VARIANT_TRANSFORM2D: {
Transform2D v;
v.columns[0].x = f->get_real();
v.columns[0].y = f->get_real();
v.columns[1].x = f->get_real();
v.columns[1].y = f->get_real();
v.columns[2].x = f->get_real();
v.columns[2].y = f->get_real();
2014-02-09 22:10:30 -03:00
r_v = v;
} break;
case VARIANT_BASIS: {
Basis v;
2022-04-24 17:07:35 -05:00
v.rows[0].x = f->get_real();
v.rows[0].y = f->get_real();
v.rows[0].z = f->get_real();
v.rows[1].x = f->get_real();
v.rows[1].y = f->get_real();
v.rows[1].z = f->get_real();
v.rows[2].x = f->get_real();
v.rows[2].y = f->get_real();
v.rows[2].z = f->get_real();
2014-02-09 22:10:30 -03:00
r_v = v;
} break;
case VARIANT_TRANSFORM3D: {
Transform3D v;
2022-04-24 17:07:35 -05:00
v.basis.rows[0].x = f->get_real();
v.basis.rows[0].y = f->get_real();
v.basis.rows[0].z = f->get_real();
v.basis.rows[1].x = f->get_real();
v.basis.rows[1].y = f->get_real();
v.basis.rows[1].z = f->get_real();
v.basis.rows[2].x = f->get_real();
v.basis.rows[2].y = f->get_real();
v.basis.rows[2].z = f->get_real();
2014-02-09 22:10:30 -03:00
v.origin.x = f->get_real();
v.origin.y = f->get_real();
v.origin.z = f->get_real();
r_v = v;
} break;
case VARIANT_PROJECTION: {
Projection v;
v.columns[0].x = f->get_real();
v.columns[0].y = f->get_real();
v.columns[0].z = f->get_real();
v.columns[0].w = f->get_real();
v.columns[1].x = f->get_real();
v.columns[1].y = f->get_real();
v.columns[1].z = f->get_real();
v.columns[1].w = f->get_real();
v.columns[2].x = f->get_real();
v.columns[2].y = f->get_real();
v.columns[2].z = f->get_real();
v.columns[2].w = f->get_real();
v.columns[3].x = f->get_real();
v.columns[3].y = f->get_real();
v.columns[3].z = f->get_real();
v.columns[3].w = f->get_real();
r_v = v;
} break;
2014-02-09 22:10:30 -03:00
case VARIANT_COLOR: {
2021-01-25 14:46:35 -05:00
Color v; // Colors should always be in single-precision.
v.r = f->get_float();
v.g = f->get_float();
v.b = f->get_float();
v.a = f->get_float();
2014-02-09 22:10:30 -03:00
r_v = v;
} break;
case VARIANT_STRING_NAME: {
r_v = StringName(get_unicode_string());
} break;
2014-02-09 22:10:30 -03:00
case VARIANT_NODE_PATH: {
Vector<StringName> names;
Vector<StringName> subnames;
bool absolute;
int name_count = f->get_16();
uint32_t subname_count = f->get_16();
absolute = subname_count & 0x8000;
subname_count &= 0x7FFF;
if (ver_format < FORMAT_VERSION_NO_NODEPATH_PROPERTY) {
subname_count += 1; // has a property field, so we should count it as well
}
2014-02-09 22:10:30 -03:00
for (int i = 0; i < name_count; i++) {
names.push_back(_get_string());
}
for (uint32_t i = 0; i < subname_count; i++) {
subnames.push_back(_get_string());
}
2014-02-09 22:10:30 -03:00
NodePath np = NodePath(names, subnames, absolute);
2014-02-09 22:10:30 -03:00
r_v = np;
} break;
case VARIANT_RID: {
r_v = f->get_32();
} break;
case VARIANT_OBJECT: {
uint32_t objtype = f->get_32();
2014-02-09 22:10:30 -03:00
switch (objtype) {
2014-02-09 22:10:30 -03:00
case OBJECT_EMPTY: {
//do none
} break;
case OBJECT_INTERNAL_RESOURCE: {
uint32_t index = f->get_32();
String path;
if (using_named_scene_ids) { // New format.
ERR_FAIL_INDEX_V((int)index, internal_resources.size(), ERR_PARSE_ERROR);
path = internal_resources[index].path;
} else {
path += res_path + "::" + itos(index);
}
//always use internal cache for loading internal resources
if (!internal_index_cache.has(path)) {
WARN_PRINT(vformat("Couldn't load resource (no cache): %s.", path));
r_v = Variant();
} else {
r_v = internal_index_cache[path];
2014-02-09 22:10:30 -03:00
}
} break;
case OBJECT_EXTERNAL_RESOURCE: {
//old file format, still around for compatibility
2014-02-09 22:10:30 -03:00
String exttype = get_unicode_string();
2014-02-09 22:10:30 -03:00
String path = get_unicode_string();
2022-02-03 21:48:38 +05:45
if (!path.contains("://") && path.is_relative_path()) {
2014-02-09 22:10:30 -03:00
// path is relative to file being loaded, so convert to a resource path
path = ProjectSettings::get_singleton()->localize_path(res_path.get_base_dir().path_join(path));
2014-02-09 22:10:30 -03:00
}
if (remaps.find(path)) {
path = remaps[path];
}
Ref<Resource> res = ResourceLoader::load(path, exttype, cache_mode_for_external);
2014-02-09 22:10:30 -03:00
if (res.is_null()) {
WARN_PRINT(vformat("Couldn't load resource: %s.", path));
2014-02-09 22:10:30 -03:00
}
r_v = res;
} break;
case OBJECT_EXTERNAL_RESOURCE_INDEX: {
//new file format, just refers to an index in the external list
int erindex = f->get_32();
if (erindex < 0 || erindex >= external_resources.size()) {
2019-01-26 22:35:31 +01:00
WARN_PRINT("Broken external resource! (index out of size)");
r_v = Variant();
} else {
Ref<ResourceLoader::LoadToken> &load_token = external_resources.write[erindex].load_token;
if (load_token.is_valid()) { // If not valid, it's OK since then we know this load accepts broken dependencies.
Error err;
Ref<Resource> res = ResourceLoader::_load_complete(*load_token.ptr(), &err);
if (res.is_null()) {
if (!ResourceLoader::is_cleaning_tasks()) {
if (!ResourceLoader::get_abort_on_missing_resources()) {
ResourceLoader::notify_dependency_error(local_path, external_resources[erindex].path, external_resources[erindex].type);
} else {
error = ERR_FILE_MISSING_DEPENDENCIES;
ERR_FAIL_V_MSG(error, vformat("Can't load dependency: '%s'.", external_resources[erindex].path));
}
}
} else {
r_v = res;
}
}
}
2014-02-09 22:10:30 -03:00
} break;
default: {
ERR_FAIL_V(ERR_FILE_CORRUPT);
} break;
}
} break;
case VARIANT_CALLABLE: {
r_v = Callable();
} break;
case VARIANT_SIGNAL: {
r_v = Signal();
} break;
2014-02-09 22:10:30 -03:00
case VARIANT_DICTIONARY: {
uint32_t len = f->get_32();
Dictionary d; //last bit means shared
len &= 0x7FFFFFFF;
for (uint32_t i = 0; i < len; i++) {
2014-02-09 22:10:30 -03:00
Variant key;
Error err = parse_variant(key);
2019-09-25 10:28:50 +02:00
ERR_FAIL_COND_V_MSG(err, ERR_FILE_CORRUPT, "Error when trying to parse Variant.");
2014-02-09 22:10:30 -03:00
Variant value;
err = parse_variant(value);
2019-09-25 10:28:50 +02:00
ERR_FAIL_COND_V_MSG(err, ERR_FILE_CORRUPT, "Error when trying to parse Variant.");
2014-02-09 22:10:30 -03:00
d[key] = value;
}
r_v = d;
} break;
case VARIANT_ARRAY: {
uint32_t len = f->get_32();
Array a; //last bit means shared
len &= 0x7FFFFFFF;
2014-02-09 22:10:30 -03:00
a.resize(len);
for (uint32_t i = 0; i < len; i++) {
2014-02-09 22:10:30 -03:00
Variant val;
Error err = parse_variant(val);
2019-09-25 10:28:50 +02:00
ERR_FAIL_COND_V_MSG(err, ERR_FILE_CORRUPT, "Error when trying to parse Variant.");
2014-02-09 22:10:30 -03:00
a[i] = val;
}
r_v = a;
} break;
case VARIANT_PACKED_BYTE_ARRAY: {
2014-02-09 22:10:30 -03:00
uint32_t len = f->get_32();
Vector<uint8_t> array;
2014-02-09 22:10:30 -03:00
array.resize(len);
uint8_t *w = array.ptrw();
f->get_buffer(w, len);
2014-02-09 22:10:30 -03:00
_advance_padding(len);
2014-02-09 22:10:30 -03:00
r_v = array;
} break;
case VARIANT_PACKED_INT32_ARRAY: {
2014-02-09 22:10:30 -03:00
uint32_t len = f->get_32();
Vector<int32_t> array;
2014-02-09 22:10:30 -03:00
array.resize(len);
int32_t *w = array.ptrw();
f->get_buffer((uint8_t *)w, len * sizeof(int32_t));
2014-02-09 22:10:30 -03:00
#ifdef BIG_ENDIAN_ENABLED
{
uint32_t *ptr = (uint32_t *)w.ptr();
for (int i = 0; i < len; i++) {
ptr[i] = BSWAP32(ptr[i]);
}
}
#endif
2014-02-09 22:10:30 -03:00
r_v = array;
} break;
case VARIANT_PACKED_INT64_ARRAY: {
2014-02-09 22:10:30 -03:00
uint32_t len = f->get_32();
Vector<int64_t> array;
2014-02-09 22:10:30 -03:00
array.resize(len);
int64_t *w = array.ptrw();
f->get_buffer((uint8_t *)w, len * sizeof(int64_t));
#ifdef BIG_ENDIAN_ENABLED
{
uint64_t *ptr = (uint64_t *)w.ptr();
for (int i = 0; i < len; i++) {
ptr[i] = BSWAP64(ptr[i]);
}
}
#endif
r_v = array;
} break;
case VARIANT_PACKED_FLOAT32_ARRAY: {
uint32_t len = f->get_32();
Vector<float> array;
array.resize(len);
float *w = array.ptrw();
f->get_buffer((uint8_t *)w, len * sizeof(float));
2014-02-09 22:10:30 -03:00
#ifdef BIG_ENDIAN_ENABLED
{
uint32_t *ptr = (uint32_t *)w.ptr();
for (int i = 0; i < len; i++) {
ptr[i] = BSWAP32(ptr[i]);
}
}
#endif
r_v = array;
} break;
case VARIANT_PACKED_FLOAT64_ARRAY: {
uint32_t len = f->get_32();
Vector<double> array;
array.resize(len);
double *w = array.ptrw();
f->get_buffer((uint8_t *)w, len * sizeof(double));
#ifdef BIG_ENDIAN_ENABLED
{
uint64_t *ptr = (uint64_t *)w.ptr();
for (int i = 0; i < len; i++) {
ptr[i] = BSWAP64(ptr[i]);
}
}
2014-02-09 22:10:30 -03:00
#endif
r_v = array;
} break;
case VARIANT_PACKED_STRING_ARRAY: {
2014-02-09 22:10:30 -03:00
uint32_t len = f->get_32();
Vector<String> array;
2014-02-09 22:10:30 -03:00
array.resize(len);
String *w = array.ptrw();
for (uint32_t i = 0; i < len; i++) {
2014-02-09 22:10:30 -03:00
w[i] = get_unicode_string();
}
2014-02-09 22:10:30 -03:00
r_v = array;
} break;
case VARIANT_PACKED_VECTOR2_ARRAY: {
2014-02-09 22:10:30 -03:00
uint32_t len = f->get_32();
Vector<Vector2> array;
2014-02-09 22:10:30 -03:00
array.resize(len);
Vector2 *w = array.ptrw();
static_assert(sizeof(Vector2) == 2 * sizeof(real_t));
const Error err = read_reals(reinterpret_cast<real_t *>(w), f, len * 2);
ERR_FAIL_COND_V(err != OK, err);
2014-02-09 22:10:30 -03:00
r_v = array;
} break;
case VARIANT_PACKED_VECTOR3_ARRAY: {
2014-02-09 22:10:30 -03:00
uint32_t len = f->get_32();
Vector<Vector3> array;
2014-02-09 22:10:30 -03:00
array.resize(len);
Vector3 *w = array.ptrw();
static_assert(sizeof(Vector3) == 3 * sizeof(real_t));
const Error err = read_reals(reinterpret_cast<real_t *>(w), f, len * 3);
ERR_FAIL_COND_V(err != OK, err);
2014-02-09 22:10:30 -03:00
r_v = array;
} break;
case VARIANT_PACKED_COLOR_ARRAY: {
2014-02-09 22:10:30 -03:00
uint32_t len = f->get_32();
Vector<Color> array;
2014-02-09 22:10:30 -03:00
array.resize(len);
Color *w = array.ptrw();
// Colors always use `float` even with double-precision support enabled
static_assert(sizeof(Color) == 4 * sizeof(float));
f->get_buffer((uint8_t *)w, len * sizeof(float) * 4);
2014-02-09 22:10:30 -03:00
#ifdef BIG_ENDIAN_ENABLED
{
uint32_t *ptr = (uint32_t *)w.ptr();
for (int i = 0; i < len * 4; i++) {
ptr[i] = BSWAP32(ptr[i]);
2014-02-09 22:10:30 -03:00
}
}
2014-02-09 22:10:30 -03:00
#endif
r_v = array;
} break;
case VARIANT_PACKED_VECTOR4_ARRAY: {
uint32_t len = f->get_32();
Vector<Vector4> array;
array.resize(len);
Vector4 *w = array.ptrw();
static_assert(sizeof(Vector4) == 4 * sizeof(real_t));
const Error err = read_reals(reinterpret_cast<real_t *>(w), f, len * 4);
ERR_FAIL_COND_V(err != OK, err);
r_v = array;
} break;
2014-02-09 22:10:30 -03:00
default: {
ERR_FAIL_V(ERR_FILE_CORRUPT);
} break;
}
return OK; //never reach anyway
}
Ref<Resource> ResourceLoaderBinary::get_resource() {
2014-02-09 22:10:30 -03:00
return resource;
}
Error ResourceLoaderBinary::load() {
if (error != OK) {
2014-02-09 22:10:30 -03:00
return error;
}
2014-02-09 22:10:30 -03:00
for (int i = 0; i < external_resources.size(); i++) {
String path = external_resources[i].path;
if (remaps.has(path)) {
path = remaps[path];
}
2014-02-09 22:10:30 -03:00
2022-02-03 21:48:38 +05:45
if (!path.contains("://") && path.is_relative_path()) {
// path is relative to file being loaded, so convert to a resource path
path = ProjectSettings::get_singleton()->localize_path(path.get_base_dir().path_join(external_resources[i].path));
}
external_resources.write[i].path = path; //remap happens here, not on load because on load it can actually be used for filesystem dock resource remap
external_resources.write[i].load_token = ResourceLoader::_load_start(path, external_resources[i].type, use_sub_threads ? ResourceLoader::LOAD_THREAD_DISTRIBUTE : ResourceLoader::LOAD_THREAD_FROM_CURRENT, cache_mode_for_external);
if (external_resources[i].load_token.is_null()) {
if (!ResourceLoader::get_abort_on_missing_resources()) {
ResourceLoader::notify_dependency_error(local_path, path, external_resources[i].type);
} else {
error = ERR_FILE_MISSING_DEPENDENCIES;
ERR_FAIL_V_MSG(error, vformat("Can't load dependency: '%s'.", path));
}
2014-02-09 22:10:30 -03:00
}
}
for (int i = 0; i < internal_resources.size(); i++) {
bool main = i == (internal_resources.size() - 1);
2014-02-09 22:10:30 -03:00
//maybe it is loaded already
String path;
String id;
2014-02-09 22:10:30 -03:00
if (!main) {
path = internal_resources[i].path;
if (path.begins_with("local://")) {
path = path.replace_first("local://", "");
id = path;
path = res_path + "::" + path;
internal_resources.write[i].path = path; // Update path.
}
if (cache_mode == ResourceFormatLoader::CACHE_MODE_REUSE && ResourceCache::has(path)) {
Ref<Resource> cached = ResourceCache::get_ref(path);
if (cached.is_valid()) {
//already loaded, don't do anything
error = OK;
internal_index_cache[path] = cached;
continue;
}
}
} else {
if (cache_mode != ResourceFormatLoader::CACHE_MODE_IGNORE && !ResourceCache::has(res_path)) {
path = res_path;
}
2014-02-09 22:10:30 -03:00
}
uint64_t offset = internal_resources[i].offset;
2014-02-09 22:10:30 -03:00
f->seek(offset);
2014-02-09 22:10:30 -03:00
String t = get_unicode_string();
2014-02-09 22:10:30 -03:00
Ref<Resource> res;
Resource *r = nullptr;
2014-02-09 22:10:30 -03:00
MissingResource *missing_resource = nullptr;
if (main) {
res = ResourceLoader::get_resource_ref_override(local_path);
r = res.ptr();
}
if (!r) {
if (cache_mode == ResourceFormatLoader::CACHE_MODE_REPLACE && ResourceCache::has(path)) {
//use the existing one
Ref<Resource> cached = ResourceCache::get_ref(path);
if (cached->get_class() == t) {
cached->reset_state();
res = cached;
}
}
2014-02-09 22:10:30 -03:00
if (res.is_null()) {
//did not replace
Object *obj = ClassDB::instantiate(t);
if (!obj) {
if (ResourceLoader::is_creating_missing_resources_if_class_unavailable_enabled()) {
//create a missing resource
missing_resource = memnew(MissingResource);
missing_resource->set_original_class(t);
missing_resource->set_recording_properties(true);
obj = missing_resource;
} else {
error = ERR_FILE_CORRUPT;
ERR_FAIL_V_MSG(ERR_FILE_CORRUPT, vformat("'%s': Resource of unrecognized type in file: '%s'.", local_path, t));
}
}
r = Object::cast_to<Resource>(obj);
if (!r) {
String obj_class = obj->get_class();
error = ERR_FILE_CORRUPT;
memdelete(obj); //bye
ERR_FAIL_V_MSG(ERR_FILE_CORRUPT, vformat("'%s': Resource type in resource field not a resource, type is: %s.", local_path, obj_class));
}
res = Ref<Resource>(r);
}
}
if (r) {
if (!path.is_empty()) {
if (cache_mode != ResourceFormatLoader::CACHE_MODE_IGNORE) {
r->set_path(path, cache_mode == ResourceFormatLoader::CACHE_MODE_REPLACE); // If got here because the resource with same path has different type, replace it.
} else {
r->set_path_cache(path);
}
}
r->set_scene_unique_id(id);
}
2014-02-09 22:10:30 -03:00
if (!main) {
internal_index_cache[path] = res;
}
int pc = f->get_32();
2014-02-09 22:10:30 -03:00
//set properties
2014-02-09 22:10:30 -03:00
Dictionary missing_resource_properties;
for (int j = 0; j < pc; j++) {
StringName name = _get_string();
if (name == StringName()) {
error = ERR_FILE_CORRUPT;
ERR_FAIL_V(ERR_FILE_CORRUPT);
}
2014-02-09 22:10:30 -03:00
Variant value;
2014-02-09 22:10:30 -03:00
error = parse_variant(value);
if (error) {
return error;
}
2014-02-09 22:10:30 -03:00
bool set_valid = true;
if (value.get_type() == Variant::OBJECT && missing_resource == nullptr && ResourceLoader::is_creating_missing_resources_if_class_unavailable_enabled()) {
// If the property being set is a missing resource (and the parent is not),
// then setting it will most likely not work.
// Instead, save it as metadata.
Ref<MissingResource> mr = value;
if (mr.is_valid()) {
missing_resource_properties[name] = mr;
set_valid = false;
}
}
if (value.get_type() == Variant::ARRAY) {
Array set_array = value;
bool is_get_valid = false;
Variant get_value = res->get(name, &is_get_valid);
if (is_get_valid && get_value.get_type() == Variant::ARRAY) {
Array get_array = get_value;
if (!set_array.is_same_typed(get_array)) {
value = Array(set_array, get_array.get_typed_builtin(), get_array.get_typed_class_name(), get_array.get_typed_script());
}
}
}
if (value.get_type() == Variant::DICTIONARY) {
Dictionary set_dict = value;
bool is_get_valid = false;
Variant get_value = res->get(name, &is_get_valid);
if (is_get_valid && get_value.get_type() == Variant::DICTIONARY) {
Dictionary get_dict = get_value;
if (!set_dict.is_same_typed(get_dict)) {
value = Dictionary(set_dict, get_dict.get_typed_key_builtin(), get_dict.get_typed_key_class_name(), get_dict.get_typed_key_script(),
get_dict.get_typed_value_builtin(), get_dict.get_typed_value_class_name(), get_dict.get_typed_value_script());
2023-06-24 13:03:28 -05:00
}
}
}
if (set_valid) {
res->set(name, value);
}
}
if (missing_resource) {
missing_resource->set_recording_properties(false);
}
if (!missing_resource_properties.is_empty()) {
res->set_meta(META_MISSING_RESOURCES, missing_resource_properties);
}
2014-02-09 22:10:30 -03:00
#ifdef TOOLS_ENABLED
res->set_edited(false);
2014-02-09 22:10:30 -03:00
#endif
if (progress) {
*progress = (i + 1) / float(internal_resources.size());
}
2014-02-09 22:10:30 -03:00
resource_cache.push_back(res);
2014-02-09 22:10:30 -03:00
if (main) {
f.unref();
resource = res;
resource->set_as_translation_remapped(translation_remapped);
error = OK;
return OK;
}
2014-02-09 22:10:30 -03:00
}
return ERR_FILE_EOF;
2014-02-09 22:10:30 -03:00
}
void ResourceLoaderBinary::set_translation_remapped(bool p_remapped) {
translation_remapped = p_remapped;
}
static void save_ustring(Ref<FileAccess> f, const String &p_string) {
CharString utf8 = p_string.utf8();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(utf8.length() + 1));
f->store_buffer((const uint8_t *)utf8.get_data(), utf8.length() + 1);
}
static String get_ustring(Ref<FileAccess> f) {
int len = f->get_32();
Vector<char> str_buf;
str_buf.resize(len);
f->get_buffer((uint8_t *)&str_buf[0], len);
return String::utf8(&str_buf[0], len);
}
String ResourceLoaderBinary::get_unicode_string() {
2014-02-09 22:10:30 -03:00
int len = f->get_32();
if (len > str_buf.size()) {
str_buf.resize(len);
}
if (len == 0) {
return String();
}
2014-02-09 22:10:30 -03:00
f->get_buffer((uint8_t *)&str_buf[0], len);
return String::utf8(&str_buf[0], len);
2014-02-09 22:10:30 -03:00
}
void ResourceLoaderBinary::get_classes_used(Ref<FileAccess> p_f, HashSet<StringName> *p_classes) {
open(p_f, false, true);
if (error) {
return;
}
for (const IntResource &res : internal_resources) {
p_f->seek(res.offset);
String t = get_unicode_string();
if (!p_f->get_error() && t != String() && ClassDB::class_exists(t)) {
p_classes->insert(t);
}
}
}
void ResourceLoaderBinary::get_dependencies(Ref<FileAccess> p_f, List<String> *p_dependencies, bool p_add_types) {
open(p_f, false, true);
if (error) {
2014-02-09 22:10:30 -03:00
return;
}
2014-02-09 22:10:30 -03:00
for (int i = 0; i < external_resources.size(); i++) {
String dep;
String fallback_path;
if (external_resources[i].uid != ResourceUID::INVALID_ID) {
dep = ResourceUID::get_singleton()->id_to_text(external_resources[i].uid);
fallback_path = external_resources[i].path; // Used by Dependency Editor, in case uid path fails.
} else {
dep = external_resources[i].path;
}
if (p_add_types && !external_resources[i].type.is_empty()) {
dep += "::" + external_resources[i].type;
}
if (!fallback_path.is_empty()) {
if (!p_add_types) {
dep += "::"; // Ensure that path comes third, even if there is no type.
}
dep += "::" + fallback_path;
}
p_dependencies->push_back(dep);
2014-02-09 22:10:30 -03:00
}
}
void ResourceLoaderBinary::open(Ref<FileAccess> p_f, bool p_no_resources, bool p_keep_uuid_paths) {
error = OK;
2014-02-09 22:10:30 -03:00
f = p_f;
uint8_t header[4];
f->get_buffer(header, 4);
if (header[0] == 'R' && header[1] == 'S' && header[2] == 'C' && header[3] == 'C') {
// Compressed.
Ref<FileAccessCompressed> fac;
fac.instantiate();
error = fac->open_after_magic(f);
if (error != OK) {
f.unref();
ERR_FAIL_MSG(vformat("Failed to open binary resource file: '%s'.", local_path));
}
2014-02-09 22:10:30 -03:00
f = fac;
} else if (header[0] != 'R' || header[1] != 'S' || header[2] != 'R' || header[3] != 'C') {
// Not normal.
2014-02-09 22:10:30 -03:00
error = ERR_FILE_UNRECOGNIZED;
f.unref();
ERR_FAIL_MSG(vformat("Unrecognized binary resource file: '%s'.", local_path));
2014-02-09 22:10:30 -03:00
}
bool big_endian = f->get_32();
bool use_real64 = f->get_32();
f->set_big_endian(big_endian != 0); //read big endian if saved as big endian
2014-02-09 22:10:30 -03:00
uint32_t ver_major = f->get_32();
uint32_t ver_minor = f->get_32();
ver_format = f->get_32();
2014-02-09 22:10:30 -03:00
print_bl("big endian: " + itos(big_endian));
#ifdef BIG_ENDIAN_ENABLED
print_bl("endian swap: " + itos(!big_endian));
#else
print_bl("endian swap: " + itos(big_endian));
#endif
2014-02-09 22:10:30 -03:00
print_bl("real64: " + itos(use_real64));
print_bl("major: " + itos(ver_major));
print_bl("minor: " + itos(ver_minor));
print_bl("format: " + itos(ver_format));
if (ver_format > FORMAT_VERSION || ver_major > GODOT_VERSION_MAJOR) {
f.unref();
ERR_FAIL_MSG(vformat("File '%s' can't be loaded, as it uses a format version (%d) or engine version (%d.%d) which are not supported by your engine version (%s).",
local_path, ver_format, ver_major, ver_minor, GODOT_VERSION_BRANCH));
2014-02-09 22:10:30 -03:00
}
type = get_unicode_string();
print_bl("type: " + type);
importmd_ofs = f->get_64();
uint32_t flags = f->get_32();
if (flags & ResourceFormatSaverBinaryInstance::FORMAT_FLAG_NAMED_SCENE_IDS) {
using_named_scene_ids = true;
}
if (flags & ResourceFormatSaverBinaryInstance::FORMAT_FLAG_UIDS) {
using_uids = true;
}
f->real_is_double = (flags & ResourceFormatSaverBinaryInstance::FORMAT_FLAG_REAL_T_IS_DOUBLE) != 0;
if (using_uids) {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
uid = ResourceUID::ID(f->get_64());
} else {
f->get_64(); // skip over uid field
uid = ResourceUID::INVALID_ID;
}
if (flags & ResourceFormatSaverBinaryInstance::FORMAT_FLAG_HAS_SCRIPT_CLASS) {
script_class = get_unicode_string();
}
2021-07-29 22:24:29 -07:00
for (int i = 0; i < ResourceFormatSaverBinaryInstance::RESERVED_FIELDS; i++) {
2014-02-09 22:10:30 -03:00
f->get_32(); //skip a few reserved fields
}
2014-02-09 22:10:30 -03:00
if (p_no_resources) {
return;
}
2014-02-09 22:10:30 -03:00
uint32_t string_table_size = f->get_32();
string_map.resize(string_table_size);
for (uint32_t i = 0; i < string_table_size; i++) {
StringName s = get_unicode_string();
string_map.write[i] = s;
2014-02-09 22:10:30 -03:00
}
print_bl("strings: " + itos(string_table_size));
uint32_t ext_resources_size = f->get_32();
for (uint32_t i = 0; i < ext_resources_size; i++) {
ExtResource er;
2014-02-09 22:10:30 -03:00
er.type = get_unicode_string();
er.path = get_unicode_string();
if (using_uids) {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
er.uid = ResourceUID::ID(f->get_64());
if (!p_keep_uuid_paths && er.uid != ResourceUID::INVALID_ID) {
if (ResourceUID::get_singleton()->has_id(er.uid)) {
// If a UID is found and the path is valid, it will be used, otherwise, it falls back to the path.
er.path = ResourceUID::get_singleton()->get_id_path(er.uid);
} else {
#ifdef TOOLS_ENABLED
// Silence a warning that can happen during the initial filesystem scan due to cache being regenerated.
if (ResourceLoader::get_resource_uid(res_path) != er.uid) {
WARN_PRINT(vformat("'%s': In external resource #%d, invalid UID: '%s' - using text path instead: '%s'.", res_path, i, ResourceUID::get_singleton()->id_to_text(er.uid), er.path));
}
#else
WARN_PRINT(vformat("'%s': In external resource #%d, invalid UID: '%s' - using text path instead: '%s'.", res_path, i, ResourceUID::get_singleton()->id_to_text(er.uid), er.path));
#endif
}
}
}
2014-02-09 22:10:30 -03:00
external_resources.push_back(er);
}
print_bl("ext resources: " + itos(ext_resources_size));
uint32_t int_resources_size = f->get_32();
for (uint32_t i = 0; i < int_resources_size; i++) {
IntResource ir;
2014-02-09 22:10:30 -03:00
ir.path = get_unicode_string();
ir.offset = f->get_64();
internal_resources.push_back(ir);
}
print_bl("int resources: " + itos(int_resources_size));
if (f->eof_reached()) {
error = ERR_FILE_CORRUPT;
f.unref();
ERR_FAIL_MSG(vformat("Premature end of file (EOF): '%s'.", local_path));
2014-02-09 22:10:30 -03:00
}
}
String ResourceLoaderBinary::recognize(Ref<FileAccess> p_f) {
2014-02-09 22:10:30 -03:00
error = OK;
f = p_f;
uint8_t header[4];
f->get_buffer(header, 4);
if (header[0] == 'R' && header[1] == 'S' && header[2] == 'C' && header[3] == 'C') {
// Compressed.
Ref<FileAccessCompressed> fac;
fac.instantiate();
error = fac->open_after_magic(f);
if (error != OK) {
f.unref();
return "";
}
2014-02-09 22:10:30 -03:00
f = fac;
} else if (header[0] != 'R' || header[1] != 'S' || header[2] != 'R' || header[3] != 'C') {
// Not normal.
error = ERR_FILE_UNRECOGNIZED;
f.unref();
2014-02-09 22:10:30 -03:00
return "";
}
bool big_endian = f->get_32();
f->get_32(); // use_real64
2014-02-09 22:10:30 -03:00
f->set_big_endian(big_endian != 0); //read big endian if saved as big endian
2014-02-09 22:10:30 -03:00
uint32_t ver_major = f->get_32();
f->get_32(); // ver_minor
uint32_t ver_fmt = f->get_32();
2014-02-09 22:10:30 -03:00
if (ver_fmt > FORMAT_VERSION || ver_major > GODOT_VERSION_MAJOR) {
f.unref();
2014-02-09 22:10:30 -03:00
return "";
}
return get_unicode_string();
2014-02-09 22:10:30 -03:00
}
String ResourceLoaderBinary::recognize_script_class(Ref<FileAccess> p_f) {
error = OK;
f = p_f;
uint8_t header[4];
f->get_buffer(header, 4);
if (header[0] == 'R' && header[1] == 'S' && header[2] == 'C' && header[3] == 'C') {
// Compressed.
Ref<FileAccessCompressed> fac;
fac.instantiate();
error = fac->open_after_magic(f);
if (error != OK) {
f.unref();
return "";
}
f = fac;
} else if (header[0] != 'R' || header[1] != 'S' || header[2] != 'R' || header[3] != 'C') {
// Not normal.
error = ERR_FILE_UNRECOGNIZED;
f.unref();
return "";
}
bool big_endian = f->get_32();
f->get_32(); // use_real64
f->set_big_endian(big_endian != 0); //read big endian if saved as big endian
uint32_t ver_major = f->get_32();
f->get_32(); // ver_minor
uint32_t ver_fmt = f->get_32();
if (ver_fmt > FORMAT_VERSION || ver_major > GODOT_VERSION_MAJOR) {
f.unref();
return "";
}
_ALLOW_DISCARD_ get_unicode_string(); // type
f->get_64(); // Metadata offset
uint32_t flags = f->get_32();
f->get_64(); // UID
if (flags & ResourceFormatSaverBinaryInstance::FORMAT_FLAG_HAS_SCRIPT_CLASS) {
return get_unicode_string();
} else {
return String();
}
}
Ref<Resource> ResourceFormatLoaderBinary::load(const String &p_path, const String &p_original_path, Error *r_error, bool p_use_sub_threads, float *r_progress, CacheMode p_cache_mode) {
if (r_error) {
*r_error = ERR_FILE_CANT_OPEN;
}
2014-02-09 22:10:30 -03:00
Error err;
Ref<FileAccess> f = FileAccess::open(p_path, FileAccess::READ, &err);
2014-02-09 22:10:30 -03:00
ERR_FAIL_COND_V_MSG(err != OK, Ref<Resource>(), vformat("Cannot open file '%s'.", p_path));
2014-02-09 22:10:30 -03:00
ResourceLoaderBinary loader;
switch (p_cache_mode) {
case CACHE_MODE_IGNORE:
case CACHE_MODE_REUSE:
case CACHE_MODE_REPLACE:
loader.cache_mode = p_cache_mode;
loader.cache_mode_for_external = CACHE_MODE_REUSE;
break;
case CACHE_MODE_IGNORE_DEEP:
loader.cache_mode = CACHE_MODE_IGNORE;
loader.cache_mode_for_external = p_cache_mode;
break;
case CACHE_MODE_REPLACE_DEEP:
loader.cache_mode = CACHE_MODE_REPLACE;
loader.cache_mode_for_external = p_cache_mode;
break;
}
loader.use_sub_threads = p_use_sub_threads;
loader.progress = r_progress;
String path = !p_original_path.is_empty() ? p_original_path : p_path;
loader.local_path = ProjectSettings::get_singleton()->localize_path(path);
loader.res_path = loader.local_path;
loader.open(f);
err = loader.load();
if (r_error) {
*r_error = err;
}
2014-02-09 22:10:30 -03:00
if (err) {
return Ref<Resource>();
}
return loader.resource;
2014-02-09 22:10:30 -03:00
}
void ResourceFormatLoaderBinary::get_recognized_extensions_for_type(const String &p_type, List<String> *p_extensions) const {
if (p_type.is_empty()) {
2014-02-09 22:10:30 -03:00
get_recognized_extensions(p_extensions);
return;
}
// res files not supported for GDExtension.
if (p_type == "GDExtension") {
return;
}
2014-02-09 22:10:30 -03:00
List<String> extensions;
ClassDB::get_extensions_for_type(p_type, &extensions);
2014-02-09 22:10:30 -03:00
extensions.sort();
for (const String &E : extensions) {
String ext = E.to_lower();
2014-02-09 22:10:30 -03:00
p_extensions->push_back(ext);
}
}
2014-02-09 22:10:30 -03:00
void ResourceFormatLoaderBinary::get_recognized_extensions(List<String> *p_extensions) const {
List<String> extensions;
ClassDB::get_resource_base_extensions(&extensions);
2014-02-09 22:10:30 -03:00
extensions.sort();
for (const String &E : extensions) {
String ext = E.to_lower();
2014-02-09 22:10:30 -03:00
p_extensions->push_back(ext);
}
}
bool ResourceFormatLoaderBinary::handles_type(const String &p_type) const {
return true; //handles all
}
void ResourceFormatLoaderBinary::get_dependencies(const String &p_path, List<String> *p_dependencies, bool p_add_types) {
Ref<FileAccess> f = FileAccess::open(p_path, FileAccess::READ);
ERR_FAIL_COND_MSG(f.is_null(), vformat("Cannot open file '%s'.", p_path));
2014-02-09 22:10:30 -03:00
ResourceLoaderBinary loader;
loader.local_path = ProjectSettings::get_singleton()->localize_path(p_path);
loader.res_path = loader.local_path;
loader.get_dependencies(f, p_dependencies, p_add_types);
}
Error ResourceFormatLoaderBinary::rename_dependencies(const String &p_path, const HashMap<String, String> &p_map) {
Ref<FileAccess> f = FileAccess::open(p_path, FileAccess::READ);
ERR_FAIL_COND_V_MSG(f.is_null(), ERR_CANT_OPEN, vformat("Cannot open file '%s'.", p_path));
Ref<FileAccess> fw;
String local_path = p_path.get_base_dir();
uint8_t header[4];
f->get_buffer(header, 4);
if (header[0] == 'R' && header[1] == 'S' && header[2] == 'C' && header[3] == 'C') {
// Compressed.
Ref<FileAccessCompressed> fac;
fac.instantiate();
Error err = fac->open_after_magic(f);
ERR_FAIL_COND_V_MSG(err != OK, err, vformat("Cannot open file '%s'.", p_path));
f = fac;
Ref<FileAccessCompressed> facw;
facw.instantiate();
facw->configure("RSCC");
err = facw->open_internal(p_path + ".depren", FileAccess::WRITE);
ERR_FAIL_COND_V_MSG(err, ERR_FILE_CORRUPT, vformat("Cannot create file '%s.depren'.", p_path));
fw = facw;
} else if (header[0] != 'R' || header[1] != 'S' || header[2] != 'R' || header[3] != 'C') {
// Not normal.
ERR_FAIL_V_MSG(ERR_FILE_UNRECOGNIZED, vformat("Unrecognized binary resource file '%s'.", local_path));
} else {
fw = FileAccess::open(p_path + ".depren", FileAccess::WRITE);
ERR_FAIL_COND_V_MSG(fw.is_null(), ERR_CANT_CREATE, vformat("Cannot create file '%s.depren'.", p_path));
uint8_t magic[4] = { 'R', 'S', 'R', 'C' };
fw->store_buffer(magic, 4);
}
bool big_endian = f->get_32();
bool use_real64 = f->get_32();
f->set_big_endian(big_endian != 0); //read big endian if saved as big endian
fw->store_32(big_endian);
fw->store_32(use_real64); //use real64
fw->set_big_endian(big_endian != 0);
uint32_t ver_major = f->get_32();
uint32_t ver_minor = f->get_32();
uint32_t ver_format = f->get_32();
if (ver_format < FORMAT_VERSION_CAN_RENAME_DEPS) {
fw.unref();
{
Ref<DirAccess> da = DirAccess::create(DirAccess::ACCESS_FILESYSTEM);
da->remove(p_path + ".depren");
}
// Use the old approach.
WARN_PRINT(vformat("This file is old, so it can't refactor dependencies, opening and resaving '%s'.", p_path));
Error err;
f = FileAccess::open(p_path, FileAccess::READ, &err);
2019-06-20 16:59:48 +02:00
ERR_FAIL_COND_V_MSG(err != OK, ERR_FILE_CANT_OPEN, vformat("Cannot open file '%s'.", p_path));
ResourceLoaderBinary loader;
loader.local_path = ProjectSettings::get_singleton()->localize_path(p_path);
loader.res_path = loader.local_path;
loader.remaps = p_map;
loader.open(f);
err = loader.load();
ERR_FAIL_COND_V(err != ERR_FILE_EOF, ERR_FILE_CORRUPT);
Ref<Resource> res = loader.get_resource();
ERR_FAIL_COND_V(res.is_null(), ERR_FILE_CORRUPT);
2022-06-03 01:33:42 +02:00
return ResourceFormatSaverBinary::singleton->save(res, p_path);
}
if (ver_format > FORMAT_VERSION || ver_major > GODOT_VERSION_MAJOR) {
ERR_FAIL_V_MSG(ERR_FILE_UNRECOGNIZED,
vformat("File '%s' can't be loaded, as it uses a format version (%d) or engine version (%d.%d) which are not supported by your engine version (%s).",
local_path, ver_format, ver_major, ver_minor, GODOT_VERSION_BRANCH));
}
// Since we're not actually converting the file contents, leave the version
// numbers in the file untouched.
fw->store_32(ver_major);
fw->store_32(ver_minor);
fw->store_32(ver_format);
save_ustring(fw, get_ustring(f)); //type
uint64_t md_ofs = f->get_position();
uint64_t importmd_ofs = f->get_64();
fw->store_64(0); //metadata offset
uint32_t flags = f->get_32();
bool using_uids = (flags & ResourceFormatSaverBinaryInstance::FORMAT_FLAG_UIDS);
uint64_t uid_data = f->get_64();
fw->store_32(flags);
fw->store_64(uid_data);
if (flags & ResourceFormatSaverBinaryInstance::FORMAT_FLAG_HAS_SCRIPT_CLASS) {
save_ustring(fw, get_ustring(f));
}
2021-07-29 22:24:29 -07:00
for (int i = 0; i < ResourceFormatSaverBinaryInstance::RESERVED_FIELDS; i++) {
2021-07-29 18:17:49 -07:00
fw->store_32(0); // reserved
f->get_32();
}
//string table
uint32_t string_table_size = f->get_32();
fw->store_32(string_table_size);
for (uint32_t i = 0; i < string_table_size; i++) {
String s = get_ustring(f);
save_ustring(fw, s);
}
//external resources
uint32_t ext_resources_size = f->get_32();
fw->store_32(ext_resources_size);
for (uint32_t i = 0; i < ext_resources_size; i++) {
String type = get_ustring(f);
String path = get_ustring(f);
if (using_uids) {
ResourceUID::ID uid = f->get_64();
if (uid != ResourceUID::INVALID_ID) {
if (ResourceUID::get_singleton()->has_id(uid)) {
// If a UID is found and the path is valid, it will be used, otherwise, it falls back to the path.
path = ResourceUID::get_singleton()->get_id_path(uid);
}
}
}
bool relative = false;
if (!path.begins_with("res://")) {
path = local_path.path_join(path).simplify_path();
relative = true;
}
if (p_map.has(path)) {
String np = p_map[path];
path = np;
}
String full_path = path;
if (relative) {
//restore relative
path = local_path.path_to_file(path);
}
save_ustring(fw, type);
save_ustring(fw, path);
if (using_uids) {
ResourceUID::ID uid = ResourceSaver::get_resource_id_for_path(full_path);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
fw->store_64(uint64_t(uid));
}
}
int64_t size_diff = (int64_t)fw->get_position() - (int64_t)f->get_position();
//internal resources
uint32_t int_resources_size = f->get_32();
fw->store_32(int_resources_size);
for (uint32_t i = 0; i < int_resources_size; i++) {
String path = get_ustring(f);
uint64_t offset = f->get_64();
save_ustring(fw, path);
fw->store_64(offset + size_diff);
}
//rest of file
uint8_t b = f->get_8();
while (!f->eof_reached()) {
fw->store_8(b);
b = f->get_8();
}
2022-05-02 16:37:50 +03:00
f.unref();
bool all_ok = fw->get_error() == OK;
fw->seek(md_ofs);
fw->store_64(importmd_ofs + size_diff);
if (!all_ok) {
return ERR_CANT_CREATE;
}
fw.unref();
Ref<DirAccess> da = DirAccess::create(DirAccess::ACCESS_RESOURCES);
if (da->exists(p_path + ".depren")) {
da->remove(p_path);
da->rename(p_path + ".depren", p_path);
}
return OK;
2014-02-09 22:10:30 -03:00
}
void ResourceFormatLoaderBinary::get_classes_used(const String &p_path, HashSet<StringName> *r_classes) {
Ref<FileAccess> f = FileAccess::open(p_path, FileAccess::READ);
ERR_FAIL_COND_MSG(f.is_null(), vformat("Cannot open file '%s'.", p_path));
ResourceLoaderBinary loader;
loader.local_path = ProjectSettings::get_singleton()->localize_path(p_path);
loader.res_path = loader.local_path;
loader.get_classes_used(f, r_classes);
// Fetch the nodes inside scene files.
if (loader.type == "PackedScene") {
ERR_FAIL_COND(loader.load() != OK);
Ref<SceneState> state = Ref<PackedScene>(loader.get_resource())->get_state();
for (int i = 0; i < state->get_node_count(); i++) {
const StringName node_name = state->get_node_type(i);
if (ClassDB::class_exists(node_name)) {
r_classes->insert(node_name);
}
// Fetch the values of properties in the node.
for (int j = 0; j < state->get_node_property_count(i); j++) {
const Variant var = state->get_node_property_value(i, j);
if (var.get_type() != Variant::OBJECT) {
continue;
}
const Object *obj = var.get_validated_object();
if (obj == nullptr) {
continue;
}
const StringName obj_name = obj->get_class_name();
if (ClassDB::class_exists(obj_name)) {
r_classes->insert(obj_name);
}
}
}
}
}
2014-02-09 22:10:30 -03:00
String ResourceFormatLoaderBinary::get_resource_type(const String &p_path) const {
Ref<FileAccess> f = FileAccess::open(p_path, FileAccess::READ);
if (f.is_null()) {
return ""; //could not read
2014-02-09 22:10:30 -03:00
}
ResourceLoaderBinary loader;
loader.local_path = ProjectSettings::get_singleton()->localize_path(p_path);
loader.res_path = loader.local_path;
String r = loader.recognize(f);
2019-10-03 17:39:08 -03:00
return ClassDB::get_compatibility_remapped_class(r);
2014-02-09 22:10:30 -03:00
}
String ResourceFormatLoaderBinary::get_resource_script_class(const String &p_path) const {
Ref<FileAccess> f = FileAccess::open(p_path, FileAccess::READ);
if (f.is_null()) {
return ""; //could not read
}
ResourceLoaderBinary loader;
loader.local_path = ProjectSettings::get_singleton()->localize_path(p_path);
loader.res_path = loader.local_path;
return loader.recognize_script_class(f);
}
ResourceUID::ID ResourceFormatLoaderBinary::get_resource_uid(const String &p_path) const {
String ext = p_path.get_extension().to_lower();
if (!ClassDB::is_resource_extension(ext)) {
return ResourceUID::INVALID_ID;
}
Ref<FileAccess> f = FileAccess::open(p_path, FileAccess::READ);
if (f.is_null()) {
return ResourceUID::INVALID_ID; //could not read
}
ResourceLoaderBinary loader;
loader.local_path = ProjectSettings::get_singleton()->localize_path(p_path);
loader.res_path = loader.local_path;
loader.open(f, true);
if (loader.error != OK) {
return ResourceUID::INVALID_ID; //could not read
}
return loader.uid;
}
bool ResourceFormatLoaderBinary::has_custom_uid_support() const {
return true;
}
2014-02-09 22:10:30 -03:00
///////////////////////////////////////////////////////////
///////////////////////////////////////////////////////////
///////////////////////////////////////////////////////////
void ResourceFormatSaverBinaryInstance::_pad_buffer(Ref<FileAccess> f, int p_bytes) {
2014-02-09 22:10:30 -03:00
int extra = 4 - (p_bytes % 4);
if (extra < 4) {
for (int i = 0; i < extra; i++) {
2014-02-09 22:10:30 -03:00
f->store_8(0); //pad to 32
}
2014-02-09 22:10:30 -03:00
}
}
void ResourceFormatSaverBinaryInstance::write_variant(Ref<FileAccess> f, const Variant &p_property, HashMap<Ref<Resource>, int> &resource_map, HashMap<Ref<Resource>, int> &external_resources, HashMap<StringName, int> &string_map, const PropertyInfo &p_hint) {
2014-02-09 22:10:30 -03:00
switch (p_property.get_type()) {
case Variant::NIL: {
f->store_32(VARIANT_NIL);
// don't store anything
} break;
case Variant::BOOL: {
f->store_32(VARIANT_BOOL);
bool val = p_property;
f->store_32(val);
} break;
case Variant::INT: {
int64_t val = p_property;
if (val > 0x7FFFFFFF || val < -(int64_t)0x80000000) {
f->store_32(VARIANT_INT64);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_64(uint64_t(val));
} else {
f->store_32(VARIANT_INT);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(p_property));
}
2014-02-09 22:10:30 -03:00
} break;
case Variant::FLOAT: {
double d = p_property;
float fl = d;
if (double(fl) != d) {
f->store_32(VARIANT_DOUBLE);
f->store_double(d);
} else {
f->store_32(VARIANT_FLOAT);
f->store_real(fl);
}
2014-02-09 22:10:30 -03:00
} break;
case Variant::STRING: {
f->store_32(VARIANT_STRING);
String val = p_property;
save_unicode_string(f, val);
2014-02-09 22:10:30 -03:00
} break;
case Variant::VECTOR2: {
f->store_32(VARIANT_VECTOR2);
Vector2 val = p_property;
f->store_real(val.x);
f->store_real(val.y);
} break;
case Variant::VECTOR2I: {
f->store_32(VARIANT_VECTOR2I);
Vector2i val = p_property;
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(val.x));
f->store_32(uint32_t(val.y));
2014-02-09 22:10:30 -03:00
} break;
case Variant::RECT2: {
f->store_32(VARIANT_RECT2);
Rect2 val = p_property;
f->store_real(val.position.x);
f->store_real(val.position.y);
2014-02-09 22:10:30 -03:00
f->store_real(val.size.x);
f->store_real(val.size.y);
} break;
case Variant::RECT2I: {
f->store_32(VARIANT_RECT2I);
Rect2i val = p_property;
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(val.position.x));
f->store_32(uint32_t(val.position.y));
f->store_32(uint32_t(val.size.x));
f->store_32(uint32_t(val.size.y));
2014-02-09 22:10:30 -03:00
} break;
case Variant::VECTOR3: {
f->store_32(VARIANT_VECTOR3);
Vector3 val = p_property;
f->store_real(val.x);
f->store_real(val.y);
f->store_real(val.z);
} break;
case Variant::VECTOR3I: {
f->store_32(VARIANT_VECTOR3I);
Vector3i val = p_property;
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(val.x));
f->store_32(uint32_t(val.y));
f->store_32(uint32_t(val.z));
} break;
case Variant::VECTOR4: {
f->store_32(VARIANT_VECTOR4);
Vector4 val = p_property;
f->store_real(val.x);
f->store_real(val.y);
f->store_real(val.z);
f->store_real(val.w);
} break;
case Variant::VECTOR4I: {
f->store_32(VARIANT_VECTOR4I);
Vector4i val = p_property;
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(val.x));
f->store_32(uint32_t(val.y));
f->store_32(uint32_t(val.z));
f->store_32(uint32_t(val.w));
2014-02-09 22:10:30 -03:00
} break;
case Variant::PLANE: {
f->store_32(VARIANT_PLANE);
Plane val = p_property;
f->store_real(val.normal.x);
f->store_real(val.normal.y);
f->store_real(val.normal.z);
f->store_real(val.d);
2014-02-09 22:10:30 -03:00
} break;
2021-01-20 07:02:02 +00:00
case Variant::QUATERNION: {
f->store_32(VARIANT_QUATERNION);
Quaternion val = p_property;
2014-02-09 22:10:30 -03:00
f->store_real(val.x);
f->store_real(val.y);
f->store_real(val.z);
f->store_real(val.w);
} break;
2017-11-16 21:09:00 -05:00
case Variant::AABB: {
f->store_32(VARIANT_AABB);
AABB val = p_property;
f->store_real(val.position.x);
f->store_real(val.position.y);
f->store_real(val.position.z);
2014-02-09 22:10:30 -03:00
f->store_real(val.size.x);
f->store_real(val.size.y);
f->store_real(val.size.z);
} break;
case Variant::TRANSFORM2D: {
f->store_32(VARIANT_TRANSFORM2D);
Transform2D val = p_property;
f->store_real(val.columns[0].x);
f->store_real(val.columns[0].y);
f->store_real(val.columns[1].x);
f->store_real(val.columns[1].y);
f->store_real(val.columns[2].x);
f->store_real(val.columns[2].y);
2014-02-09 22:10:30 -03:00
} break;
case Variant::BASIS: {
f->store_32(VARIANT_BASIS);
Basis val = p_property;
2022-04-24 17:07:35 -05:00
f->store_real(val.rows[0].x);
f->store_real(val.rows[0].y);
f->store_real(val.rows[0].z);
f->store_real(val.rows[1].x);
f->store_real(val.rows[1].y);
f->store_real(val.rows[1].z);
f->store_real(val.rows[2].x);
f->store_real(val.rows[2].y);
f->store_real(val.rows[2].z);
2014-02-09 22:10:30 -03:00
} break;
case Variant::TRANSFORM3D: {
f->store_32(VARIANT_TRANSFORM3D);
Transform3D val = p_property;
2022-04-24 17:07:35 -05:00
f->store_real(val.basis.rows[0].x);
f->store_real(val.basis.rows[0].y);
f->store_real(val.basis.rows[0].z);
f->store_real(val.basis.rows[1].x);
f->store_real(val.basis.rows[1].y);
f->store_real(val.basis.rows[1].z);
f->store_real(val.basis.rows[2].x);
f->store_real(val.basis.rows[2].y);
f->store_real(val.basis.rows[2].z);
2014-02-09 22:10:30 -03:00
f->store_real(val.origin.x);
f->store_real(val.origin.y);
f->store_real(val.origin.z);
} break;
case Variant::PROJECTION: {
f->store_32(VARIANT_PROJECTION);
Projection val = p_property;
f->store_real(val.columns[0].x);
f->store_real(val.columns[0].y);
f->store_real(val.columns[0].z);
f->store_real(val.columns[0].w);
f->store_real(val.columns[1].x);
f->store_real(val.columns[1].y);
f->store_real(val.columns[1].z);
f->store_real(val.columns[1].w);
f->store_real(val.columns[2].x);
f->store_real(val.columns[2].y);
f->store_real(val.columns[2].z);
f->store_real(val.columns[2].w);
f->store_real(val.columns[3].x);
f->store_real(val.columns[3].y);
f->store_real(val.columns[3].z);
f->store_real(val.columns[3].w);
2014-02-09 22:10:30 -03:00
} break;
case Variant::COLOR: {
f->store_32(VARIANT_COLOR);
Color val = p_property;
// Color are always floats
f->store_float(val.r);
f->store_float(val.g);
f->store_float(val.b);
f->store_float(val.a);
2014-02-09 22:10:30 -03:00
} break;
case Variant::STRING_NAME: {
f->store_32(VARIANT_STRING_NAME);
String val = p_property;
save_unicode_string(f, val);
2014-02-09 22:10:30 -03:00
} break;
case Variant::NODE_PATH: {
f->store_32(VARIANT_NODE_PATH);
NodePath np = p_property;
f->store_16(np.get_name_count());
uint16_t snc = np.get_subname_count();
if (np.is_absolute()) {
2014-02-09 22:10:30 -03:00
snc |= 0x8000;
}
2014-02-09 22:10:30 -03:00
f->store_16(snc);
for (int i = 0; i < np.get_name_count(); i++) {
if (string_map.has(np.get_name(i))) {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(string_map[np.get_name(i)]));
} else {
save_unicode_string(f, np.get_name(i), true);
}
}
for (int i = 0; i < np.get_subname_count(); i++) {
if (string_map.has(np.get_subname(i))) {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(string_map[np.get_subname(i)]));
} else {
save_unicode_string(f, np.get_subname(i), true);
}
}
2014-02-09 22:10:30 -03:00
} break;
case Variant::RID: {
2014-02-09 22:10:30 -03:00
f->store_32(VARIANT_RID);
WARN_PRINT("Can't save RIDs.");
2014-02-09 22:10:30 -03:00
RID val = p_property;
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(val.get_id()));
2014-02-09 22:10:30 -03:00
} break;
case Variant::OBJECT: {
f->store_32(VARIANT_OBJECT);
Ref<Resource> res = p_property;
if (res.is_null() || res->get_meta(SNAME("_skip_save_"), false)) {
2014-02-09 22:10:30 -03:00
f->store_32(OBJECT_EMPTY);
return; // Don't save it.
2014-02-09 22:10:30 -03:00
}
2021-07-10 21:17:41 +02:00
if (!res->is_built_in()) {
f->store_32(OBJECT_EXTERNAL_RESOURCE_INDEX);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(external_resources[res]));
2014-02-09 22:10:30 -03:00
} else {
if (!resource_map.has(res)) {
2014-02-09 22:10:30 -03:00
f->store_32(OBJECT_EMPTY);
ERR_FAIL_MSG("Resource was not pre cached for the resource section, most likely due to circular reference.");
2014-02-09 22:10:30 -03:00
}
f->store_32(OBJECT_INTERNAL_RESOURCE);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(resource_map[res]));
2014-02-09 22:10:30 -03:00
//internal resource
}
} break;
case Variant::CALLABLE: {
f->store_32(VARIANT_CALLABLE);
WARN_PRINT("Can't save Callables.");
} break;
case Variant::SIGNAL: {
f->store_32(VARIANT_SIGNAL);
WARN_PRINT("Can't save Signals.");
} break;
2014-02-09 22:10:30 -03:00
case Variant::DICTIONARY: {
f->store_32(VARIANT_DICTIONARY);
Dictionary d = p_property;
f->store_32(uint32_t(d.size()));
2014-02-09 22:10:30 -03:00
2025-03-12 06:43:48 +08:00
for (const KeyValue<Variant, Variant> &kv : d) {
write_variant(f, kv.key, resource_map, external_resources, string_map);
write_variant(f, kv.value, resource_map, external_resources, string_map);
2014-02-09 22:10:30 -03:00
}
} break;
case Variant::ARRAY: {
f->store_32(VARIANT_ARRAY);
Array a = p_property;
f->store_32(uint32_t(a.size()));
for (const Variant &var : a) {
write_variant(f, var, resource_map, external_resources, string_map);
2014-02-09 22:10:30 -03:00
}
} break;
case Variant::PACKED_BYTE_ARRAY: {
f->store_32(VARIANT_PACKED_BYTE_ARRAY);
Vector<uint8_t> arr = p_property;
2014-02-09 22:10:30 -03:00
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const uint8_t *r = arr.ptr();
f->store_buffer(r, len);
_pad_buffer(f, len);
2014-02-09 22:10:30 -03:00
} break;
case Variant::PACKED_INT32_ARRAY: {
f->store_32(VARIANT_PACKED_INT32_ARRAY);
Vector<int32_t> arr = p_property;
2014-02-09 22:10:30 -03:00
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const int32_t *r = arr.ptr();
for (int i = 0; i < len; i++) {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(r[i]));
}
2014-02-09 22:10:30 -03:00
} break;
case Variant::PACKED_INT64_ARRAY: {
f->store_32(VARIANT_PACKED_INT64_ARRAY);
Vector<int64_t> arr = p_property;
2014-02-09 22:10:30 -03:00
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const int64_t *r = arr.ptr();
for (int i = 0; i < len; i++) {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_64(uint64_t(r[i]));
}
} break;
case Variant::PACKED_FLOAT32_ARRAY: {
f->store_32(VARIANT_PACKED_FLOAT32_ARRAY);
Vector<float> arr = p_property;
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const float *r = arr.ptr();
2014-02-09 22:10:30 -03:00
for (int i = 0; i < len; i++) {
f->store_float(r[i]);
2014-02-09 22:10:30 -03:00
}
} break;
case Variant::PACKED_FLOAT64_ARRAY: {
f->store_32(VARIANT_PACKED_FLOAT64_ARRAY);
Vector<double> arr = p_property;
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const double *r = arr.ptr();
for (int i = 0; i < len; i++) {
f->store_double(r[i]);
}
2014-02-09 22:10:30 -03:00
} break;
case Variant::PACKED_STRING_ARRAY: {
f->store_32(VARIANT_PACKED_STRING_ARRAY);
Vector<String> arr = p_property;
2014-02-09 22:10:30 -03:00
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const String *r = arr.ptr();
2014-02-09 22:10:30 -03:00
for (int i = 0; i < len; i++) {
save_unicode_string(f, r[i]);
2014-02-09 22:10:30 -03:00
}
} break;
case Variant::PACKED_VECTOR2_ARRAY: {
f->store_32(VARIANT_PACKED_VECTOR2_ARRAY);
Vector<Vector2> arr = p_property;
2014-02-09 22:10:30 -03:00
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const Vector2 *r = arr.ptr();
2014-02-09 22:10:30 -03:00
for (int i = 0; i < len; i++) {
f->store_real(r[i].x);
f->store_real(r[i].y);
}
} break;
case Variant::PACKED_VECTOR3_ARRAY: {
f->store_32(VARIANT_PACKED_VECTOR3_ARRAY);
Vector<Vector3> arr = p_property;
2014-02-09 22:10:30 -03:00
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const Vector3 *r = arr.ptr();
2014-02-09 22:10:30 -03:00
for (int i = 0; i < len; i++) {
f->store_real(r[i].x);
f->store_real(r[i].y);
f->store_real(r[i].z);
2014-02-09 22:10:30 -03:00
}
} break;
case Variant::PACKED_COLOR_ARRAY: {
f->store_32(VARIANT_PACKED_COLOR_ARRAY);
Vector<Color> arr = p_property;
2014-02-09 22:10:30 -03:00
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const Color *r = arr.ptr();
2014-02-09 22:10:30 -03:00
for (int i = 0; i < len; i++) {
f->store_float(r[i].r);
f->store_float(r[i].g);
f->store_float(r[i].b);
f->store_float(r[i].a);
2014-02-09 22:10:30 -03:00
}
} break;
case Variant::PACKED_VECTOR4_ARRAY: {
f->store_32(VARIANT_PACKED_VECTOR4_ARRAY);
Vector<Vector4> arr = p_property;
int len = arr.size();
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(len));
const Vector4 *r = arr.ptr();
for (int i = 0; i < len; i++) {
f->store_real(r[i].x);
f->store_real(r[i].y);
f->store_real(r[i].z);
f->store_real(r[i].w);
}
2014-02-09 22:10:30 -03:00
} break;
default: {
ERR_FAIL_MSG("Invalid variant.");
2014-02-09 22:10:30 -03:00
}
}
}
void ResourceFormatSaverBinaryInstance::_find_resources(const Variant &p_variant, bool p_main) {
switch (p_variant.get_type()) {
case Variant::OBJECT: {
Ref<Resource> res = p_variant;
2014-02-09 22:10:30 -03:00
if (res.is_null() || external_resources.has(res) || res->get_meta(SNAME("_skip_save_"), false)) {
2014-02-09 22:10:30 -03:00
return;
}
2014-02-09 22:10:30 -03:00
2021-07-10 21:17:41 +02:00
if (!p_main && (!bundle_resources) && !res->is_built_in()) {
if (res->get_path() == path) {
ERR_PRINT(vformat("Circular reference to resource being saved found: '%s' will be null next time it's loaded.", local_path));
return;
}
int idx = external_resources.size();
external_resources[res] = idx;
2014-02-09 22:10:30 -03:00
return;
}
if (resource_set.has(res)) {
2014-02-09 22:10:30 -03:00
return;
}
2014-02-09 22:10:30 -03:00
resource_set.insert(res);
2014-02-09 22:10:30 -03:00
List<PropertyInfo> property_list;
res->get_property_list(&property_list);
for (const PropertyInfo &E : property_list) {
if (E.usage & PROPERTY_USAGE_STORAGE) {
Variant value = res->get(E.name);
if (E.usage & PROPERTY_USAGE_RESOURCE_NOT_PERSISTENT) {
NonPersistentKey npk;
npk.base = res;
npk.property = E.name;
non_persistent_map[npk] = value;
Ref<Resource> sres = value;
if (sres.is_valid()) {
resource_set.insert(sres);
saved_resources.push_back(sres);
} else {
_find_resources(value);
}
} else {
_find_resources(value);
}
2014-02-09 22:10:30 -03:00
}
}
saved_resources.push_back(res);
} break;
case Variant::ARRAY: {
Array varray = p_variant;
_find_resources(varray.get_typed_script());
for (const Variant &v : varray) {
2014-02-09 22:10:30 -03:00
_find_resources(v);
}
} break;
case Variant::DICTIONARY: {
Dictionary d = p_variant;
2023-06-24 13:03:28 -05:00
_find_resources(d.get_typed_key_script());
_find_resources(d.get_typed_value_script());
2025-03-12 06:43:48 +08:00
for (const KeyValue<Variant, Variant> &kv : d) {
_find_resources(kv.key);
_find_resources(kv.value);
2014-02-09 22:10:30 -03:00
}
} break;
case Variant::NODE_PATH: {
//take the chance and save node path strings
NodePath np = p_variant;
for (int i = 0; i < np.get_name_count(); i++) {
2014-02-09 22:10:30 -03:00
get_string_index(np.get_name(i));
}
for (int i = 0; i < np.get_subname_count(); i++) {
2014-02-09 22:10:30 -03:00
get_string_index(np.get_subname(i));
}
2014-02-09 22:10:30 -03:00
} break;
default: {
}
2014-02-09 22:10:30 -03:00
}
}
void ResourceFormatSaverBinaryInstance::save_unicode_string(Ref<FileAccess> p_f, const String &p_string, bool p_bit_on_len) {
2014-02-09 22:10:30 -03:00
CharString utf8 = p_string.utf8();
if (p_bit_on_len) {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
p_f->store_32(uint32_t((utf8.length() + 1) | 0x80000000));
} else {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
p_f->store_32(uint32_t(utf8.length() + 1));
}
p_f->store_buffer((const uint8_t *)utf8.get_data(), utf8.length() + 1);
2014-02-09 22:10:30 -03:00
}
int ResourceFormatSaverBinaryInstance::get_string_index(const String &p_string) {
StringName s = p_string;
if (string_map.has(s)) {
2014-02-09 22:10:30 -03:00
return string_map[s];
}
2014-02-09 22:10:30 -03:00
string_map[s] = strings.size();
strings.push_back(s);
return strings.size() - 1;
}
static String _resource_get_class(Ref<Resource> p_resource) {
Ref<MissingResource> missing_resource = p_resource;
if (missing_resource.is_valid()) {
return missing_resource->get_original_class();
} else {
return p_resource->get_class();
}
}
Error ResourceFormatSaverBinaryInstance::save(const String &p_path, const Ref<Resource> &p_resource, uint32_t p_flags) {
Resource::seed_scene_unique_id(p_path.hash());
2014-02-09 22:10:30 -03:00
Error err;
Ref<FileAccess> f;
2014-02-09 22:10:30 -03:00
if (p_flags & ResourceSaver::FLAG_COMPRESS) {
Ref<FileAccessCompressed> fac;
fac.instantiate();
2014-02-09 22:10:30 -03:00
fac->configure("RSCC");
f = fac;
err = fac->open_internal(p_path, FileAccess::WRITE);
2014-02-09 22:10:30 -03:00
} else {
f = FileAccess::open(p_path, FileAccess::WRITE, &err);
}
ERR_FAIL_COND_V_MSG(err != OK, err, vformat("Cannot create file '%s'.", p_path));
2014-02-09 22:10:30 -03:00
relative_paths = p_flags & ResourceSaver::FLAG_RELATIVE_PATHS;
skip_editor = p_flags & ResourceSaver::FLAG_OMIT_EDITOR_PROPERTIES;
bundle_resources = p_flags & ResourceSaver::FLAG_BUNDLE_RESOURCES;
big_endian = p_flags & ResourceSaver::FLAG_SAVE_BIG_ENDIAN;
takeover_paths = p_flags & ResourceSaver::FLAG_REPLACE_SUBRESOURCE_PATHS;
if (!p_path.begins_with("res://")) {
takeover_paths = false;
}
2014-02-09 22:10:30 -03:00
local_path = p_path.get_base_dir();
path = ProjectSettings::get_singleton()->localize_path(p_path);
2014-02-09 22:10:30 -03:00
_find_resources(p_resource, true);
if (!(p_flags & ResourceSaver::FLAG_COMPRESS)) {
//save header compressed
static const uint8_t header[4] = { 'R', 'S', 'R', 'C' };
f->store_buffer(header, 4);
}
if (big_endian) {
f->store_32(1);
} else {
2014-02-09 22:10:30 -03:00
f->store_32(0);
}
2014-02-09 22:10:30 -03:00
f->store_32(0); //64 bits file, false for now
f->set_big_endian(big_endian);
f->store_32(GODOT_VERSION_MAJOR);
f->store_32(GODOT_VERSION_MINOR);
2014-02-09 22:10:30 -03:00
f->store_32(FORMAT_VERSION);
if (f->get_error() != OK && f->get_error() != ERR_FILE_EOF) {
return ERR_CANT_CREATE;
}
save_unicode_string(f, _resource_get_class(p_resource));
f->store_64(0); //offset to import metadata
String script_class;
{
uint32_t format_flags = FORMAT_FLAG_NAMED_SCENE_IDS | FORMAT_FLAG_UIDS;
#ifdef REAL_T_IS_DOUBLE
format_flags |= FORMAT_FLAG_REAL_T_IS_DOUBLE;
#endif
if (!p_resource->is_class("PackedScene")) {
Ref<Script> s = p_resource->get_script();
if (s.is_valid()) {
script_class = s->get_global_name();
if (!script_class.is_empty()) {
format_flags |= ResourceFormatSaverBinaryInstance::FORMAT_FLAG_HAS_SCRIPT_CLASS;
}
}
}
f->store_32(format_flags);
}
ResourceUID::ID uid = ResourceSaver::get_resource_id_for_path(p_path, true);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_64(uint64_t(uid));
if (!script_class.is_empty()) {
save_unicode_string(f, script_class);
}
2021-07-29 22:24:29 -07:00
for (int i = 0; i < ResourceFormatSaverBinaryInstance::RESERVED_FIELDS; i++) {
2014-02-09 22:10:30 -03:00
f->store_32(0); // reserved
}
2014-02-09 22:10:30 -03:00
List<ResourceData> resources;
{
for (const Ref<Resource> &E : saved_resources) {
Dictionary missing_resource_properties = E->get_meta(META_MISSING_RESOURCES, Dictionary());
2014-02-09 22:10:30 -03:00
ResourceData &rd = resources.push_back(ResourceData())->get();
rd.type = _resource_get_class(E);
2014-02-09 22:10:30 -03:00
List<PropertyInfo> property_list;
E->get_property_list(&property_list);
2014-02-09 22:10:30 -03:00
for (const PropertyInfo &F : property_list) {
if (skip_editor && F.name.begins_with("__editor")) {
2014-02-09 22:10:30 -03:00
continue;
}
if (F.name == META_PROPERTY_MISSING_RESOURCES) {
continue;
}
if ((F.usage & PROPERTY_USAGE_STORAGE) || missing_resource_properties.has(F.name)) {
2014-02-09 22:10:30 -03:00
Property p;
p.name_idx = get_string_index(F.name);
if (F.usage & PROPERTY_USAGE_RESOURCE_NOT_PERSISTENT) {
NonPersistentKey npk;
npk.base = E;
npk.property = F.name;
if (non_persistent_map.has(npk)) {
p.value = non_persistent_map[npk];
}
} else {
p.value = E->get(F.name);
}
if (F.type == Variant::OBJECT && missing_resource_properties.has(F.name)) {
// Was this missing resource overridden? If so do not save the old value.
Ref<Resource> res = p.value;
if (res.is_null()) {
p.value = missing_resource_properties[F.name];
}
}
bool is_script = F.name == CoreStringName(script);
Variant default_value = is_script ? Variant() : PropertyUtils::get_property_default_value(E.ptr(), F.name);
if (default_value.get_type() != Variant::NIL && bool(Variant::evaluate(Variant::OP_EQUAL, p.value, default_value))) {
2014-02-09 22:10:30 -03:00
continue;
}
p.pi = F;
2014-02-09 22:10:30 -03:00
rd.properties.push_back(p);
}
}
}
}
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(strings.size())); //string table size
2014-02-09 22:10:30 -03:00
for (int i = 0; i < strings.size(); i++) {
save_unicode_string(f, strings[i]);
2014-02-09 22:10:30 -03:00
}
// save external resource table
f->store_32(external_resources.size()); //amount of external resources
Vector<Ref<Resource>> save_order;
save_order.resize(external_resources.size());
for (const KeyValue<Ref<Resource>, int> &E : external_resources) {
2021-08-09 14:13:42 -06:00
save_order.write[E.value] = E.key;
}
2014-02-09 22:10:30 -03:00
for (int i = 0; i < save_order.size(); i++) {
save_unicode_string(f, save_order[i]->get_save_class());
String res_path = save_order[i]->get_path();
res_path = relative_paths ? local_path.path_to_file(res_path) : res_path;
save_unicode_string(f, res_path);
ResourceUID::ID ruid = ResourceSaver::get_resource_id_for_path(save_order[i]->get_path(), false);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_64(uint64_t(ruid));
2014-02-09 22:10:30 -03:00
}
// save internal resource table
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(saved_resources.size())); //amount of internal resources
2014-02-09 22:10:30 -03:00
Vector<uint64_t> ofs_pos;
HashSet<String> used_unique_ids;
for (Ref<Resource> &r : saved_resources) {
2021-07-10 21:17:41 +02:00
if (r->is_built_in()) {
if (!r->get_scene_unique_id().is_empty()) {
if (used_unique_ids.has(r->get_scene_unique_id())) {
r->set_scene_unique_id("");
} else {
used_unique_ids.insert(r->get_scene_unique_id());
}
}
}
}
HashMap<Ref<Resource>, int> resource_map;
int res_index = 0;
for (Ref<Resource> &r : saved_resources) {
2021-07-10 21:17:41 +02:00
if (r->is_built_in()) {
if (r->get_scene_unique_id().is_empty()) {
String new_id;
while (true) {
new_id = _resource_get_class(r) + "_" + Resource::generate_scene_unique_id();
if (!used_unique_ids.has(new_id)) {
break;
}
}
r->set_scene_unique_id(new_id);
used_unique_ids.insert(new_id);
}
save_unicode_string(f, "local://" + r->get_scene_unique_id());
if (takeover_paths) {
r->set_path(p_path + "::" + r->get_scene_unique_id(), true);
}
#ifdef TOOLS_ENABLED
r->set_edited(false);
#endif
} else {
save_unicode_string(f, r->get_path()); //actual external
}
ofs_pos.push_back(f->get_position());
2014-02-09 22:10:30 -03:00
f->store_64(0); //offset in 64 bits
resource_map[r] = res_index++;
2014-02-09 22:10:30 -03:00
}
Vector<uint64_t> ofs_table;
//now actually save the resources
for (const ResourceData &rd : resources) {
ofs_table.push_back(f->get_position());
save_unicode_string(f, rd.type);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(rd.properties.size()));
2014-02-09 22:10:30 -03:00
for (const Property &p : rd.properties) {
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
f->store_32(uint32_t(p.name_idx));
write_variant(f, p.value, resource_map, external_resources, string_map, p.pi);
2014-02-09 22:10:30 -03:00
}
}
for (int i = 0; i < ofs_table.size(); i++) {
f->seek(ofs_pos[i]);
f->store_64(ofs_table[i]);
}
f->seek_end();
f->store_buffer((const uint8_t *)"RSRC", 4); //magic at end
if (f->get_error() != OK && f->get_error() != ERR_FILE_EOF) {
return ERR_CANT_CREATE;
}
2014-02-09 22:10:30 -03:00
return OK;
}
Error ResourceFormatSaverBinaryInstance::set_uid(const String &p_path, ResourceUID::ID p_uid) {
Ref<FileAccess> f = FileAccess::open(p_path, FileAccess::READ);
ERR_FAIL_COND_V_MSG(f.is_null(), ERR_CANT_OPEN, vformat("Cannot open file '%s'.", p_path));
Ref<FileAccess> fw;
local_path = p_path.get_base_dir();
uint8_t header[4];
f->get_buffer(header, 4);
if (header[0] == 'R' && header[1] == 'S' && header[2] == 'C' && header[3] == 'C') {
// Compressed.
Ref<FileAccessCompressed> fac;
fac.instantiate();
Error err = fac->open_after_magic(f);
ERR_FAIL_COND_V_MSG(err != OK, err, vformat("Cannot open file '%s'.", p_path));
f = fac;
Ref<FileAccessCompressed> facw;
facw.instantiate();
facw->configure("RSCC");
err = facw->open_internal(p_path + ".uidren", FileAccess::WRITE);
ERR_FAIL_COND_V_MSG(err, ERR_FILE_CORRUPT, vformat("Cannot create file '%s.uidren'.", p_path));
fw = facw;
} else if (header[0] != 'R' || header[1] != 'S' || header[2] != 'R' || header[3] != 'C') {
// Not a binary resource.
return ERR_FILE_UNRECOGNIZED;
} else {
fw = FileAccess::open(p_path + ".uidren", FileAccess::WRITE);
ERR_FAIL_COND_V_MSG(fw.is_null(), ERR_CANT_CREATE, vformat("Cannot create file '%s.uidren'.", p_path));
uint8_t magich[4] = { 'R', 'S', 'R', 'C' };
fw->store_buffer(magich, 4);
}
big_endian = f->get_32();
bool use_real64 = f->get_32();
f->set_big_endian(big_endian != 0); //read big endian if saved as big endian
fw->store_32(big_endian);
fw->store_32(use_real64); //use real64
fw->set_big_endian(big_endian != 0);
uint32_t ver_major = f->get_32();
uint32_t ver_minor = f->get_32();
uint32_t ver_format = f->get_32();
if (ver_format < FORMAT_VERSION_CAN_RENAME_DEPS) {
fw.unref();
{
Ref<DirAccess> da = DirAccess::create(DirAccess::ACCESS_FILESYSTEM);
da->remove(p_path + ".uidren");
}
// Use the old approach.
WARN_PRINT(vformat("This file is old, so it does not support UIDs, opening and resaving '%s'.", p_path));
return ERR_UNAVAILABLE;
}
if (ver_format > FORMAT_VERSION || ver_major > GODOT_VERSION_MAJOR) {
ERR_FAIL_V_MSG(ERR_FILE_UNRECOGNIZED,
vformat("File '%s' can't be loaded, as it uses a format version (%d) or engine version (%d.%d) which are not supported by your engine version (%s).",
local_path, ver_format, ver_major, ver_minor, GODOT_VERSION_BRANCH));
}
// Since we're not actually converting the file contents, leave the version
// numbers in the file untouched.
fw->store_32(ver_major);
fw->store_32(ver_minor);
fw->store_32(ver_format);
save_ustring(fw, get_ustring(f)); //type
fw->store_64(f->get_64()); //metadata offset
uint32_t flags = f->get_32();
flags |= ResourceFormatSaverBinaryInstance::FORMAT_FLAG_UIDS;
f->get_64(); // Skip previous UID
fw->store_32(flags);
Core ubsan fixes This fixes UBSAN errors reported by running our testsuite, importing the TPS demo, and running the TPS demo. I have tried, wherever possible, to fix issues related to reported issues but not directly reported by UBSAN because thse code paths just happened to not have been exercised in these cases. These fixes apply only to errors reported, and caused by, core/ The following things have been changed: * Make sure there are no implicit sign changing casts in core. * Explicitly type enums that are part of a public API such that users of the API cannot pass in wrongly-sized values leading to potential stack corruption. * Ensure that memcpy is never called with invalid or null pointers as this is undefined behavior, and when the engine is built with optimizations turned on leads to memory corruption and hard to debug crashes. * Replace enum values only used as static values with constexpr static const values instead. This has no runtime overhead. This makes it so that the size of the enums is explicit. * Make sure that nan and inf is handled consistently in String. * Implement a _to_int template to ensure that all of the paths use the same algorhithm, and correct the negative integer case. * Changed the way the json serializer precision work, and added tests to verify the new behavior. The behavior doesn't quite match master in particulary for negative doubles as the original code tried to cast -inf to an int. This then led to negative doubles losing all but one of their decimal points when serializing. Behavior in GDScript remains unchanged.
2024-12-14 02:17:09 +01:00
fw->store_64(uint64_t(p_uid));
if (flags & ResourceFormatSaverBinaryInstance::FORMAT_FLAG_HAS_SCRIPT_CLASS) {
save_ustring(fw, get_ustring(f));
}
//rest of file
uint8_t b = f->get_8();
while (!f->eof_reached()) {
fw->store_8(b);
b = f->get_8();
}
f.unref();
bool all_ok = fw->get_error() == OK;
if (!all_ok) {
return ERR_CANT_CREATE;
}
fw.unref();
Ref<DirAccess> da = DirAccess::create(DirAccess::ACCESS_RESOURCES);
da->remove(p_path);
da->rename(p_path + ".uidren", p_path);
return OK;
}
2022-06-03 01:33:42 +02:00
Error ResourceFormatSaverBinary::save(const Ref<Resource> &p_resource, const String &p_path, uint32_t p_flags) {
String local_path = ProjectSettings::get_singleton()->localize_path(p_path);
2014-02-09 22:10:30 -03:00
ResourceFormatSaverBinaryInstance saver;
return saver.save(local_path, p_resource, p_flags);
}
Error ResourceFormatSaverBinary::set_uid(const String &p_path, ResourceUID::ID p_uid) {
String local_path = ProjectSettings::get_singleton()->localize_path(p_path);
ResourceFormatSaverBinaryInstance saver;
return saver.set_uid(local_path, p_uid);
}
bool ResourceFormatSaverBinary::recognize(const Ref<Resource> &p_resource) const {
2014-02-09 22:10:30 -03:00
return true; //all recognized
}
void ResourceFormatSaverBinary::get_recognized_extensions(const Ref<Resource> &p_resource, List<String> *p_extensions) const {
2014-02-09 22:10:30 -03:00
String base = p_resource->get_base_extension().to_lower();
2016-06-28 18:58:40 +02:00
p_extensions->push_back(base);
if (base != "res") {
p_extensions->push_back("res");
}
2014-02-09 22:10:30 -03:00
}
ResourceFormatSaverBinary::ResourceFormatSaverBinary() {
singleton = this;
}