mirrors/Atmosphere
1
0
Fork 0
mirror of https://github.com/Atmosphere-NX/Atmosphere.git synced 2024-11-18 18:06:40 +00:00
Code Issues Releases Activity
Atmosphere/libraries/libstratosphere/include/stratosphere/fs/fs_path.hpp

676 lines
26 KiB
C++
Raw Normal View History

ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/*
* Copyright (c) Atmosphère-NX
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#pragma once
#include <stratosphere/fs/fs_common.hpp>
#include <stratosphere/fs/fs_memory_management.hpp>
#include <stratosphere/fs/impl/fs_common_mount_name.hpp>
#include <stratosphere/fs/fs_path_utility.hpp>
namespace ams::fs {
class DirectoryPathParser;
fs: for my sanity, begin tracking version where code was last checked for accuracy
2022-03-24 15:43:40 +00:00
/* ACCURATE_TO_VERSION: 13.4.0.0 */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
/* NOTE: Intentional inaccuracy in custom WriteBuffer class, to save 0x10 bytes (0x28 -> 0x18) over Nintendo's implementation. */
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
class Path {
NON_COPYABLE(Path);
NON_MOVEABLE(Path);
private:
static constexpr const char *EmptyPath = "";
static constexpr size_t WriteBufferAlignmentLength = 8;
private:
friend class DirectoryPathParser;
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
public:
class WriteBuffer {
NON_COPYABLE(WriteBuffer);
private:
char *m_buffer;
size_t m_length_and_is_normalized;
public:
constexpr WriteBuffer() : m_buffer(nullptr), m_length_and_is_normalized(0) { /* ... */ }
constexpr ~WriteBuffer() {
if (m_buffer != nullptr) {
::ams::fs::impl::Deallocate(m_buffer, this->GetLength());
this->ResetBuffer();
}
}
constexpr WriteBuffer(WriteBuffer &&rhs) : m_buffer(rhs.m_buffer), m_length_and_is_normalized(rhs.m_length_and_is_normalized) {
rhs.ResetBuffer();
}
constexpr WriteBuffer &operator=(WriteBuffer &&rhs) {
fs: fix memory leak when path is reallocated (closes #1842)
2022-04-08 19:02:17 +01:00
if (m_buffer != nullptr) {
::ams::fs::impl::Deallocate(m_buffer, this->GetLength());
}
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
m_buffer = rhs.m_buffer;
m_length_and_is_normalized = rhs.m_length_and_is_normalized;
rhs.ResetBuffer();
return *this;
}
std::unique_ptr<char[], ::ams::fs::impl::Deleter> ReleaseBuffer() {
auto released = std::unique_ptr<char[], ::ams::fs::impl::Deleter>(m_buffer, ::ams::fs::impl::Deleter(this->GetLength()));
this->ResetBuffer();
return released;
}
constexpr ALWAYS_INLINE void ResetBuffer() {
m_buffer = nullptr;
this->SetLength(0);
}
constexpr ALWAYS_INLINE char *Get() const {
return m_buffer;
}
constexpr ALWAYS_INLINE size_t GetLength() const {
return m_length_and_is_normalized >> 1;
}
constexpr ALWAYS_INLINE bool IsNormalized() const {
return static_cast<bool>(m_length_and_is_normalized & 1);
}
constexpr ALWAYS_INLINE void SetNormalized() {
m_length_and_is_normalized |= static_cast<size_t>(1);
}
constexpr ALWAYS_INLINE void SetNotNormalized() {
m_length_and_is_normalized &= ~static_cast<size_t>(1);
}
private:
constexpr ALWAYS_INLINE WriteBuffer(char *buffer, size_t length) : m_buffer(buffer), m_length_and_is_normalized(0) {
this->SetLength(length);
}
public:
static WriteBuffer Make(size_t length) {
if (void *alloc = ::ams::fs::impl::Allocate(length); alloc != nullptr) {
return WriteBuffer(static_cast<char *>(alloc), length);
} else {
return WriteBuffer();
}
}
private:
constexpr ALWAYS_INLINE void SetLength(size_t size) {
m_length_and_is_normalized = (m_length_and_is_normalized & 1) | (size << 1);
}
};
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
private:
const char *m_str;
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
WriteBuffer m_write_buffer;
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
public:
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
constexpr Path() : m_str(EmptyPath), m_write_buffer() {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* ... */
}
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
constexpr Path(const char *s, util::ConstantInitializeTag) : m_str(s), m_write_buffer() {
m_write_buffer.SetNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
constexpr ~Path() { /* ... */ }
std::unique_ptr<char[], ::ams::fs::impl::Deleter> ReleaseBuffer() {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Check pre-conditions. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
AMS_ASSERT(m_write_buffer.Get() != nullptr);
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Reset. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
m_str = EmptyPath;
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
/* Release our write buffer. */
return m_write_buffer.ReleaseBuffer();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
constexpr Result SetShallowBuffer(const char *buffer) {
/* Check pre-conditions. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
AMS_ASSERT(m_write_buffer.GetLength() == 0);
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Check the buffer is valid. */
R_UNLESS(buffer != nullptr, fs::ResultNullptrArgument());
/* Set buffer. */
this->SetReadOnlyBuffer(buffer);
/* Note that we're normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
fs: update LocalFileSystem
2022-03-28 09:33:40 +01:00
constexpr const char *GetString() const {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Check pre-conditions. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
AMS_ASSERT(this->IsNormalized());
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
return m_str;
}
fs: update LocalFileSystem
2022-03-28 09:33:40 +01:00
constexpr size_t GetLength() const {
if (std::is_constant_evaluated()) {
return util::Strlen(this->GetString());
} else {
return std::strlen(this->GetString());
}
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
fs: update LocalFileSystem
2022-03-28 09:33:40 +01:00
constexpr bool IsEmpty() const {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
return *m_str == '\x00';
}
fs: update LocalFileSystem
2022-03-28 09:33:40 +01:00
constexpr bool IsMatchHead(const char *p, size_t len) const {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
return util::Strncmp(this->GetString(), p, len) == 0;
}
Result Initialize(const Path &rhs) {
/* Check the other path is normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
const bool normalized = rhs.IsNormalized();
R_UNLESS(normalized, fs::ResultNotNormalized());
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Allocate buffer for our path. */
const auto len = rhs.GetLength();
R_TRY(this->Preallocate(len + 1));
/* Copy the path. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
const size_t copied = util::Strlcpy<char>(m_write_buffer.Get(), rhs.GetString(), len + 1);
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_UNLESS(copied == len, fs::ResultUnexpectedInPathA());
/* Set normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
Result Initialize(const char *path, size_t len) {
/* Check the path is valid. */
R_UNLESS(path != nullptr, fs::ResultNullptrArgument());
/* Initialize. */
R_TRY(this->InitializeImpl(path, len));
/* Set not normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNotNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
Result Initialize(const char *path) {
/* Check the path is valid. */
R_UNLESS(path != nullptr, fs::ResultNullptrArgument());
ams: globally prefer R_RETURN to return for ams::Result
2022-03-26 21:48:33 +00:00
R_RETURN(this->Initialize(path, std::strlen(path)));
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
Result InitializeWithFormat(const char *fmt, ...) __attribute__((format (printf, 2, 3))) {
/* Check the format string is valid. */
R_UNLESS(fmt != nullptr, fs::ResultNullptrArgument());
/* Create the va_list for formatting. */
std::va_list vl;
va_start(vl, fmt);
/* Determine how big the string will be. */
char dummy;
const auto len = util::VSNPrintf(std::addressof(dummy), 0, fmt, vl);
/* Allocate buffer for our path. */
R_TRY(this->Preallocate(len + 1));
/* Format our path into our new buffer. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
const auto real_len = util::VSNPrintf(m_write_buffer.Get(), m_write_buffer.GetLength(), fmt, vl);
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
AMS_ASSERT(real_len == len);
AMS_UNUSED(real_len);
/* Finish the va_list. */
va_end(vl);
/* Set not normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNotNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
Result InitializeWithReplaceBackslash(const char *path) {
/* Check the path is valid. */
R_UNLESS(path != nullptr, fs::ResultNullptrArgument());
/* Initialize. */
R_TRY(this->InitializeImpl(path, std::strlen(path)));
/* Replace slashes as desired. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
if (const auto write_buffer_length = m_write_buffer.GetLength(); write_buffer_length > 1) {
fs::Replace(m_write_buffer.Get(), write_buffer_length - 1, '\\', '/');
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
/* Set not normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNotNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
Result InitializeWithReplaceForwardSlashes(const char *path) {
/* Check the path is valid. */
R_UNLESS(path != nullptr, fs::ResultNullptrArgument());
/* Initialize. */
R_TRY(this->InitializeImpl(path, std::strlen(path)));
/* Replace slashes as desired. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
if (m_write_buffer.GetLength() > 1) {
if (auto *p = m_write_buffer.Get(); p[0] == '/' && p[1] == '/') {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
p[0] = '\\';
p[1] = '\\';
}
}
/* Set not normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNotNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
Result InitializeWithReplaceUnc(const char *path) {
/* Check the path is valid. */
R_UNLESS(path != nullptr, fs::ResultNullptrArgument());
/* Initialize. */
R_TRY(this->InitializeImpl(path, std::strlen(path)));
/* Set not normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNotNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Replace unc as desired. */
if (m_str[0]) {
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
auto *p = m_write_buffer.Get();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Replace :/// -> \\ as needed. */
if (auto *sep = std::strstr(p, ":///"); sep != nullptr) {
sep[0] = '\\';
sep[1] = '\\';
}
/* Edit path prefix. */
if (!util::Strncmp(p, AMS_FS_IMPL_HOST_ROOT_FILE_SYSTEM_MOUNT_NAME AMS_FS_IMPL_MOUNT_NAME_DELIMITER "/", AMS_FS_IMPL_HOST_ROOT_FILE_SYSTEM_MOUNT_NAME_LEN + AMS_FS_IMPL_MOUNT_NAME_DELIMITER_LEN + 1)) {
static_assert((AMS_FS_IMPL_HOST_ROOT_FILE_SYSTEM_MOUNT_NAME AMS_FS_IMPL_MOUNT_NAME_DELIMITER)[AMS_FS_IMPL_HOST_ROOT_FILE_SYSTEM_MOUNT_NAME_LEN + AMS_FS_IMPL_MOUNT_NAME_DELIMITER_LEN - 1] == '/');
p[AMS_FS_IMPL_HOST_ROOT_FILE_SYSTEM_MOUNT_NAME_LEN + AMS_FS_IMPL_MOUNT_NAME_DELIMITER_LEN - 1] = '\\';
p[AMS_FS_IMPL_HOST_ROOT_FILE_SYSTEM_MOUNT_NAME_LEN + AMS_FS_IMPL_MOUNT_NAME_DELIMITER_LEN - 0] = '\\';
}
if (p[0] == '/' && p[1] == '/') {
p[0] = '\\';
p[1] = '\\';
}
}
R_SUCCEED();
}
Result InitializeWithNormalization(const char *path, size_t size) {
/* Check the path is valid. */
R_UNLESS(path != nullptr, fs::ResultNullptrArgument());
/* Initialize. */
R_TRY(this->InitializeImpl(path, size));
/* Set not normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNotNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Perform normalization. */
fs::PathFlags path_flags;
if (fs::IsPathRelative(m_str)) {
path_flags.AllowRelativePath();
} else if (fs::IsWindowsPath(m_str, true)) {
path_flags.AllowWindowsPath();
} else {
/* NOTE: In this case, Nintendo checks is normalized, then sets is normalized, then returns success. */
/* This seems like a bug. */
size_t dummy;
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
bool normalized;
R_TRY(PathFormatter::IsNormalized(std::addressof(normalized), std::addressof(dummy), m_str));
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
/* Normalize. */
R_TRY(this->Normalize(path_flags));
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
Result InitializeWithNormalization(const char *path) {
/* Check the path is valid. */
R_UNLESS(path != nullptr, fs::ResultNullptrArgument());
R_RETURN(this->InitializeWithNormalization(path, std::strlen(path)));
}
Result InitializeAsEmpty() {
/* Clear our buffer. */
this->ClearBuffer();
/* Set normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
Result AppendChild(const char *child) {
/* Check the path is valid. */
R_UNLESS(child != nullptr, fs::ResultNullptrArgument());
/* Basic checks. If we hvea a path and the child is empty, we have nothing to do. */
const char *c = child;
if (m_str[0]) {
/* Skip an early separator. */
if (*c == '/') {
++c;
}
R_SUCCEED_IF(*c == '\x00');
}
/* If we don't have a string, we can just initialize. */
auto cur_len = std::strlen(m_str);
if (cur_len == 0) {
R_RETURN(this->Initialize(child));
}
/* Remove a trailing separator. */
if (m_str[cur_len - 1] == '/' || m_str[cur_len - 1] == '\\') {
--cur_len;
}
/* Get the child path's length. */
auto child_len = std::strlen(c);
/* Reset our write buffer. */
WriteBuffer old_write_buffer;
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
if (m_write_buffer.Get() != nullptr) {
old_write_buffer = std::move(m_write_buffer);
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
this->ClearBuffer();
}
/* Pre-allocate the new buffer. */
R_TRY(this->Preallocate(cur_len + 1 + child_len + 1));
/* Get our write buffer. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
auto *dst = m_write_buffer.Get();
if (old_write_buffer.Get() != nullptr && cur_len > 0) {
util::Strlcpy<char>(dst, old_write_buffer.Get(), cur_len + 1);
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
/* Add separator. */
dst[cur_len] = '/';
/* Copy the child path. */
const size_t copied = util::Strlcpy<char>(dst + cur_len + 1, c, child_len + 1);
R_UNLESS(copied == child_len, fs::ResultUnexpectedInPathA());
R_SUCCEED();
}
Result AppendChild(const Path &rhs) {
R_RETURN(this->AppendChild(rhs.GetString()));
}
Result Combine(const Path &parent, const Path &child) {
/* Get the lengths. */
const auto p_len = parent.GetLength();
const auto c_len = child.GetLength();
/* Allocate our buffer. */
R_TRY(this->Preallocate(p_len + c_len + 1));
/* Initialize as parent. */
R_TRY(this->Initialize(parent));
/* If we're empty, we can just initialize as child. */
if (this->IsEmpty()) {
R_TRY(this->Initialize(child));
} else {
/* Otherwise, we should append the child. */
R_TRY(this->AppendChild(child));
}
R_SUCCEED();
}
Result RemoveChild() {
/* If we don't have a write-buffer, ensure that we have one. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
if (m_write_buffer.Get() == nullptr) {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
if (const auto len = std::strlen(m_str); len > 0) {
R_TRY(this->Preallocate(len));
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
util::Strlcpy<char>(m_write_buffer.Get(), m_str, len + 1);
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
}
/* Check that it's possible for us to remove a child. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
auto *p = m_write_buffer.Get();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
s32 len = std::strlen(p);
R_UNLESS(len != 1 || (p[0] != '/' && p[0] != '.'), fs::ResultNotImplemented());
/* Handle a trailing separator. */
if (len > 0 && (p[len - 1] == '\\' || p[len - 1] == '/')) {
--len;
}
/* Remove the child path segment. */
while ((--len) >= 0 && p[len]) {
if (p[len] == '/' || p[len] == '\\') {
if (len > 0) {
p[len] = 0;
} else {
p[1] = 0;
len = 1;
}
break;
}
}
/* Check that length remains > 0. */
R_UNLESS(len > 0, fs::ResultNotImplemented());
R_SUCCEED();
}
Result Normalize(const PathFlags &flags) {
/* If we're already normalized, nothing to do. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
R_SUCCEED_IF(this->IsNormalized());
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Check if we're normalized. */
bool normalized;
size_t dummy;
R_TRY(PathFormatter::IsNormalized(std::addressof(normalized), std::addressof(dummy), m_str, flags));
/* If we're not normalized, normalize. */
if (!normalized) {
/* Determine necessary buffer length. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
auto len = m_write_buffer.GetLength();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
if (flags.IsRelativePathAllowed() && fs::IsPathRelative(m_str)) {
len += 2;
}
if (flags.IsWindowsPathAllowed() && fs::IsWindowsPath(m_str, true)) {
len += 1;
}
/* Allocate a new buffer. */
const size_t size = util::AlignUp(len, WriteBufferAlignmentLength);
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
auto buf = WriteBuffer::Make(size);
R_UNLESS(buf.Get() != nullptr, fs::ResultAllocationMemoryFailedMakeUnique());
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Normalize into it. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
R_TRY(PathFormatter::Normalize(buf.Get(), size, m_write_buffer.Get(), m_write_buffer.GetLength(), flags));
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Set the normalized buffer as our buffer. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetModifiableBuffer(std::move(buf));
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
/* Set normalized. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetNormalized();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_SUCCEED();
}
private:
void ClearBuffer() {
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
m_write_buffer.ResetBuffer();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
m_str = EmptyPath;
}
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
void SetModifiableBuffer(WriteBuffer &&buffer) {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Check pre-conditions. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
AMS_ASSERT(buffer.Get() != nullptr);
AMS_ASSERT(buffer.GetLength() > 0);
AMS_ASSERT(util::IsAligned(buffer.GetLength(), WriteBufferAlignmentLength));
/* Get whether we're normalized. */
if (m_write_buffer.IsNormalized()) {
buffer.SetNormalized();
} else {
buffer.SetNotNormalized();
}
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Set write buffer. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
m_write_buffer = std::move(buffer);
m_str = m_write_buffer.Get();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
constexpr void SetReadOnlyBuffer(const char *buffer) {
m_str = buffer;
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
m_write_buffer.ResetBuffer();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
Result Preallocate(size_t length) {
/* Allocate additional space, if needed. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
if (length > m_write_buffer.GetLength()) {
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Allocate buffer. */
const size_t size = util::AlignUp(length, WriteBufferAlignmentLength);
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
auto buf = WriteBuffer::Make(size);
R_UNLESS(buf.Get() != nullptr, fs::ResultAllocationMemoryFailedMakeUnique());
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
/* Set write buffer. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
this->SetModifiableBuffer(std::move(buf));
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
R_SUCCEED();
}
Result InitializeImpl(const char *path, size_t size) {
if (size > 0 && path[0]) {
/* Pre allocate a buffer for the path. */
R_TRY(this->Preallocate(size + 1));
/* Copy the path. */
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
const size_t copied = util::Strlcpy<char>(m_write_buffer.Get(), path, size + 1);
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
R_UNLESS(copied >= size, fs::ResultUnexpectedInPathA());
} else {
/* We can just clear the buffer. */
this->ClearBuffer();
}
R_SUCCEED();
}
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
constexpr char *GetWriteBuffer() {
AMS_ASSERT(m_write_buffer.Get() != nullptr);
return m_write_buffer.Get();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
constexpr ALWAYS_INLINE size_t GetWriteBufferLength() const {
return m_write_buffer.GetLength();
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
fs: reduce path size 0x28 -> 0x18 This implements two optimizations on fs::Path, which N added in 12.0.0. The current structure looks like: ```cpp struct Path { const char *m_str; // Points to the read-only path string char *m_write_buffer_buffer; // Part of std::unique_ptr<char[], ams::fs::impl::Deleter> ams::fs::impl::Deleter m_write_buffer_deleter; // Parse of std::unique_ptr<char[], ams::fs::impl::Deleter>, stores the size of the buffer. size_t m_write_buffer_length; // Copy of the write buffer's size accessible to the Path() structure. bool m_is_normalized; // Whether the path buffer is normalized }; ``` This is pretty wasteful. The write buffer size is stored twice, wasting 8 bytes, because one copy of the size isn't accessible to the path. In addition, due to alignment, the bool wastes 7 padding bytes. This commit: * Encodes normalized in the low bit of the write buffer length, saving 8 bytes. * Use a custom WriteBuffer class rather than generic unique_ptr, to avoid needing to store the WriteBuffer twice. These each save 8 bytes, for a final size of 0x18 rather than 0x28.
2022-03-25 03:22:47 +00:00
constexpr ALWAYS_INLINE bool IsNormalized() const { return m_write_buffer.IsNormalized(); }
constexpr ALWAYS_INLINE void SetNormalized() { m_write_buffer.SetNormalized(); }
constexpr ALWAYS_INLINE void SetNotNormalized() { m_write_buffer.SetNotNormalized(); }
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
public:
ALWAYS_INLINE bool operator==(const fs::Path &rhs) const { return std::strcmp(this->GetString(), rhs.GetString()) == 0; }
ALWAYS_INLINE bool operator!=(const fs::Path &rhs) const { return !(*this == rhs); }
ALWAYS_INLINE bool operator==(const char *p) const { return std::strcmp(this->GetString(), p) == 0; }
ALWAYS_INLINE bool operator!=(const char *p) const { return !(*this == p); }
};
consteval fs::Path MakeConstantPath(const char *s) { return fs::Path(s, util::ConstantInitializeTag{}); }
inline Result SetUpFixedPath(fs::Path *out, const char *s) {
/* Verify the path is normalized. */
bool normalized;
size_t dummy;
R_TRY(PathNormalizer::IsNormalized(std::addressof(normalized), std::addressof(dummy), s));
R_UNLESS(normalized, fs::ResultInvalidPathFormat());
/* Set the fixed path. */
R_RETURN(out->SetShallowBuffer(s));
}
inline Result SetUpFixedPathSingleEntry(fs::Path *out, char *buf, size_t buf_size, const char *e) {
/* Print the path into the buffer. */
const size_t len = util::TSNPrintf(buf, buf_size, "/%s", e);
R_UNLESS(len < buf_size, fs::ResultInvalidArgument());
/* Set up the path. */
R_RETURN(SetUpFixedPath(out, buf));
}
inline Result SetUpFixedPathDoubleEntry(fs::Path *out, char *buf, size_t buf_size, const char *e, const char *e2) {
/* Print the path into the buffer. */
const size_t len = util::TSNPrintf(buf, buf_size, "/%s/%s", e, e2);
R_UNLESS(len < buf_size, fs::ResultInvalidArgument());
/* Set up the path. */
R_RETURN(SetUpFixedPath(out, buf));
}
inline Result SetUpFixedPathSaveId(fs::Path *out, char *buf, size_t buf_size, u64 id) {
/* Print the path into the buffer. */
const size_t len = util::TSNPrintf(buf, buf_size, "/%016" PRIx64 "", id);
R_UNLESS(len < buf_size, fs::ResultInvalidArgument());
/* Set up the path. */
R_RETURN(SetUpFixedPath(out, buf));
}
inline Result SetUpFixedPathSaveMetaName(fs::Path *out, char *buf, size_t buf_size, u32 type) {
/* Print the path into the buffer. */
const size_t len = util::TSNPrintf(buf, buf_size, "/%08" PRIx32 ".meta", type);
R_UNLESS(len < buf_size, fs::ResultInvalidArgument());
/* Set up the path. */
R_RETURN(SetUpFixedPath(out, buf));
}
inline Result SetUpFixedPathSaveMetaDir(fs::Path *out, char *buf, size_t buf_size, u64 id) {
/* Print the path into the buffer. */
const size_t len = util::TSNPrintf(buf, buf_size, "/saveMeta/%016" PRIx64 "", id);
R_UNLESS(len < buf_size, fs::ResultInvalidArgument());
/* Set up the path. */
R_RETURN(SetUpFixedPath(out, buf));
}
fs: update LocalFileSystem
2022-03-28 09:33:40 +01:00
constexpr inline bool IsWindowsDriveRootPath(const fs::Path &path) {
const char * const str = path.GetString();
return fs::IsWindowsDrive(str) && (str[2] == StringTraits::DirectorySeparator || str[2] == StringTraits::AlternateDirectorySeparator) && str[3] == StringTraits::NullTerminator;
}
ams: support building unit test programs on windows/linux/macos
2022-03-06 20:08:20 +00:00
}
Reference in a new issue Copy permalink