Files were not being truncated on Windows anymore when using `io_open` with `IOFLAG_WRITE` due to a regression from #7254. Instead, the existing file contents were kept and the file pointer was set to the beginning of the file.
This caused broken demo files to be created (#7349) when recording a shorter demo with the same filename as an existing longer demo. It also caused the map tools to produce maps with additional junk data at the end, if an existing map is overridden by a smaller map.
This is fixed by using the creation disposition `CREATE_ALWAYS` instead of `OPEN_EXISTING` with [`CreateFileW`](https://learn.microsoft.com/en-us/windows/win32/api/fileapi/nf-fileapi-createfilew), which ensures that a file is always created and truncated.
A regression test case is added, which fails without this change.
We support little and big endian but not PDP endian (Middle-endian).
Define endianness as string `CONF_ARCH_ENDIAN_STRING` to avoid conditional compilation when printing endianness.
On Windows the function `setsockopt` does not set `errno` on errors but instead `WSAGetLastError` must be used. This is encapsulated in the `net_errno` function.
Additionally, ensure snapshots are cleared and dummy is disconnected when disconnecting programatically with `DisconnectWithReason`.
Add `NETADDR_ZEROED` and `UUID_ZEROED` for more convenient initialization without using `mem_zero`.
Add stricter error handling when converting between UTF-16 (wide characters) and UTF-8 (multi-byte) on Windows.
The `windows_wide_to_utf8` function now returns an `std::optional`, which will be empty if the argument contains invalid UTF-16. Files/folders with names containing invalid UTF-16 are now ignored on Windows. It was previously not possible to use these files either, as converting their names to UTF-8 changed the invalid codepoints to unicode replacement characters.
The `windows_utf8_to_wide` function now fails with an assertion error if the argument contains invalid UTF-8, as this should never happen.
Closes#7486.
Initialize nontrivial types with a constructor instead. Make the
compiler aware that some of our constructors are indeed trivial.
This allows `mem_zero` calls to actually always zero the memory.
Partially replaces #5690.
POD types are just memset. Other types are either destructed if not
trivial and/or constructed if not trivial. Types need to have a default
constructor.
Virtual classes can be mem_zeroed only if they already have been
constructed, otherwise it is UB.
On Unix, the encoding of filenames is unspecified, although UTF-8 is likely used in most cases. Detecting and converting from several possible encodings to UTF-8 would be too much effort, considering that most systems already use UTF-8 per default. Therefore, any filenames which are not valid UTF-8 will be ignored when listing directories with `fs_listdir(_fileinfo)`, `fs_storage_path` will fail if the storage location is not valid UTF-8 and `fs_getcwd` will fail if the current working directory is not valid UTF-8. Paths specified in `storage.cfg` must also be valid UTF-8.
On Windows, we already ensure that all filenames are converted to UTF-8.
Use `CLock` and `CLockScope` instead of `std::mutex` and add clang thread-safety analysis annotations everywhere except for usages in engine graphics and video, as those usages also involve `std::condition_variable`.
Fix lock not being unlocked on all code paths in `CFutureLogger::Log`, which is caught by the static analysis.
Replace usages of platform specific `lock_*` functions with `std::mutex` through the wrapper class `CLock`. Move lock classes to `base/lock.h`.
The `CLock` wrapper class is only necessary because the clang thread-safety attributes are not available for `std::mutex` except when explicitly using libc++.
Round to nearest integer instead of truncating in `f2fx` to ensure correct round-trip with `fx2f`.
Add test to ensure correct round-trip with maximum `0.0005f` absolute error.
After the error message popup is shown for graphics assertions, the client window was destroyed but the process was not terminated properly. Now the graphics assertion error is handled like a normal assertion error, as those are also shown in an error message popup and correctly cause the client to break into the debugger or terminate.
However, calling `dbg_assert` while already owning the lock that `WaitForIdle` waits on will cause a deadlock, so error handling must be delayed until after the lock is released.
The buffer size for assertion messages is increased, as it was not sufficient for some graphics assertions.
The `ICommandProcessor::GetError` and `ICommandProcessor::GetWarning` functions are marked as `const`.
Add easter, halloween and xmas to `enum ETimeSeason` and `time_season` function. This unifies the logic for detecting these events and seasons in one function. This allows adding support for special easter, halloween and xmas theme maps in the future. The `localtime` function is not used outside `system.c` anymore. The code for detecting easter is taken from upstream.
The function returns 0 on success and -1 on error, as it delegates to `io_seek`, instead of returning the number of bytes skipped, which the documentation previously suggested. The return type is changed from `unsigned` to `int`, as this is also the type of `io_seek` and this better handles -1 as a return value.
The TODO in the `Dilate` function is removed, as the code already appears to be safe without additional checks. The variable `k` is at most `(w * h - 1) * BPP`, as `ix` and `iy` are clamped to maximum `w - 1` and `h - 1` respectively. Because `p < BPP - 1` the index `k + p` is therefore always valid for the buffers. (The caller must ensure that the source and destination buffers are of size `w * h * BPP`.)
This mostly reverts #6937 by making our `IOHANDLE` effectively `FILE *` on all systems again. We still use `CreateFileW` to open a `HANDLE` initially so we can specify the necessary flag so the file can be moved/deleted while open, which we can't do with the `FILE *` based `fopen` functions on Windows. This brings back the automatic I/O buffering on Windows, causing significantly less system calls when saving files. Closes#7226.
Add more efficient function for formatting integer values as strings.
A benchmark shows that using this function is significantly faster than using `str_format`. It is faster by a factor of 220 with Clang 15.0 O2 (https://quick-bench.com/q/BlNoLnlyqxipf4jvsFTUxKMHDJU) and by a factor of 11 with GCC 12.2 O2 (https://quick-bench.com/q/Fxf9lDCTqXBF4pIa_IyZ5R0IqYg).
This increases FPS in the editor by ~25% when many numbers are rendered for switch/tele/speedup/tune layers or with "Show Info" being enabled.
The additional static analysis for `std::to_chars` revealed that the wrong size was used in `CHud` for `aScoreTeam[TEAM_RED]` and `aScoreTeam[TEAM_BLUE]`.
This requires incrementing the macOS deployment target from 10.13 to 10.15.
Add `str_utf8_offset_bytes_to_chars` and `str_utf8_offset_chars_to_bytes` functions to base system to convert between byte and UTF-8 character offsets in UTF-8 strings.
Previously, this was separately implemented in the textrender and in the lineinput helper.
These textrender functions are entirely replaced by the new functions:
- `ITextRender::SelectionToUTF8OffSets` (by `str_utf8_offset_chars_to_bytes`)
- `ITextRender::UTF8OffToDecodedOff` (by `str_utf8_offset_bytes_to_chars`)
- `ITextRender::DecodedOffToUTF8Off` (by `str_utf8_offset_chars_to_bytes`)
These lineinput helper functions are reimplemented using the new functions:
- `CLineInput::OffsetFromActualToDisplay` (uses `str_utf8_offset_bytes_to_chars`)
- `CLineInput::OffsetFromDisplayToActual` (uses `str_utf8_offset_chars_to_bytes`)
This still works
$ ./DDNet "connect 127.0.0.1"
But now also this works
$ ./DDNet "connect tw-0.6+udp://ger10.ddnet.org:8303"
2023-07-16 14:07:50 I engine: running on unix-linux-amd64
2023-07-16 14:07:50 I client: starting...
2023-07-16 14:07:50 I client: version 17.1.1 on linux amd64
2023-07-16 14:07:50 I client: git revision hash: 7f100e2620
2023-07-16 14:07:50 I client: connecting to 'tw-0.6+udp://ger10.ddnet.org:8303'
2023-07-16 14:07:50 I host_lookup: host='ger10.ddnet.org' port=8303 3
2023-07-16 14:07:51 I client: connected, sending info
Switch to using the Windows-specific `HANDLE` and Windows API functions instead of `FILE *` and standard library functions in all Windows IO function implementations.
Using `CreateFileW` instead of `_wfsopen` to open files is necessary, as only `CreateFileW` allows specifying all necessary sharing flags. The sharing flags `FILE_SHARE_READ` and `FILE_SHARE_WRITE` were previously already specified by using `_SH_DENYNO` with `_wfsopen`, but `FILE_SHARE_DELETE` can only be set when using `CreateFileW`. The flags are necessary so files which are in use by the game can still be opened by other processes. In particular, `FILE_SHARE_DELETE` allows deleting/renaming of open files, which was previously not possible. This was causing the smart editor saving process that renames files to fail if a map file is currently in use.
Because `CreateFileW` returns a `HANDLE` instead of a `FILE *` we have to use the `HANDLE`-based Windows IO functions everywhere.
See: https://devblogs.microsoft.com/oldnewthing/20211022-00/?p=105822Closes#6922.
`stdout_output_level` for printing to stdout, `console_output_level` for
printing to local console and remote console and `loglevel` for the log
file.
Keep the old log level filters 0 for info and more severe, 1 for debug
and more severe and 2 for trace and more severe, introducing -1 for
warn, and -2 for error. -3 will show no log messages at all.
Port map and editor support for `CURVETYPE_BEZIER` from upstream, i.e. support bezier curves with configurable in- and out-tangents for every envelope point.
The in- and out-tangents are represented by triangles and can be dragged in the envelope editor like the envelope points.
Support reading and writing the bezier information as a separate UUID-based map item. If the bezier information is not found, bezier will default to linear behavior. Old clients will still be able to read the new maps and ignore the unknown map item. The unknown curvetype will also be handled as linear by old clients.
Allow reading upstream maps that use `CMapItemEnvelope` version 3. On upstream, a different struct is used to store all envelope points including bezier information, which broke compatibility to old clients.
Fix holding Ctrl for slow envelope point editing not working for vertical movement.
Highlight the currently selected element (envelope point or bezier tangent marker) which is being used with the value/time edit boxes.
Hide the value/time edit boxes when no element is selected.
Add optional `end` output parameter to `str_utf8_find_nocase` which is set to a pointer into the haystack after the matched string.
This is necessary because there are pairs of matching upper case and lower case Unicode characters with different byte length (e.g. both `I` and `İ` map to `i`), so the byte length of the string matched in the haystack may not be identical to the length of the needle string.
6755: Remove individual RGBA sliders from editor color picker, add shift+rightclick/leftclick to copy/paste color in editor r=Jupeyy a=Robyt3
Remove the individual RGBA sliders for editor color pickers and only show one button that opens the color picker popup instead.
Decrease size of layer and point popups that previously had color properties which need less space now.
Support shift-rightclicking color picker buttons to copy the color to the clipboard in RRGGBBAA hex format.
Support shift-leftclicking color picker buttons to paste a color from the clipboard in RGB, RGBA, RRGGBB or RRGGBBAA format with optional leading `#` or `$`.
Screenshots:
- Before:
- After:
- After (color picker open):
Suggested by `@HiRavie` in https://github.com/ddnet/ddnet/pull/6743#issuecomment-1593886873, though it was easier and also looks better to me when the color picker button has exactly the same size as the other value selectors.
## Checklist
- [X] Tested the change ingame
- [X] Provided screenshots if it is a visual change
- [ ] Tested in combination with possibly related configuration options
- [ ] Written a unit test (especially base/) or added coverage to integration test
- [ ] Considered possible null pointers and out of bounds array indexing
- [ ] Changed no physics that affect existing maps
- [ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addresssanitizer--undefinedbehavioursanitizer-or-valgrinds-memcheck) (optional)
Co-authored-by: Robert Müller <robytemueller@gmail.com>
6429: Minor refactoring of `os_version_str` r=edg-l a=Robyt3
## Checklist
- [X] Tested the change ingame
- [ ] Provided screenshots if it is a visual change
- [ ] Tested in combination with possibly related configuration options
- [ ] Written a unit test (especially base/) or added coverage to integration test
- [ ] Considered possible null pointers and out of bounds array indexing
- [ ] Changed no physics that affect existing maps
- [ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addresssanitizer--undefinedbehavioursanitizer-or-valgrinds-memcheck) (optional)
Co-authored-by: Robert Müller <robytemueller@gmail.com>
To convert between UTF-8 encoded strings and wide character strings used with the Windows API.
The functions use `std::wstring` and `std::string`, to simplify the usage and support arguments with arbitrary length.
Also add some tests for the functions.
Port the line input (UI edit boxes, chat, console) and Input Method Editor (IME) support from upstream. Closes#4397.
General
------------------------------
Fix issues with the text input. Closes#4346. Closes#4524.
Word skipping (when holding Ctrl) is overhauled to be consistent with the Windows / Firefox experience that I took as reference.
Improve usability by not blinking (i.e. always rendering) the caret shortly after is has been moved.
UI text input
------------------------------
Fix inconsistent mouse-based left and right scrolling (closes#4347).
Support smooth left and right scrolling.
Chat
------------------------------
Support keyboard-based text selection of the chat input.
Mouse-based selection could be support in the future when we decide to add something like an ingame UI cursor.
Support smooth up and down scrolling of the chat input, removing the old hack that offsets the input string to simulate scrolling.
Console
------------------------------
Also support mouse-based text selection of the command input.
Only text from either the command input or the console log can be selected at the same time. This ensures that Ctrl+C will always copy the text that is currently visually selected in the console.
Check for Ctrl+C input event in event handler instead of in render function, to hopefully fix the issue that copying does not work sometimes (closes#5974 until further notice).
When Ctrl+C is used to copy text from the console log, the selection is cleared. This should make it more clear when text was copied from the log.
Fix an issue that was preventing the console log selection from being cleared, when all log lines are selected.
Remove Ctrl+A/E hotkeys that move cursor to beginning/end respectively. Ctrl+A now selectes all text like for all other inputs. Home and End keys can still be used to go the beginning and end.
Remove Ctrl+U/K hotkeys that clear everything before/after the cursor respectively. Hold shift and use Home/End to select everything instead.
IME support
------------------------------
Render list of IME candidates in the client on Windows, so the candidate list can also be viewed in fullscreen mode. There is no API available to retrieve a candidate list on the other operating systems.
Improve composition rendering by underlining the composition text instead of putting it in square brackets.
Track active input globally to properly activate and deactivate IME through the SDL functions.
Closes#1030. Closes#1008.
Password rendering
------------------------------
Fix rendering of passwords containing unicode. Instead of rendering one star character for each UTF-8 `char`, render on star for every unicode codepoint.
Show the composition text also for passwords. Without seeing the composition text it's hard to type a password containing those characters. The candidate window exposes the composition anyway. If you don't want to expose your password this way, e.g. while streaming, you could:
1. Use a latin password and switch off the IME for the password input with the IME hotkey.
2. Blank your screen with an external program while you are streaming and entering passwords.
3. Use binds to authenticate in rcon or to set the server browser password.
Refactoring
------------------------------
Move all text input logic and general rendering to `CLineInput`.
A `CLineInput` is associated with a particular `char` buffer given as a pointer either in the constructor or with `SetBuffer`. The maximum byte size of the buffer must also be specified. The maximum length in unicode codepoints can also be specified separately (e.g. on upstream, name are limited by the number of unicode codepoints instead).
Add `CLineInputBuffered`, which is a `CLineInput` that own a `char` buffer of a fixed size, which is specified as a template argument. As `CLineInput` does not own a buffer anymore, this reduces duplicate code for line inputs that need their own buffer.
Add `CLineInputNumber` which has additional convenience functions to consider the text as an `int` or `float`, to reduce duplicate code in those cases. In the future we could also add an input filter function so that only numbers can be entered in the number input.
Add `CLineInput::SetClipboardLineCallback` to handle the case that multiple lines of text are pasted into a lineinput. This reduces duplicate code, as this behavior was previously implemented separately for chat and console. The behavior is also fixed to be consistent with the console on Windows, so the first line being pasted edits the current input text and then sends it instead of being sent on its own without the existing input text.
Add `CalcFontSizeAndBoundingBox` to UI to reduce duplicate code. Expose `CalcAlignedCursorPos` as static member function to reuse it for line input.
Dispatch input events to UI inputs through the event handler instead of storing them in a duplicate buffer.
Use `size_t` for line input cursor position, length etc. and for `str_utf8_stats`.
Add `IButtonColorFunction` to UI to describe a functions that defines colors for the Default, Active and Hovered states of UI elements. Add some default button color functions. Use button color function to reduce duplicate code in scrollbar rendering.
Use `vec2` instead of two `floats` to represent the mouse positions in the text renderer.
Remove `CaretPosition` again, as it does not calculate the correct Y position near line breaks due to the wrapping being different when not rendering the entire string. Instead, calculate the exact caret position when rending a text container and store the caret position in the text cursor for later use.
IME usage guide (Windows)
------------------------------
1. Install the respective language and the Microsoft-IME keyboard (e.g. for Chinese, Japanese or Korean).
2. Launch the game (or a text editor to first try out the IME). Note that Windows may track the input language separately for every application. You can change this in the Windows input settings so the input language is changed globally.
2. Switch the input language using the hotkey Windows+Space or another hotkey that you configured in the Windows input settings (Alt+Shift is the default, but you should consider disabling it, to avoid accidentally changing the input language while playing).
3. Switch from Latin/English input mode to the respective asian input mode.
- Chinese: Use Ctrl+Space to switch between English and Chinese input mode. You can change this hotkey in the IME's settings.
- Japanese: Use Ctrl+Space to switch between Alphanumeric and Hiragana/Katakana input mode. You can change this hotkey in the IME's settings.
- Korean: Use Right Alt to switch between English and Hangul input mode. You cannot change this hotkey as of yet.
- Note that the input mode is also tracked per application, but there is no setting to change this behavior as far as I know, so you'll need to switch for every application separately.
4. Start typing. The underlined text is the current composition text. While a composition is active, you can only edit the composition text. Confirm the composition with Space or by selecting a candidate from the candidate list with the arrow keys. Cancel the composition with Escape or by using Backspace to delete the composition text. Note that not all languages offer a candidate list.
SDL version-specific issues
------------------------------
- 2.26.5, 2.24.2, 2.0.22: IME candidates work. But there are minor bugs when moving the composition cursor.
- 2.0.18, 2.0.20: IME candidates work.
- 2.0.16 (our current version): IME candidates cannot be determined with Windows API. Windows tries to draw the composition window like before, so this does not work in fullscreen mode.
- 2.0.8 (upstream 0.7): IME candidates work. But this SDL version is too old for us.
6460: Replace usage of `fmodf` with `std::fmod` r=heinrich5991 a=Robyt3
Leftover from #6372.
## Checklist
- [ ] Tested the change ingame
- [ ] Provided screenshots if it is a visual change
- [ ] Tested in combination with possibly related configuration options
- [ ] Written a unit test (especially base/) or added coverage to integration test
- [ ] Considered possible null pointers and out of bounds array indexing
- [ ] Changed no physics that affect existing maps
- [ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addresssanitizer--undefinedbehavioursanitizer-or-valgrinds-memcheck) (optional)
Co-authored-by: Robert Müller <robytemueller@gmail.com>
6439: Refactor `random` functions r=heinrich5991 a=Robyt3
## Checklist
- [ ] Tested the change ingame
- [ ] Provided screenshots if it is a visual change
- [ ] Tested in combination with possibly related configuration options
- [ ] Written a unit test (especially base/) or added coverage to integration test
- [ ] Considered possible null pointers and out of bounds array indexing
- [ ] Changed no physics that affect existing maps
- [ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addresssanitizer--undefinedbehavioursanitizer-or-valgrinds-memcheck) (optional)
Co-authored-by: Robert Müller <robytemueller@gmail.com>
Reduces duplicate code. This also ensures that the log messages are properly zero-terminated and that they do not end with truncated unicode characters.
So the string formatting can be reused in other places. Passing variadic arguments directly to another function is not possible.
The function must have a different name, as overloading `str_format` would cause the wrong function to be called when passing the `va_list`. The same naming with suffix `_v` is used for the logging functions.
Add `random_float(float min, float max)` to generate a random `float` between `min` and `max`.
Add `random_float(float max)` to generate a random `float` between `0.0f` and `max`.
Replace existing `RandomDir` with this function. This function was biased towards the corners, as it first generates a random `vec2` with `x` and `y` in `[-0.5f;0.5f]` and then normalizes it, which doesn't result in a uniform distribution of directions. Now the random direction is generated by taking a random angle and then converting it to a direction which is per definition already normalized.
To generate a random angle in the range `[0.0f;2.0f * pi[`. This ensures that the random angle cannot be `2.0f * pi`, which would be identical to the angle `0.0f` and therefore cause the random angle to be less uniformly selected.
Note that this first casts `RAND_MAX` to a `float` and then uses `std::nextafter` to get the next larger `float`. Using `RAND_MAX + 1` would cause an integer overflow on systems where `RAND_MAX == INT_MAX` (e.g. Ubuntu and macOS).
6324: Optimize client ids map update r=def- a=0xfaulty
I have launched support for 128 players in testing mode on my server, it's means it's time for another step in this direction here. Perhaps not all changes will be clear, I am ready to answer/discuss them as usual. Let me know if I forgot or didn't take anything into account.
## Checklist
- [x] Tested the change ingame
- [ ] Provided screenshots if it is a visual change
- [ ] Tested in combination with possibly related configuration options
- [ ] Written a unit test (especially base/) or added coverage to integration test
- [ ] Considered possible null pointers and out of bounds array indexing
- [ ] Changed no physics that affect existing maps
- [ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addresssanitizer--undefinedbehavioursanitizer-or-valgrinds-memcheck) (optional)
6345: Set constant and uniform names for the server and client CMake targets r=def- a=Kaffeine
This MR set the server and client CMake target names to `game-server` and `game-client` which are uniform with other targets such as `engine-gfx`, `engine-shared`, and `game-shared`.
Set uniform names for the targets to:
1. Avoid long rebuilds on the target _executable_ name changed
2. Provide the same name for different executables and/or different setup (use-cases are `cmake --build . --target game-server` and various CMake target lookups in IDEs)
Those changes are necessary but not enough to allow CI for custom app name on all platforms.
Personally this 'll allow me switch to between DDNet and Infclass branches faster and without breaking the targets configuration (e.g. with that patchset I can have `game-client` as an active target in both branches), and also reduces the conflicts in `CMakeLists.txt`.
Easier to switch between the forks => easier to contribute.
Co-authored-by: Valentin Bashkirov <valenteen3d@ya.ru>
Co-authored-by: Valentin Bashkirov <v.bashkirov@dev.tassta.com>
Co-authored-by: Alexander Akulich <akulichalexander@gmail.com>
Replacing the C standard headers with the C++ standard headers causes various `error: call to 'floor' promotes float to double [performance-type-promotion-in-math-fn,-warnings-as-errors]`, which are fixed by using the C++ std math functions or our own math functions instead of the C math functions.
- Use `absolute` instead of `abs` and `fabs`.
- Use `std::floor` instead of `floor` and `floorf`.
- Use `std::ceil` instead of `ceil`, `ceilf` and `round_ceil`.
- Use `std::round` instead of `round` and `roundf`.
- Use `std::sin` instead of `sin` and `sinf`.
- Use `std::asin` instead of `asin` and `asinf`.
- Use `std::cos` instead of `cos` and `cosf`.
- Use `std::acos` instead of `acos` and `acosf`.
- Use `std::tan` instead of `tan` and `tanf`.
- Use `std::atan` instead of `atan` and `atanf`.
- Use `std::pow` instead of `pow` and `powf`.
- Use `std::log` instead of `log` and `logf`.
- Use `std::log2` instead of `log2` and `log2f`.
- Use `std::log10` instead of `log10` and `log10f`.
- Use `std::pow` instead of `pow` and `powf`.
- Use `std::sqrt` instead of `sqrt` and `sqrtf`.
- Use `std::fmod` instead of `fmod` and `fmodf`.
- Use `direction(Angle)` instead of `vec2(std::cos(Angle), std::sin(Angle))`.
- Use `length(vec2(x, y))` instead of `std::sqrt(x * x + y * y)`.
- Remove unused `NormalizeAngular` and `AngularDistance` functions.
Instead of returning the number of bytes written, which are platform specific, return `true` on success and `false` on failure, so no platform specific code is required when checking the result.
ChillerDragon
