Fix edit boxes, scrollbars, value selectors and the editor sound seekbar being activated by holding the left mouse button from outside the UI element and then hovering it, which is inconsistent with the behavior of the `DoButtonLogic` function.
Use the `DoButtonLogic` function directly for the key reader buttons in the controls settings. Now, the key reader will only be activated when the left click is initiated and completed inside the key reader button. Previously, the key reader would be activated immediately when holding the left mouse button from outside and then hovering the key reader, and the key reader was already activated by the press-event, whereas now it will only be activated by the release event of the left mouse button. The active key reader UI ID is now tracked as a member variable of `CMenusKeyBinder` to avoid using static variables for this purpose.
The scrollregion scrollbar was not being deactivated if the handle is continued to be hovered after releasing it, causing the handle to become offset with the position of the click on the next activation of the handle.
The seekbar is immediately activated if the mouse button is held while moving the mouse over the seekbar, which is inconsistent with the general `DoButtonLogic` behavior. This is problematic on Android, where the first mouse button is always pressed while the mouse is being moved, causing the seekbar to steal focus immediately when hovered.
The tooltip was not shown if the seekbar is the active item, i.e. when the first mouse button is held down and the seekbar is currently updating the demo player.
Assume every server that does not have ddnet or race in its gametype to be a pvp server.
This allows spectators to follow multiple players during a fng/ctf/gctf/ictf round.
Related issue #7509
The `-1` in the size calculation for the ringbuffer allocation of chat history entries was incorrect, so the entries potentially didn't fit the zero terminator. Because `mem_copy` was used, the string was also not zero terminated explicitly.
This reverts commit f319ed239a.
Initializing these variables with junk data doesn't seem to be an
improvement over not initializing them. It hides potential Valgrind
warnings about data accesses to uninitialized memory though.
f319ed239a (r1590801501)
Nudge the initial position for explosions' smoke particles slightly towards the edge of the closest non-solid tile, if it would otherwise be inside a solid tile, so the smoke particles do not get stuck inside solid tiles on explosion events happening at the edges of tiles but slightly inside them. The physical position of the explosion event is unchanged, so this does not affect physics. The explosion sprite is still rendered at the physical position of the explosion, to preserve the apprearance.
Add additional checks to ensure that the `CVideo::Stop` function and the functions called by it will correctly stop the current video also if the video was not started successfully, i.e. if `CVideo::Start` returned `false` from any of the error branches.
In the `CVideo::Stop` function, iterate over the vectors of video and audio threads directly instead of using `m_VideoThreads` and `m_AudioThreads`, which do not reflect the actual count if the initialization failed before the threads were created.
In the `CVideo::Stop` function, only call `av_write_trailer` if the video recording was stated successfully, i.e. only if `avformat_write_header` was called successfully, as this will otherwise cause the client to crash. Closes#6375.
In the `CVideo::Stop` function, only call `avio_closep` if the format context was allocated.
In the `CVideo::FinishFrames` function, ensure that the codec has been allocated and opened, otherwise using it is not allowed.
Add assertions to the `CVideo::Start` and `Stop` functions to ensure that the same video is not started/stopped multiple times.
Crash with assertion when the size of the graphics is different from the video currently being rendered, instead of causing weirder bugs and a corrupted video file.
The wrong sampling rate was being used for video recording if the client is not restarted after changing the `snd_rate` config variable.
Ensure that the correct bit rate is used if the sample rate was adjusted because the selected value is not supposed.
Use `log_error` for all errors and consistently format all error messages.
Handle all ffmpeg errors and output the formatted ffmpeg error message when possible.
Register a log callback for ffmpeg log messages to delegate them to our logging system, to fix the log messages being interleaved with our log messages and not using the correct line breaks on Windows.
Stop video and demo immediately and show an error message popup if the video could not be started successfully.
Remove unnecessary debug output from ffmpeg.
Removed member prefix m_ used for local variables.
Removed all b, c, i hungarian notation prefixes for boolean, const and integers.
Fixed local variables using lower_snake_case instead of UpperCamelCase.
Renamed all ``float wSearch = TextRender()->TextWidth..`` to ``float SearchWidth = ..``.
When using the `remove_vote` command, the entire heap of vote options is allocated again without the entry being removed. This allocation was not considering the required alignment for `CVoteOptionServer` objects and potentially wasting space by aligning with `alignof(std::max_align_t)`. When allocating the entries with the `add_vote` command, the alignment is already specified correctly.
Relevant upstream commit:
d2924b5ad6
Closes https://github.com/ChillerDragon/ddnet/issues/7
The snap item obj_character contains a field called m_TriggeredEvents
https://chillerdragon.github.io/teeworlds-protocol/07/snap_items.html#obj_character
It is responsible for effects and sounds. Those flags are set in the gamecore.
So if the servers gamecore ticks twice and resets the flags before a snap is
sent the client misses the information. Which is not too big of a
problem since the client has his own gamecore running (prediction) which
also sets those flags. But it is still wrong and teeworlds does always
include the triggered events in the snap.
So this commit fixes it using the same approach as teeworlds.
By not resetting the triggered events until a snap was sent.
Always use the server address from the network client to reconnect to the correct server after being disconnected due to wrong password. The password popup is only shown after being disconnected, so the server address of the network client should always the address of the server which the client should reconnect to with the password. Using `ui_server_address` was causing the wrong address to be used after being redirected to another server and if the config variable is changed while the password popup is open.
Closes#8260.
Fix leak of pending future logger log messages if the future logger is not set, in particular when the `logfile` config variable is not set or the file could not be opened, by setting a logger that discards all log messages in this case.
Closes#8265.
When changing the screen width, height or refresh rate config variables to 0 or negative values, which are not allowed by the backend, automatically revert the config variables to the actual values again to ensure that the config variables stay in sync with the state of the window. This fixes the client crashing in the graphics settings when setting the screen width and height to 0 via the console, which causes a division by zero when calculating the aspect ratio.
According to the documentation, this function returns `0` on success and a negative number (error code) otherwise, which would cause an allocation of an invalid size.
The third parameter of the `op_read` function specifies the remaining size of the buffer, but we always passed the total size of the buffer without respecting the position at which the data is written into the buffer.
Set the data pointer of the sample only when the sample has been loaded successfully, so the invalid sample data is not freed again when decoding fails.
Otherwise the snapshots may incorrectly be invalidated because a change of the `cl_dummy` variable is detected when connecting to another server after disconnecting while the dummy is active.
Send dummy ready and enter game when connection is online based on its state instead of when the security token is not unknown anymore, which should effectively be the same condition but more understandable. This is also how it's already done for the main connection so the `SecurityTokenUnknown` function can be removed as it's unused now.
The console was not keeping its current scoll position if entries from the backlog are removed due to being recycled when new entries are added. For this purpose, a callback function is added to the ringbuffer to handle popped items, so the scrolling position of the console can be updated based on the number of lines of the removed backlog entries.
Fix backlog corruption in `CConsole::PumpBacklogPending` when many backlog entries are allocated at the same time. When allocating many entries from the `m_Backlog` ringbuffer at the same time, the first entries being allocated may already have been recycled again, so the pointers to them being stored in the temporary vector of new backlog entries were pointing arbitrarily into the ringbuffer data, which could cause corruption of the structural and user data of the ringbuffer. Now, we iterate over the entire backlog and only update uninitialized entries instead of storing the new entries separately.
This was sometimes caught as a misaligned access with UBSan:
```
src/engine/shared/ringbuffer.cpp:160:14: runtime error: member access within misaligned address 0x00014126f4df for type 'struct CItem', which requires 8 byte alignment
0x00014126f4df: note: pointer points here
<memory cannot be printed>
0 0x5825349a6a1c in CRingBufferBase::Prev(void*) src/engine/shared/ringbuffer.cpp:160
1 0x5825334e8934 in CTypedRingBuffer<CGameConsole::CInstance::CBacklogEntry>::Prev(CGameConsole::CInstance::CBacklogEntry*) src/engine/shared/ringbuffer.h:59
2 0x5825334d13e6 in CGameConsole::OnRender() src/game/client/components/console.cpp:1259
3 0x582533bce058 in CGameClient::OnRender() src/game/client/gameclient.cpp:715
4 0x582532f3cc44 in CClient::Render() src/engine/client/client.cpp:894
5 0x582532f9d236 in CClient::Run() src/engine/client/client.cpp:2971
6 0x582533002e5e in main src/engine/client/client.cpp:4523
```
Fix out-of-bounds accesses when rendering ghost players, which use the client ID `-2`. This was causing the hook collision line of ghost players to be affected by real players, which is not correct, as ghosts do not interact with other characters.
Closes#8239.
Add optional filename argument to `start_video` command, to start recording to a video file with a specific filename, instead of always using the current timestamp.
Add log messages to `start_video` and `stop_video` commands to indicate success and errors.
Make the `CClient::StartVideo` function non-`static` and reduce duplicate code in the `Con_StartVideo` function.
Determine the video filename outside of the `CVideo` constructor, same as for demos.
Fix game times and prediction not being updated when only exactly two snapshots have been received, due to the conditions `m_aReceivedSnapshots[...] >= 3`. These specific condition are not necessary and replaced with simpler checks whether the current snapshot is set. Some duplicate nested conditions are also removed.
Pump the network first in `CClient::Update` before updating anything else, to ensure that snapshots are received from the network client before the game times and prediction are being updated based on the current snapshots.
Fix current and previous game tick always being `0` on the first call of `IGameClient::OnNewSnapshot` when two snapshot have been received. Now, the game ticks are properly initialized from the two initial snapshots.
Fix old inputs sometimes being used in prediction after changing map because inputs with tick `-1` were not being ignored.
Ensure all snapshots and game times are properly cleared when entering the game.
`GetConsoleMode` can fail with `ERROR_INVALID_HANDLE` when redirecting output to `nul`, which is considered a character file but cannot be used as a console.
The purpose of this condition is to only update the freeze time every 50 ticks so the freeze bar keeps being refilled after one second when standing in freeze, but it incorrectly prevented freeze from being applied during the first 50 ticks after the map has been loaded. Now, freeze is also applied if not currently frozen, so the behavior directly after changing the map is identical to subsequent respawns.
ChillerDragon
