Prevent potential issues when creating the Graphics window (and thus
calling PopulateBackendInfo) while the core state is Stopping, like we
already do while it's Starting or Running.
Remove the PopulateBackendInfoFromUI function, which had a single caller
(GraphicsWindow::OnBackendChanged) and checked that the core wasn't
running or starting before calling PopulateBackendInfo.
Move the core state check into PopulateBackendInfo and have
OnBackendChanged call that instead. This guarantees the check is
performed by all callers of PopulateBackendInfo, preventing
potential reintroduction of the crash fixed in 3d4ae63f if another call
to PopulateBackendInfo is added.
As of the previous commit the only other caller of PopulateBackendInfo
is Core::Init shortly before s_state is set to Starting, so it will
always pass the check and so maintain its current behavior.
Remove the second of two calls to PopulateBackendInfo during emulation
startup. This call was originally the only one, but b214e0e added an
earlier call to handle the backend being changed by the GameINI.
The PR discussion doesn't explain why the original call was left in; I
suspect it was just overlooked.
As a bonus, this removes one of the extra copies of the "Video Info" On
Screen Display message at startup when using OpenGL.
Fix a crash when opening the Graphics window for the first time during
emulation startup when the backend is Vulkan, D3D11, or D3D12.
Don't call PopulateBackendInfo() from the Host thread when the core is
starting up. First, the function has already been called in Core::Init()
so we don't need to again. More importantly, PopulateBackendInfo() calls
g_video_backend->InitBackendInfo(), and the Vulkan and D3D
implementations of those functions load and then unload libraries (and
their associated function pointers) which are potentially in use by
other threads.
This crash was reliably reproducible with the following steps:
1) Select an affected backend.
2) Enable "Compile Shaders Before Starting"
3) Delete the cached shaders (but not the .uidcache file) for the game
you're testing.
4) Close and reopen Dolphin.
5) Start the game.
6) While the game is still booting or compiling shaders, open the
Graphics window for the first time in that Dolphin session.
Fixes https://bugs.dolphin-emu.org/issues/13634.
Arbitrary Mipmap Detection doesn't work when GPU Texture Decoding is
enabled, so disable GPU Texture Decoding for games where the .ini
enables Arbitrary Mipmap Detection.
Co-authored-by: jeremyie <[email protected]>
If texture dumping is enabled, notify the user on emulation startup
using an On Screen Display message.
Also notify the user when texture dumping is toggled.
Addresses https://bugs.dolphin-emu.org/issues/12445.
Set the Render Window as the parent of the Confirm On Stop confirmation
dialog when Keep Window On Top is enabled, ensuring it will always be
visible.
Previously, when Confirm On Stop and Keep Window On Top were both
enabled the Confirm On Stop dialog could be hidden by the render window
in the following situations:
* Clicking Stop in the Main Window
* Clicking the Main Window's close button
* Pressing the Stop hotkey while in FullScreen mode
This was particularly troublesome because the confirm dialog is modal,
preventing the user from moving the render window out of the way if it
was obscuring the dialog.
Fixes https://bugs.dolphin-emu.org/issues/13247.
Allow connecting or disconnecting multiple Wii Remotes simultaneously
instead of only handling the highest index whose hotkey is pressed. This
allows using a single hotkey to toggle multiple remotes.
Before the call to OnSelectionChange, m_code_edit and m_code_remove are
disabled and UpdateList calls m_code_list->clear(), thereby deselecting
any selected items.
When no items are selected, OnSelectionChange disables m_code_edit and
m_code_remove and then returns. Since that was already done, the call
doesn't change anything and can be removed.
Create ARCodeWidget and GeckoCodeWidget once on startup rather than
every time a game is launched or shutdown.
In addition to losing focus on the tab (since the previous widget and
tab no longer existed), the behavior prior to this commit could cause a
crash if the user initiated a game shutdown and then opened a code edit
window since the AR/GeckoCodeWidget would get deleted in the meantime.
Also some minor refactoring of nearby/related code:
* Make non-obvious variable types explicit instead of auto.
* Throw some consts around.
* Use setDisabled(empty) instead of setEnabled(!empty).
Update the checkmarked slot in the Select State Slot menu when the
Increase Selected State Slot or Decrease Selected State Slot hotkeys are
pressed.
The actual selected save slot was being changed correctly before this
commit; this just fixes the menu checkmark.
Use QObject->deleteLater() instead of the delete operator to destroy
child widgets of the layout. This prevents crashes caused by pending
events trying to access the now-destroyed widget.
Enable emulator hotkeys and controller input (when that option is
enabled) when a TAS Input window has focus, as if it was the render
window instead. This allows TASers to use frame advance and the like
without having to switch the focused window or disabling Hotkeys Require
Window Focus which also picks up keypresses while other apps are active.
Cursor updates are disabled when the TAS Input window has focus, as
otherwise the Wii IR widget (and anything else controlled by the mouse)
becomes unusable. The cursor continues to work normally when the render
window has focus.
Move CheatManager's child widgets into scroll areas to allow making the
window smaller than the default.
In CheatSearchWidget, enable word wrapping for the label describing the
address space and search type to help it fit better inside a narrower
window.
* Fix irregularly shaped corners
* Remove extra space for BalloonTips with no message or no title
* When the target tip location is not on a screen, put the tooltip on
the mouse's screen instead of the primary screen
* Fix description getting cut off when the title was too long
* Expose border width as a parameter
* Fix spacing and sizing issues with larger border widths
Use "bzip2" instead of "bzip" in optparse's compression choices for the
convert command. This is both more accurate and matches what the
ParseCompressionTypeString function expects.
The mismatch between the two parsing functions prevented compression
using bzip2 because either ParseCompressionTypeString or optparse would
generate an error when using "bzip" or "bzip2" respectively.
Fixes https://bugs.dolphin-emu.org/issues/13427
Use 1 of the same type as the stored value when shifting left. This
prevents undefined behavior caused by shifting an int more than 31 bits.
Previously iterator incrementation could either hang or prematurely
report it had reached the end of the bitset.
Keep a shared_ptr to NetKDTimeDevice inside NetKDRequestDevice.
This allows the KDDownload task to finish its work without potentially
trying to dereference nullptr, which can potentially come from either
GetIOS() or GetDeviceByName() if EmulationKernel's destructor has
started running.
Explicitly shut down work queues in NetKDRequestDevice's destructor to
prevent their threads from accessing members after they've been freed.
This crash would occur sporadically if NetKDRequestDevice's periodic
download or mail checks happened to overlap with emulation shutdown in
the wrong way.
An example sequence of events that could cause the crash:
* m_scheduler_timer_thread queues a periodic Download event in
m_scheduler_work_queue, then waits for m_shutdown_event.
* A request to stop emulation results in s_ios being reset by the CPU
thread. This triggers NetKDRequestDevice's destructor which sets
m_shutdown_event and joins m_scheduler_timer_thread.
* m_scheduler_timer_thread wakes from m_shutdown_event and returns from
its thread function, ending the thread.
* The CPU thread resumes execution at the end of NetKDRequestDevice's
destructor and begins destroying NetKDRequestDevice's members in
reverse declaration order.
* m_http is declared after m_scheduler_work_queue and is therefore
destroyed earlier.
* m_scheduler_work_queue's destructor calls its Shutdown function, which
by default finishes the work items in the queue.
* The queued Download event calls KDDownload which calls m_http.Get()
which calls Fetch() which passes garbage data from the freed m_curl
into curl_easy_setopt().
* Curl promptly crashes.
Shutting down the work queues manually in the destructor prevents the
above because m_http and the other members don't get freed until after
the queue threads finish.
The move assignment operator for a class is implicitly deleted when the
class has a non-static reference data member, which is true of
WiiSocket's m_socket_manager member.
Explicitly declaring the operator as default generates a
-Wdefaulted-function-deleted warning on Clang.
Delete the move constructor as well for consistency.
Fix -WSwitch warning about unhandled enum value SDL_NUM_LOG_PRIORITIES.
log_level is initialized to LNOTICE right before the switch statement so
this doesn't cause any behavior changes.
Use RunOnCPUThread instead of RunAsCPUThread in BeginRecordingInput.
Most OpenGL functions require an OpenGL context to have been created on
that thread before calling the function; when that isn't the case they
return invalid results which can cause crashes when passed into other
functions.
Dolphin creates the OpenGL context in the EmuThread which then becomes
either the CPU-GPU thread or the Video thread for single and dual core
respectively. OpenGL functions must therefore be called from that
thread.
Movie::BeginRecordingInput is called from the Host thread and runs a
block of code which ultimately creates a savestate, which in turn embeds
the framebuffer which requires calling various OpenGL functions.
In single core the use of RunAsCPUThread leads to this all happening on
the Host thread, eventually leading to invalid OpenGL calls and a crash.
In Dual core the crash is avoided because VideoBackendBase::DoState uses
the AsyncRequests::DO_SAVE_STATE event which causes VideoCommon_DoState
and its subsequent OpenGL calls to safely run on the Video thread.
This commit uses RunOnCPUThread instead of RunAsCPUThread, which causes
the subsequent code to run on the CPU-GPU thread in single core which
has the valid OpenGL context and so doesn't crash.
At the end of each frame automatically update the Current Value for
visible table rows in the selected and visible CheatSearchWidget (if
any). Also update all Current Values in all CheatSearchWidgets when the
State changes to Paused.
Only updating visible table rows serves to minimize the performance cost
of this feature. If the user scrolls to an un-updated cell it will
promptly be updated by either the next VIEndFieldEvent or the State
transitioning to Paused.
The table only needs to be recreated when the displayed addresses might
change. If we're just refreshing the current values then update those
table cells and leave the rest of the table alone.
Using GTEST_SKIP instead of just returning from the function shows that
a test was skipped in the test summary. If GTEST_SKIP is called the rest
of the function won't be run, just like with the return.
GTEST_SKIP wasn't available until gtest 1.10, and we updated to 1.12 in
597f8f1b87.
Resolve warning caused by using values from two different enums in a
conditional expression which was deprecated in c++20.
The warning in question is clang -Wdeprecated-anon-enum-enum-conversion
and gcc -Wenum-compare.
Modify PPCSTATE_OFF and PPCSTATE_OFF_ARRAY macros when using GCC to
avoid useless log spam. Specifically, use a consteval lambda with gcc
_Pragma statements to disable the -Winvalid-offsetof warning inside the
macros.
Each successful build (and many failing ones) on the Android buildbot
generates almost 300 cases of -Winvalid-offsetof, resulting in thousands
of lines of log spam per build. In addition to bloating the log filesize
these spurious warnings make it harder to find actual warnings.
These warnings are generated by calls to the macros PPCSTATE_OFF and
PPCSTATE_OFF_ARRAY, which in turn are used by many other macros used by
the JIT. The ultimate cause is that offsetof is only conditionally
supported on non-standard-layout types, which includes the PowerPCState
struct.
To address potential questions of whether there's a better way to handle
this:
The obvious solution would be to modify PowerPCState so that it does
have a standard layout. This is unfortunately impractical.
To have a standard layout a type can only contain other types with
standard layouts. None of the stl containers are guaranteed to have
standard layouts, and PowerPCState contains a std::tuple and std::array.
PowerPCState also contains a PowerPC::Cache and InstructionCache which
themselves contain std:arrays and std::vectors.
Furthermore InstructionCache derives from Cache, and a derived class can
only have standard layout if at most one class in its hierarchy has a
non-static data member, but both classes have such members. Making
InstructionCache have a standard layout would require duplicating all
the functionality of Cache so it no longer derived from it, as well as
replacing the stl containers. This might require having a raw pointer to
said containers, with the manual memory management that implies.
All of that would be much more disruptive than would be justified to get
rid of some warnings (however annoying they might be). This is
compounded by the fact that PowerPCState hasn't had a standard layout
for a long time, if ever, and if the PPCSTATE_OFF macros weren't working
reliably it would have become obvious a long time ago.
As to why I picked the lambda solution over other potential changes:
- Keeping the define as-is and wrapping some gcc #pragmas around it
doesn't work because the pragmas don't get included when the define is
substituted to the call site.
- Keeping the define as a non-lambda expression and using inline
_Pragma() statements would ideally be better and works fine for msvc,
but fails for GCC with "'#pragma' is not allowed here".
- Turning off -Winvalid-offsetof globally for gcc would work, but there
might be other contexts where offsetof is problematic and GCC seems to
be the only compiler warning about it.
Android interprets char as unsigned char, so comparing with 0 triggers a
tautological-unsigned-char-zero-compare warning.
Casting c to an unsigned char and removing the comparison with 0
resolves the warning while needing one less comparison on all platforms.
Check bytes directly to avoid ambiguity in the disassembly between short
and near jumps, which could hypothetically cause the test to pass when
it shouldn't.
Verify that DXFramebuffer's integer RTV descriptor's cpu_handle has been
allocated before using it, and if it hasn't use the non-integer RTV
descriptor instead. This fixes a Dolphin crash in Twilight Princess, and
possibly other games (Issue 13312).
As an optimization to save space in the descriptor heap, DXFramebuffer's
integer descriptor is only initialized if the given abstract texture
format has different integer and non-integer RTV formats. This
previously wasn't accounted for by GetIntRTVDescriptorArray, which could
cause DX12::Gfx::BindFramebuffer to call OMSetRenderTargets with an
invalid descriptor which would lead to a crash.
Triggering the bug was fortunately rare because integer formats are only
used when blending is disabled and logic ops are enabled. Furthermore,
the standard integer abstract format is RGBA8 which has different
integer and non-integer RTV formats, causing the integer descriptor to
be initialized and avoiding the bug.
The crash started appearing in a2702c6 because it changed the
swapchain's abstract texture format from RGBA8 to RGB10_A2. Unlike
RGBA8, RGB10_A2 has the same integer and non-integer RTV formats and so
the bug can be triggered if the other requirements are met.
The ExpansionInterfaceManager::PauseAndLock function does nothing but
call other functions that have no effect.
ExpansionInterfaceManager::PauseAndLock calls CEXIChannel::PauseAndLock,
which in turn calls IEXIDevice::PauseAndLock. None of the classes
deriving from IEXIDevice override PauseAndLock, and the implementation
in IEXIDevice does nothing.
Dentomologist
