IOLinux.cpp should include <sys/select.h> as it uses select() functionality.
On certain platforms it's included implicitly by other headers, which is why
it compiled before. This makes it also work on musl platforms.
libusb transfer callbacks might be called immediately during transfer
submission in some cases. (libusb doesn't even specify what thread
the callback is invoked on.) In other words, it is possible to reach
the transfer callback from the CPU thread, and not just from the
USB event handling thread.
So CoreTiming::FromThread::NON_CPU is incorrect and should instead
be ANY.
Unfortunately, it appears that using libusb's synchronous transfer API
from several threads causes nasty race conditions in event handling and
can lead to deadlocks, despite the fact that libusb's synchronous API
is documented to be perfectly fine to use from several threads (only
the manual polling functionality is supposed to require special
precautions).
Since usbdk was the only real reason for using a single libusb context
and since usbdk (currently) has so many issues with Dolphin, I think
dropping support for it in order to fix other backends is acceptable.
Now that the floating point members are assigned in bulk, we can remove
their setter macro. While we're at it, we can also remove the setter for
unsigned int, given it's not used.
We were doing quite a bit of unnecessary work within CMake to handle and
make sure the necessary libraries were copied over. That approach has
several downsides:
1. It's not possible to handle multi-configuration generators (like
Visual Studio) in an easy manner. The existing script would fail to
copy over the necessary libraries if one configuration was built, and
then another one was built.
2. If you have Qt already installed (properly) by the official binary,
the existing script would copy *all* dlls even if they weren't
necessary. This is pretty bad, since it can waste quite a bit of
space.
Instead, we can just delegate off to the official deployment application
bundled with Qt's libraries that determines what the necessary libraries
are and copies them over as necessary. This also means we can properly
support both release and debug binaries in the same directory, like how
the old handcrafted Visual Studio project files allowed.
Its sufficient to simply specify a debug postfix instead of using an
separate variable. What's nice about this approach is that it will
actually work :p
Previously the code wouldn't work for multi-configuration generators
like Visual Studio.
When the bluetooth adapter device is opened/closed by dolphin, the
kernel driver is automatically detached/reattached.
This enables transparent sharing of the same bluetooth wiimotes and
bluetooth adapters between the hosts system and the emulated one using
the same.
ImGui::Text() assumes that the incoming text is intended to be
formatted, but we don't actually use it to format anything. We can be
explicit by using the relevant function.
This also has a plus of not needing to go through the formatter itself,
but the gains from that are probably minimal.
Previously these functions were declared without the static specifier,
giving them external linkage, which isn't really ideal.
Instead, we can place these functions up by the relevant file-scope
variables and place them inside an anonymous namespace with said variables,
giving them internal linkage.
Avoids the use of the null pointer to represent an empty string.
Instead, we can simply pass an empty string_view instance. Using
std::string_view enforces this invariant at the API level.
Due to the lack of cast here, this will actually print out the ascii
value, rather than the character itself, due to promoting to integral
values. Instead, we can eliminate the use of character operands and just
print the value itself directly, given it's equivalent behavior with
less code.
Allows these arrays to be placed within the read-only segment (and
enforces the immutability in the code itself). While we're at it, we can
make use of std::array here.
Now that the std::map less-than comparitor is capable of being used with
heterogenous lookup, we're able to convert many of the querying
functions that took std::string references over to std::string_view.
Now these functions may be used without potentially allocating a
std::string instance unnecessarily.
Previously, when performing find() operations or indexing operations on
the section map, it would need to operate on a std::string key.
This means cases like:
map.find(some_string_view)
aren't usable, which kind of sucks, especially given for most cases, we
use regular string literals to perform operations in calling code.
However, since C++14, it's possible to use heterogenous lookup to avoid
needing to construct exact key types. In otherwords, we can perform the
above or use string literals without constructing a std::string instance
around them implicitly.
We simply need to specify a member type within our comparison struct
named is_transparent, to allow std::map to perform automatic type
deduction.
We also slightly alter the algorithm to an equivalent compatible with
std::string_view (which need not be null-terminated), as strcasecmp
requires null-terminated strings.
While we're at it, we can also provide a helper function to the struct
for comparing string equality rather than only less than. This allows
removing other usages of strcasecmp in other functions, allowing for the
transition of them to std::string_view.
fmt diverges from printf in that '.' as a precision specifier may only
be used for floating-point values (makes sense, given it's indicating
precision after the decimal point).
This fixes the problem where OBS game capture only grabs the region
inside an ImGui window whenever one is open, when using the OpenGL
backend. Shouldn't have any negative effects, as the scissor would've
been something completely arbitrary anyways.
This may affect other capture software that uses the same hooking
method, but I've only tested OBS.
In a few cases we needed to alter... less than ideal parameter types.
While u8 may have been OK with printf-style formatting, which promotes
most smaller types back to int, this won't work with fmt. fmt preserves
the type of the passed in arguments, meaning that u8, being an alias of
uint8_t (itself being an alias of unsigned char on all the platforms we
support), will print out as a character, not a numeric value.
As such, we amend some functions to operate on u32 values for two
reasons:
1. We actually want it to print out as a value
2. Arithmetic on unsigned types smaller than unsigned int will actually promote to an int,
not unsigned int. This is very non-obvious to some and makes for
error-prone code. < sizeof(int) types are great for storage, not so
much for performing unsigned arithmetic, despite the signedness of
the type.
While we do have this library as part of the public linkage interface in
the common library target, which will be used in the future for the
logging macros, we should still be explicit that we're using this
library. Therefore, we privately link it in to be explicit about it.
Begins the transition to using fmt for string formatting where
applicable. Given fmt supports formatting std::string instances out of
the box, we can remove now-unnecessary calls to .c_str() and .data().
Note that this change does not touch the actual logging subsystem aside
from converting the final StringFromFormat call in the process over to
fmt::format. Given our logging system is heavily used throughout the
entire codebase, and converting that over will be quite a large change
by itself, this will be tackled near the end of the conversion process.
Currently, it is possible for the DiscordHandler thread to be in the
middle of sleeping when Dolphin is closing. This results in a very
noticeable delay of up to 2 seconds that is unacceptable, especially
for people who don't use the Discord integration.
This fixes the issue by making the thread wait on an Event instead
and signalling it when shutting down.
It already is disabled for other backends, but this didn't happen with the software renderer. Attempting to change it while running causes the change to visually happen (including switching to the normal render settings UI instead of the barren one for the software renderer), but doesn't actually change the backend itself (it'll still use the software renderer at the next launch).
Previously, this array potentially wouldn't be placed within the
read-only segment, since it wasn't marked const. We can make the lookup
table const, along with any other nearby variables.
It appears that some older drivers do not support
CreateSwapChainForHwnd, resulting in DXGI_ERROR_INVALID_CALL. For these
cases, fall back to the base CreateSwapChain() from DXGI 1.0.
In theory this should also let us run on Win7 without the platform
update, but in reality we require the newer shader compiler so this
probably won't work regardless. Also any hardware of this vintage is
unlikely to run Dolphin well.
We can use std::array and const char* to make these capable of fully
being stored in the read-only segment, and get rid of a few static
constructors (144 of them).
Avoids dragging in IniFile, EXI device and SI device headers in this header which is
quite widely used throughout the codebase.
This also uncovered a few cases where indirect inclusions were being
relied upon, which this also fixes.
We can std::move the std::string parameter in Label's constructor,
allowing the constructor to be moved into in calling code.
We can cascade this outwards in the interface as well.
Given this is a private struct and it's used in a container that
supports incomplete types, we can forward-declare it and move it into
the cpp file. While we're at it, we can change the name to Label to
follow our formatting guidelines.
We also move the destructor definition into the cpp file, as that will
allow us to make the entire label_t type hidden from external view in a
following change.
This aims to emulate rotational latency as part of the disc drive
timings, which makes loading times have more variance (like on a
real disc drive) but should not affect the average loading times.
Starting with C++17, this allows for the same behavior as the existing
code, but without the need to manually write the out-of-class
definitions as well.
Allows for both string types to be non-allocating. We can't remove the
const char* overload in this case due to the fact that pointers can
implicitly convert to bool, so if we removed the overload all const
char arrays passed in would begin executing the bool overload instead of
the string_view overload, which is definitely not what we want to occur.
This fixes the desync on playback of start-from-boot input recordings
made while using the GC adapter, as well as other desyncs that could
potentially occur in other circumstances where this bit is used.
I used a previously reserved bit in the ControllerState to store the
new data, so this shouldn't significantly break backwards
compatibility. However, tools that aren't aware of this new bit may set
it to 0, which will break input recordings that contain it.
Since C++17, non-member std::size() is present in the standard library
which also operates on regular C arrays. Given that, we can just replace
usages of ArraySize with that where applicable.
In many cases, we can just change the actual C array ArraySize() was
called on into a std::array and just use its .size() member function
instead.
In some other cases, we can collapse the loops they were used in, into a
ranged-for loop, eliminating the need for en explicit bounds query.
MessageData must be a trivially copyable type, given it's copied into
emulated memory via our memory copy function CopyToEmu. Under the
covers, this function utilizes memcpy. One of memcpy's requirements is
that pointers to it point to types that are trivially copyable,
otherwise the behavior is undefined.
Given that, we can enforce this requirement at compile-time.
Simplifies initialization code quite a bit, and replaces a pointer
variable for SMessageData with a type properly representing the whole
set of data it needs.
These aren't modified by the class, nor do they directly need anything
related to the class state, so they can solely live within the cpp file,
hidden from external view, and also be made const, so the compiler can
place it within the read-only segment.
Makes the names consistent between declaration and definition and
adjusts them to follow our code formatting guidelines.
Now all functions in the translation unit follow our formatting
guidelines.
Constructs the strings directly within the container instead of
performing a construction, then a copy.
The reasoning is that the BACKEND_* strings are const char arrays, so
the push_back code is equivalent to:
push_back(std::string(BACKEND_WHATEVER)) instead of forwarding the
arguments to a constructed instance directly in the container.
This changes the Host Input Authority and Golf Mode checkboxes into a
set of radio buttons, consisting of Fair Input Delay, Host Input
Authority, and Golf Mode. This represents the 3 network modes we have.
Although Golf Mode is just an extension of Host Input Authority, it's
more logical to the user to present it as a separate option, rather
than enabling the Golf Mode checkbox only when Host Input Authority is
enabled. This also eliminates the need to first enable Host Input
Authority before Golf Mode can be enabled.
This also adds tooltips to provide brief descriptions of the options,
as well as reintroducing tooltips that were previously removed.
We can use u32 instead of unsigned int to shorten up these definitions
and make them much nicer to read.
While we're at it, change the size array to house u32 elements
to match the return value of the function.
We can use u32 instead of unsigned int to shorten up these definitions
and make them much nicer to read.
While we're at it, change the size array to house u32 elements to match
the return value of the function.
This is only ever used to retrieve a raw view of the given UID data
structure, however it's already valid C++ to retrieve a char/unsigned
char view of an object for bytewise inspection.
u8 maps to unsigned char on all platforms we support, so we can just do
this directly with a reinterpret cast, simplifying the overall
interface.
Zero-initialization zeroes out all members and padding bits, so this is
safe to do. While we're at it, also add static assertions that enforce
the necessary requirements of a UID type explicitly within the ShaderUid
class.
This way, we can remove several memset calls around the shader
generation code that makes sure the underlying UID data is zeroed out.
Now our ShaderUid class enforces this for us, so we don't need to care about
it at the usage sites.
Now that we utilize C++17, we can simply return an optional containing
the code instead of using an out variable and a boolean result,
essentially combining them into one.
This provides a much more straightforward interface.
Greatly simplifies the overall interface when it comes to compiling
shaders. Also allows getting rid of a std::string overload of the same
name. Now std::string and const char* both go through the same function.
Makes VertexLoader_Normal completely stateless, eliminating the need for
an Init() function, and by extension, also gets rid of the need for the
FifoAnalyzer to have an Init() function.
Many of the arrays defined within this file weren't declared as
immutable, which can inhibit the strings being put into the read-only
segment. We can declare them constexpr to make them immutable.
While we're at it, we can use std::array, to allow bounds conditional
bounds checking with standard libraries. The declarations can also be
shortened in the future when all platform toolchain versions we use
support std::array deduction guides. Currently macOS and FreeBSD
builders fail on them.
The vector is only ever queryied and it's contents aren't modified, so
there's no reason to take the vector by value. We can take a constant
reference to it to avoid unnecessary allocating.
In these cases, the given string is only ever compared against other
string, so std::string can be turned into a std::string_view to allow
non-allocating inputs.
If opening the adapter fails, report the libusb error message in the GUI
instead of “No Adapter Detected”.
The error condition is removed when the adapter is unplugged.
Ensures that the destruction logic is kept local to the translation unit
(making it nicer when it comes to forward declaring non-trivial types).
It also doesn't really do much to define it in the header.
Given we're simply storing the std::string into a deque. We can emplace
it and move it. Completely avoiding copies with the current usage of the
function.
The pack is already has its validity checked at the beginning of the
function, so we don't need to check this again after inserting it.
Also resolves a use-after-move case.
Same behavior but doesn't take up 8 bytes in the executable to store a
pointer. While we're at it, move it into an anonymous namespace within
the UICommon namespace.
We already check ahead of time if the optional contains a value within
it before accessing it, so we don't need to use the throwing value()
accessor. We can just directly use operator->
Rather than construct strings twice, we can just construct it once and
reuse it. While we're at it, we can move variables closer to where
they're actually used within DownloadDefaultCover()
* Simplifies libusb context usage and allows us to set options for
all contexts easily. Notably, this lets us enable usbdk support
in libusb, which is now opt-in in the latest version.
* Moves the libusb config descriptor wrapper class to LibusbUtils too
since that could easily be reused.
* Moves device listing to LibusbUtils too and add a lock around it
as some libusb backends are not thread safe.
* Consequences: only a single context and a single event handling
thread is used now, which is more efficient.
This allows the same code to be used to read into a std::string, which
allows for eliminating the vector->string transfer when reading the
manifest file.
A ContiguousContainer is a concept that includes std::array,
std::string, and std::vector.
Makes it way harder to introduce resource leaks, and plugs the existing
resource leaks in the constructor and Install() where the file wouldn't
be closed in some error cases.
Now that we assume C++17, the in-file definition of the std::array can
be removed. This is all that's necessary, as constexpr used on a static
member variable implies inline (and so, automatically has C++17's static
inline behavior).
Removes redundant initializers from the constructor and provides
initializers for all members that don't already have one for consistency
(and deterministic initial state).
Given the volume verifier has quite a few non-trivial object within it,
it's best to default the destructor within the cpp file to prevent
inlining complex destruction logic elsewhere, while also making it nicer
if a forward-declared type is ever used in a member variable.
std::call_once is guaranteed to execute the given callable object
exactly once. This guarantee holds even if the function is called
concurrently from several threads.
Given that, we can replace the mutex and boolean flag with
std::call_once and a std::once_flag to perform the same behavior.
Previously, every entry pair within the map would be copied. The reason
for this is subtle.
A std::map's internal entry type is defined as:
std::pair<const Key, Value>
but the loop was declaring it as:
std::pair<Key, Value>
These two types aren't synonymous with one another and so the compiler
is required to always perform a copy.
Using structured bindings avoids this (as would plain auto or correcting
the explicit type), while also allowing the use of more appropriate
names compared to first and second.
std::function is allowed to heap allocate in order to hold any necessary
bound data in order to execute properly (e.g. lambdas with captures), so
this avoids unnecessary reallocating.
The previous implementation of cheat search would reconvert the input
string for every single memory value. Now we do it once and construct
a comparison lambda which we pass to the search code.
In addition, I also added input validation. So, for example, if you've
selected Decimal input and you try to compare against "FF",
it won't search and will instead let the user know they've entered an
invalid value. Similar logic for if you enter "1.2" in a search for
bytes. Before, it would just use 0 if it failed to convert the value.
- Files for the current game are now guaranteed to be loaded, space and validity permitting.
- Avoid showing PanicAlerts for every problem encountered, most of them aren't really important enough and will probably just annoy the user.
- And for the only error the user will definitely care about, when the save of the game they're trying to play fails to load, show an imgui message instead.
To avoid having to claim/release interfaces all the time, and having to
trigger interface changes from several places, all interfaces are now
claimed ahead of time.
This commit also makes us avoid changing the active interface when it's
not necessary.
Changing the active interface has side effects such as resetting the
active alternate setting -- which is extremely undesirable because it
would require the emulated software to change the alternate setting
again, which isn't supposed to be necessary at all.
This fixes Your Shape, which submits isochronous transfers on an
endpoint that only exists in alt setting 6 right after submitting
control transfers (which would have reset to alt setting 0 prior to
this fix).
Changed itemSelectionChanged and itemClicked signal to itemPressed in CodeWidget.
Holding mouse down and moving will only travel up/down the stack one time.
This fixes the common occurrence of unintentionally traveling deeper down the stack or higher up the callstack than intended.
This allows us to update the rich presence description if a channel
is launched from the Wii Menu. It also handles other PPC title
launches, e.g. Smash Bros. Masterpieces.
Host.h: Added Host_TitleChanged().
DolphinNoGUI/MainNoGUI.cpp: Implemented Host_TitleChanged().
DolphinQt/Host.cpp: Implemented Host_TitleChanged().
Android/jni/MainAndroid.cpp: Stubbed Host_TitleChanged().
DSPTool/StubHost.cpp: Stubbed Host_TitleChanged().
UnitTests/StubHost.cpp: Stubbed Host_TitleChanged().
While current usages of ParseLine aren't problematic, this is still a
public function that can be used for other purposes. Essentially makes
the function handle potential external inputs a little nicer.
We can just utilize map's insert_or_assign() function and check the
return value to determine whether or not we need to insert the key into
the keys_order vector.
Even though libusb is supposed to be thread-safe, in practice
it's not (at least on Windows); getting a list of devices from two
different threads can result in libusb crashes. This is easily
fixed by waiting for the scan thread to complete scanning instead
of running the scan on the CPU thread.
This also fixes an issue that I had overlooked in the initial
implementation: IOS interfaces such as OH0 are sometimes opened
every frame, in which case we were doing a full device scan every
single frame on the CPU thread!
Fixes an embarrassing bug that made the implementation utterly useless.
This fixes Your Shape hanging on shutdown. The game was waiting for an
interrupt transfer to be cancelled, and Dolphin wasn't cancelling
transfers on the correct endpoint.
The total buffer size for isochronous transfers should be a u32,
not a u16. It is easy to hit the bug with devices such as cameras,
which require larger buffers.
This fixes Your Shape.
This also fixes the length type for bulk and interrupt transfers,
which should be u32 as that's what IOS supports. I'm not sure why
I made them u16... probably because OH0 uses u16 for most lengths...
In terms of order of operations, the move would occur first before the
construction of the relevant reader would occur. However, given the
local variable 'path' was declared const, this bug actually wouldn't
occur, as std::move on a const variable does nothing (in a non-mutable
context), resulting in a copy instead, masking this issue.
Given this is a bug waiting to happen, we correct the code.
This adds the center to the calibration menu when it is nonzero.
It also refactors CENTER_COLOR to be a function, similar to other colors
after an earlier commit.
This is useful in far out-of-calibration controllers, such as the
Switch Pro controller. This also adds support for configuring the center
in the Mapping widget.
* Use font based sizing for row height. Fits more rows on screen.
* Adds whitespace for better formatting and minimum column width.
Helps prevent frequent resizing while scrolling.
The main EmuThread (in Core) is responsible for both initialising the
audio stream and shutting it down properly.
When the core is shutting down (when state is State::PowerDown), it is
possible that the CPU or CPU-GPU thread and the UI thread will both
try to stop the audio stream at the same time, which is an issue
because some audio backends such as cubeb are not thread-safe.
This commit prevents the race from ever happening in the first place
by removing the call to AudioCommon::SetSoundStreamRunning from
CPU::RunAdjacentSystems, which is actually completely unnecessary when
shutting down because Core::EmuThread is going to stop the stream and
perform more cleanup anyway.
Should fix https://bugs.dolphin-emu.org/issues/11722
WindowModal allows alt+tabing to the render window, but prohibits interaction
with parent windows (controller settings window and the main dolphin window).
The system menu is passing the SHA1 hash of the save data to ES
to sign, not the save data itself.
Fixes save import in the System Menu for saves that were exported
by Dolphin.
All supported platforms now have easy access to a compiler with C++17
support.
C++17 potentially allows for some nice cleanups and removes the need
for standard library backports (optional/variant).
See discussion at https://dolp.in/pr6264#discussion_r158134178
Different address spaces can be chosen in the memory view panel.
* Effective (or virtual): Probably the view people mostly want. Address
translation goes through MMU.
* Auxiliary: ARAM address space. Does not display anything in Wii mode.
* Physical: Physical address space. Only supports mem1 and mem2 (wii
mode) so far.
MemoryWatcher only works on Linux and affects emulation determinism due
to scheduling additional events, which causes NetPlay to desync.
Considering that this interface is a rather specialized use case, the
communication with it is kinda crappy *and* it's affecting emulation, I
think it's best to just axe it and come up with a better implementation
of the functionality.
* The high half of the register is immediately masked so the value in it is irrelevant.
* MOVSD produces an unnecessary dependency on the high half of regOp.
* MOVAPS is implemented as a register rename on modern microarchitectures.
* The high half of regOp is immediately overwritten so the value in it is irrelevant.
* MOVSD produces an unnecessary dependency on the high half of regOp.
* MOVAPS is implemented as a register rename on modern microarchitectures.
There is no reason to use MOVAPD over MOVAPS, for two reasons:
* There has never been a microarchitecture with separate single and double domains.
* MOVAPD is one byte longer than MOVAPS
If bounding box is enabled when a UID cache is created, then later disabled,
we shouldn't emit the bounding box portion of the shader.
Fixes pipeline creation errors on D3D12 backend for this case.
Because the GC language setting cannot be set to Japanese, we
need a special condition for Japanese GC games. I accidentally
removed it in PR 7816, but here it is again in a new form.
We could do the same thing with Korean GC games if we want to
(which we couldn't do before PR 7816), but due to how spotty
GameTDB is with having Korean names for Korean GC releases,
things will be more consistent if we just use English for them.
psq_l with QUANTIZE_FLOAT does not use the FPU, so it does not trim the precision of the u32 input data.
We already have the helper ConvertToDouble for floating point u32->u64 convertion used in lfs, so let's use it here as well.
The logic didn't account for the case where a player leaves, so the
host would be left in a dangling state where the UI is disabled but the
game won't start, requiring a full restart of Dolphin to fix.
We can't join a detached thread, so NetPlayIndex gets deleted before
the notification thread exits, creating a race condition. We switch to
using Common::Event because just sleeping leaves the UI hung on the
thread join for a few seconds.
This is an extension of host input authority that allows switching the
host (who has zero latency) on the fly, at the further expense of
everyone else's latency. This is useful for turn-based games where the
latency of players not on their turn doesn't matter.
To become the so-called golfer, the player simply presses a hotkey.
When the host is the golfer, latency is identical to normal host input
authority.
Since the copy X and Y coordinates/sizes are 10-bit, the game can configure a
copy region up to 1024x1024. Hardware tests have found that the number of bytes
written does not depend on the configured stride, instead it is based on the
size registers, writing beyond the length of a single row. The data written
for the pixels which lie outside the EFB bounds does not wrap around instead
returning different colors based on the pixel format of the EFB.
This suggests it's not based on coordinates, but instead on memory addresses.
The effect of a within-bounds size but out-of-bounds offset
(e.g. offset 320,0, size 640,480) are the same.
As it would be difficult to emulate the exact behavior of out-of-bounds reads,
instead of writing the junk data, we don't write anything to RAM at all for
over-sized copies, and clamp to the EFB borders for over-offset copies.
Connor McLaughlin