* video_core/renderer_opengl: Move SurfaceParams into its own file
Some of its enums are needed outside of the rasterizer cache
and trying to use it caused circular dependencies.
* video_core/renderer_opengl: Overhaul the texture filter framework
This should make it less intrusive.
Now texture filtering doesn't have any mutable global state.
The texture filters now always upscale to the internal rendering resolution.
This simplifies the logic in UploadGLTexture and it simply takes the role of BlitTextures at the end of the function.
This also prevent extra blitting required when uploading to a framebuffer surface with a mismatched size.
* video_core/renderer_opengl: Use generated mipmaps for filtered textures
The filtered guest mipmaps often looked terrible.
* core/settings: Remove texture filter factor
* sdl/config: Remove texture filter factor
* qt/config: Remove texture filter factor
* vm_manager: Handle multiple areas in ChangeMemoryState
It is possible that a few areas have the same permisson and state, but with different backing pointers. Currently, this function assumes that only one continous area is found, but this is not always the case.
* service/ldr_ro: Handle multiple areas in VerifyBufferState
It is possible that the buffer passed from the game is made up of multiple areas with the same permisson and state but different backing pointers. Change the check to allow that.
The most important one being adding a mutex to protect the format_context. Apparently it wasn't thread safe (as one'd expect) but I didn't think about that.
Should fix some of the strange issues happening with MP4 muxers, etc.
The file's size is stored in FileSessionSlot and retrieved when the game calls GetSize. However, it is not updated when the file is written to, which can possibly change the file size. Therefore, this can cause GetSize to return incorrect results.
According to HW tests, this vsync event is signaled for activated cameras at about the same frequency as the frame rate. The last 5 vsync timings are recorded (in microseconds) and can be retrieved with the service function.
Also, corrected the default frame_rate to 15, according to HW test.
This should fix the missing camera images in certain games.
You can now directly place ExeFS overrides/patches inside the mod folder (instead of the exefs subfolder). This allows us to have drop-in compatibility with Luma3DS mods.
These two functions allow the frontend to get a list of encoders/formats and their specific options.
Retrieving the options is harder than it sounds due to FFmpeg's strange AVClass and AVOption system. For example, for integer and flags options, 'named constants' can be set. They are of type `AV_OPT_TYPE_CONST` and are categoried according to the `unit` field. An option can recognize all constants of the same `unit`.
Previously, we just used the native sample rate for encoding. However, some encoders like libmp3lame doesn't support it. Therefore, we now use a supported sample rate (preferring the native one if possible).
FFmpeg requires audio data to be sent in a sequence of frames, each containing the same specific number of samples. Previously, we buffered input samples in FFmpegBackend. However, as the source and destination sample rates can now be different, we should buffer resampled data instead. swresample have an internal input buffer, so we now just forward all data to it and 'gradually' receive resampled data, at most one frame_size at a time. When there is not enough resampled data to form a frame, we will record the current offset and request for less data on the next call.
Additionally, this commit also fixes a flaw. When an encoder supports variable frame sizes, its frame size is reported to be 0, which breaks our buffering system. Now we treat variable frame size encoders as having a frame size of 160 (the size of a HLE audio frame).
We previously assumed that the first preferred sample format is planar, but that may not be true for all codecs. Instead we should find a supported sample format that is planar.
While YUV420P is widely used, not all encoders accept it (e.g. Intel QSV only accepts NV12). We should use the codec's preferred pixel format instead as we need to rescale the frame anyway.
This uses the mailbox model to move pixel downloading to its own thread, eliminating Nvidia's warnings and (possibly) making use of GPU copy engine.
To achieve this, we created a new mailbox type that is different from the presentation mailbox in that it never discards a rendered frame.
Also, I tweaked the projection matrix thing so that it can just draw the frame upside down instead of having the CPU flip it.
This slider affects the number of cycles that the guest cpu emulation
reports that have passed since the last time slice. This option scales
the result returned by a percentage that the user selects. In some games
underclocking the CPU can give a major speedup. Exposing this as an
option will give users something to toy with for performance, while also
potentially enhancing games that experience lag on the real console
* Core::Timing: Add multiple timer, one for each core
* revert clang-format; work on tests for CoreTiming
* Kernel:: Add support for multiple cores, asserts in HandleSyncRequest because Thread->status == WaitIPC
* Add some TRACE_LOGs
* fix tests
* make some adjustments to qt-debugger, cheats and gdbstub(probably still broken)
* Make ARM_Interface::id private, rework ARM_Interface ctor
* ReRename TimingManager to Timing for smaler diff
* addressed review comments
* HTTP_C::Implement Context::MakeRequest
* httplib: Add add_client_cert_ASN1 and set_verify
* HTTP_C: Fix request methode strings case in MakeRequest
* HTTP_C: clang-format and cleanups
* HTTP_C: Add comment about async in BeginRequest and BeginRequestAsync
* Update httplib to contain all the changes we need; adapt http_c and web_services to the changes in httplib; addressed minor review comments
* Add android-ifaddrs
10 slots are offered along with 'Save to Oldest Slot' and 'Load from Newest Slot'.
The savestate format is similar to the movie file format. It is called CST (Citra SavesTate), and is basically a 0x100 byte header (consisting of magic, revision, creation time and title ID) followed by Zstd compressed raw savestate data.
The savestate files are saved to the `states` folder in Citra's user folder. The files are named like `<Title ID>.<Slot ID>.cst`.
This holds the archives which include the SelfNCCH archive which holds the RomFS files. If we don't reset it the LayeredFS class can't get destructed and mods files won't be released.
The original path (file_name.exefsdir) is still supported, but alternatively users can choose to put exefs patches in the same place as LayeredFS files (`load/mods/<Title ID>/exefs`).
Only enabled for NCCHs that do not have an override romfs.
LayeredFS files should be put in the `load` directory in User Directory. The directory structure is similar to yuzu's but currently does not allow named mods yet. Replacement files should be put in `load/mods/<Title ID>/romfs` while patches/stubs should be put in `load/mods/<Title ID>/romfs_ext`.
This implementation is different from Luma3DS's which directly hooks the SDK functions. Instead, we read the RomFS's metadata and figure out the directory and file structure. Then, relocations (i.e. replacements/deletions/patches) are applied. Afterwards, we rebuild the metadata, and assign 'fake' data offsets to the files. When we want to read file data from this rebuilt RomFS, we use binary search to find the last data offset smaller or equal to the given offset and read from that file (either from the original RomFS, or from replacement files, or from buffered data with patches applied) and any later files when length is not enough.
The code that rebuilds the metadata is pretty complex and uses quite a few variables to keep track of necessary information like metadata offsets. According to my tests, it is able to build RomFS-es identical to the original (but without trailing garbage data) when no relocations are applied.
The DIGIT filter was incorrectly implemented as preventing all digits. It actually limits the maximum digit count to max_digits, according to ctrulib and hardware testing.
This is based on what was done using additional layouts, but modified
to have a variable to control rotation and making it so Single Screen
Layout behaves like Upright Single would, and Default Layout behaves
like Upright Double would, when the new variable is used.
Large Layout and Side Layout currently ignore the new variable.
New variable still currently doesn't have a hotkey.
The BPS format allows distributing patches that are smaller and that do
not contain copyrighted content if data is relocated
(unlike non-trivial IPS patches).
This is essential for games such as MM3D that have three barely
different code revisions. Supporting all three versions would
demand an unreasonable amount of work; with BPS patches only one
version has to be supported.
This changes ApplyCodePatch to return a ResultStatus, which makes it
possible to determine whether patch applying has failed. Previously,
only a boolean was returned, and false was returned when no patch
was found OR when a patch was found but applying it failed.
This also changes AppLoader_NCCH to return an error if patching fails
because the executable is likely to be left in an inconsistent state
and we should not proceed booting in that case.
The text shared memory wasn't supposed to be cleared according to my comparison with the LLE swkbd. This can cause issues in certain games such as Harvest Moon.
A null terminator is added to the text copied to mark the end of the string.
While QOpenGLWidget sounds like a good idea, it has issues which are
harder to debug due to how Qt manages the context behind the scenes. We
could probably work around any of these issues over time, but its
probably easier to do it ourselves with a QWindow directly.
Plus using QWindow + createWindowContainer is the easiest to use
configuration for Qt + Vulkan so this is probably much better in the
long run.
Fixes a regression from #4862, which caused the NCCH title ID checking
heuristic to be skipped whenever the exheader is replaced.
I was thinking the heuristic wouldn't be needed in that case, but it
turns out that many users still have pathological NCCHs that indicate
they are encrypted but are actually decrypted...
Now the original exheader is always read and used for the heuristic
to determine whether the NCCH is actually encrypted; only then do we
load a replacement exheader (if it exists) to avoid affecting the
heuristic.
Separate options are now provided for FFmpeg AAC audio decoder and FFmpeg video dumper. This allows users to configure Citra with greater freedom.
Also, previously for Linux builds, AAC decoder is accidentally enabled along with the dumper, which could potentially cause patent issues (?). This commit fixes it by only enabling video dumper.
This is for displaying the service names. This function is only used in the frontend, because Recorder which is in the Kernel cannot and should not have access to SM in the System.
Pretty much the same as LLE requests, the 'translate' part is chosen. A function is added to the context class to record requests that involves unimplemented HLE functions.
The 'translate' function is a great place to put this in IMO as it is possible to get both untranslated and translated cmdbufs. However a kernel reference has to be passed here, but it is not too hard fortunately.
This class resides in Kernel mainly because that, it's hard for kernel objects to get references to the System (and therefore to the Recorder), while much easier for KernelSystem. If this is to be moved to System the code will likely get much more complex with System (or Recorder) references passed everywhere.
This is for displaying the function name for HLE requests. Probably it is possible to do the same for LLE ones but it would require having the HLE handlers available even when not using them, which doesn't seem to make sense and is more of a hack than a proper solution in my opinion.
Added a new state amiibo_in_range. This state is akin to the real world
physical relationship between a 3DS machine and an amiibo, which is
independent from the service state (or even the machine is powered on or
not). The service state nfc_tag_state is then synchronized with this
physical state on every potential point when the state changes. This
solves the issue where user might load an amiibo before NFC service
initializes, or remove an amiibo after NFC service shutdown, which
previously causes inconsistent state change.
Also removed std::atomic on nfc_tag_state, because
1. It is already protected by g_hle_lock
2. It wasn't properly used in the code anyway. For example, there are
many double loading on this variable, which effectively make it
non-atomic.
Previously, telemetry results couldn't give a good estimate for
performance over time, because it didn't include any fields related to
performance. With this, devs should be able to query metabase for mean
frametime to check for performance regressions after a change is made.
For better tracking of performance regressions on incoming changes, this
change adds a way to dump frametime to file by changing an ini config
option. This is intentionally hidden as its only useful to a small
number of individuals, and not really applicable to the general
userbase.
* Add Anaglyph 3D
Change 3D slider in-game
Change shaders while game is running
Move shader loading into function
Disable 3D slider setting when stereoscopy is off
The rest of the shaders
Address review issues
Documentation and minor fixups
Forgot clang-format
Fix shader release on SDL2-software rendering
Remove unnecessary state changes
Respect 3D factor setting regardless of stereoscopic rendering
Improve shader resolution passing
Minor setting-related improvements
Add option to toggle texture filtering
Rebase fixes
* One final clang-format
* Fix OpenGL problems
Require and assume that override exheaders are decrypted for
consistency with Luma's loader behaviour and to ensure consistent
behaviour regardless of whether the NCCH is marked as encrypted or not.
Currently, exheader overriding with an encrypted NCCH would cause
the title ID checking heuristic to mistakenly disable encryption,
which would then make exefs loading fail.
* yuzu/bootmanager: Remove unnecessary pointer casts
We can just invoke these functions by qualifying the object name before
the function.
* yuzu/bootmanager: unsigned -> u32
Same thing (for platforms we support), less reading.
* yuzu/bootmanager: Default EmuThread's destructor in the cpp file
This class contains non-trivial members, so we should default the
destructor's definition within the cpp file.
* yuzu/bootmanager: Treat the resolution factor as a u32
Treating it as a u16 can result in a sign-conversion warning when
performing arithmetic with it, as u16 promotes to an int when aritmetic
is performed on it, not unsigned int.
This also makes the interface more uniform, as the layout interface now
operates on u32 across the board.
* yuzu/bootmanager: Log out screenshot destination path
We can make this message more meaningful by indicating the location the
screenshot has been saved to. We can also log out whenever a screenshot
could not be saved (e.g. due to filesystem permissions or some other
reason).
* Fix compilation
This allows an IPS patch to edit .bss. This is useful for game patches
that need to add code, as putting things in .bss allows adding new code
*without* editing .code and thus without having to relocate everything.
Currently, having a replacement code.bin causes Citra to crash as it is
trying to load icon data from code.bin because of a typo. This commit
fixes that typo.
This attempts to fix segfault in some tests where page table is set before initializing cpu core (intended behaviour? might be worth a check...)
see: src/tests/core/arm/arm_test_common.cpp
see: src/tests/core/arm/dyncom/arm_dyncom_vfp_tests.cpp
It will both change the page table in memory and notify the CPU about the change by itself. This way there is no need to call memory.SetCurrentPageTable() when kernel.setCurrentProcess() and the management is kept internally in the kernel
This is performing more work than would otherwise be necessary during
VMManager's destruction. All we actually want to occur in this scenario
is for any allocated memory to be freed, which will happen automatically
as the VMManager instance goes out of scope.
Anything else being done is simply unnecessary work.
This makes cpu_core and memory being completely independent components inside the system, having a simpler and more understandable initialization process
The thread which casues page table changes in memory will be responsible to notify the cpu_core too
Makes the dependency explicit in the TelemetrySession's interface
instead of making it a hidden dependency.
This also revealed a hidden issue with the way the telemetry session was
being initialized. It was attempting to retrieve the app loader and log
out title-specific information. However, this isn't always guaranteed to
be possible.
During the initialization phase, everything is being constructed. It
doesn't mean an actual title has been selected. This is what the Load()
function is for. This potentially results in dead code paths involving
the app loader. Instead, we explicitly add this information when we know
the app loader instance is available.
* common/thread: Remove unused functions
Many of these functions are carried over from Dolphin (where they aren't
used anymore). Given these have no use (and we really shouldn't be
screwing around with OS-specific thread scheduler handling from the
emulator, these can be removed.
The function for setting the thread name is left, however, since it can
have debugging utility usages.
* input_common/sdl: Use a type alias to shorten declaration of GetPollers
Just makes the definitions a little bit more tidy.
* input_common/sdl: Correct return values within implementations of GetPollers()
In both cases, we weren't actually returning anything, which is
undefined behavior.
* yuzu/debugger/graphics_surface: Fill in missing surface format listings
Fills in the missing surface types that were marked as unknown. The
order corresponds with the TextureFormat enum within
video_core/texture.h.
We also don't need to all of these strings as translatable (only the
first string, as it's an English word).
* yuzu/debugger/graphics_surface: Clean up connection overload deduction
We can utilize qOverload with the signal connections to make the
function deducing a little less ugly.
* yuzu/debugger/graphics_surface: Tidy up SaveSurface
- Use QStringLiteral where applicable.
- Use const where applicable
- Remove unnecessary precondition check (we already assert the pixbuf
being non null)
* yuzu/debugger/graphics_surface: Display error messages for file I/O errors
* core: Add missing override specifiers where applicable
Applies the override specifier where applicable. In the case of
destructors that are defaulted in their definition, they can
simply be removed.
This also removes the unnecessary inclusions being done in audin_u and
audrec_u, given their close proximity.
* kernel/thread: Make parameter of GetWaitObjectIndex() const qualified
The pointed to member is never actually modified, so it can be made
const.
* kernel/thread: Avoid sign conversion within GetCommandBufferAddress()
Previously this was performing a u64 + int sign conversion. When dealing
with addresses, we should generally be keeping the arithmetic in the
same signedness type.
This also gets rid of the static lifetime of the constant, as there's no
need to make a trivial type like this potentially live for the entire
duration of the program.
* kernel/codeset: Make CodeSet's memory data member a regular std::vector
The use of a shared_ptr is an implementation detail of the VMManager
itself when mapping memory. Because of that, we shouldn't require all
users of the CodeSet to have to allocate the shared_ptr ahead of time.
It's intended that CodeSet simply pass in the required direct data, and
that the memory manager takes care of it from that point on.
This means we just do the shared pointer allocation in a single place,
when loading modules, as opposed to in each loader.
* kernel/wait_object: Make ShouldWait() take thread members by pointer-to-const
Given this is intended as a querying function, it doesn't make sense to
allow the implementer to modify the state of the given thread.
* Set accepted EULA version to max value
CFG: write the max value of 0x7F7F to the default cfg savegame and
auto update on init
CECD: Actually read the EULA version from CFG
* HTTP_C: Implement SetClientCertContext and GetSSLError; Stubbed BeginRequest and BeginRequestAsync
* HTTP_C: Move logs to beginning of function calls
Allows updating the credentials of the announce session, thus allowing credentials changes to be reflected before citra restart. To avoid race conditions and web errors (you can only update the room that you created, i.e. changing credentials halfway will make it break), now you can only use the Citra Web Services settings when not hosting a public room.
The `Register()` function can now handle error results and the error will be passed immediately to the Qt frontend, instead of being ignored silently and failing later with a "Room is not registered".
The backend is not used until we decide to submit the testcase/telemetry, and creating it early prevents users from updating the credentials properly while the games are running.
Keeps the return type consistent with the function name. While we're at
it, we can also reduce the amount of boilerplate involved with handling
these by using structured bindings.
Some objects declare their handle type as const, while others declare it
as constexpr. This makes the const ones constexpr for consistency, and
prevent unexpected compilation errors if these happen to be attempted to be
used within a constexpr context.
These can just be passed regularly, now that we use fmt instead of our
old logging system.
While we're at it, make the parameters to MakeFunctionString
std::string_views.
Since C++17, the introduction of deduction guides for locking facilities
means that we no longer need to hardcode the mutex type into the locks
themselves, making it easier to switch mutex types, should it ever be
necessary in the future.
This reduces the boilerplate that services have to write out the current thread explicitly. Using current thread instead of client thread is also semantically incorrect, and will be a problem when we implement multicore (at which time there will be multiple current threads)
This causes a reference cycle because ServerPort also holds a shared pointer to SessionRequestHandler (inherited by ServiceFrameworkBase). Given that the member port is never used in ServiceFrameworkBase, we can simply remove it. The port object is kept alive by ServiceManager|KernelSystem::named_ports -> ClientPort -> ServerPort
Services can hold kernel objects and do cleanup upon destruction, so we need to keep the kernel alive longer. The new order approximnately resembles the reverse construction order. I will revisit the ordering issue and make it less error-prone after global state cleanup
When making the initial implementation, I forgot to add the series variable to the AmiiboConfig struct.
With this PR it is added and many of the AmiiboConfig fields get their proper values now.
The loading of the Amiibo data that is added here has been hwtested.
This fixes Amiibos in Yoshis Woolly World, Smash (partially) and probably other games too.
* gdbstub: fix IsMemoryBreak() returning false while connected to client
As a result, the only existing codepath for a memory watchpoint hit to break into GDB (InterpeterMainLoop, GDB_BP_CHECK, ARMul_State::RecordBreak) is finally taken,
which exposes incorrect logic* in both RecordBreak and ServeBreak.
* a blank BreakpointAddress structure is passed, which sets r15 (PC) to NULL
* gdbstub: DynCom: default-initialize two members/vars used in conditionals
* gdbstub: DynCom: don't record memory watchpoint hits via RecordBreak()
For now, instead check for GDBStub::IsMemoryBreak() in InterpreterMainLoop and ServeBreak.
Fixes PC being set to a stale/unhit breakpoint address (often zero) when a memory watchpoint (rwatch, watch, awatch) is handled in ServeBreak() and generates a GDB trap.
Reasons for removing a call to RecordBreak() for memory watchpoints:
* The``breakpoint_data`` we pass is typed Execute or None. It describes the predicted next code breakpoint hit relative to PC;
* GDBStub::IsMemoryBreak() returns true if a recent Read/Write operation hit a watchpoint. It doesn't specify which in return, nor does it trace it anywhere. Thus, the only data we could give RecordBreak() is a placeholder BreakpointAddress at offset NULL and type Access. I found the idea silly, compared to simply relying on GDBStub::IsMemoryBreak().
There is currently no measure in the code that remembers the addresses (and types) of any watchpoints that were hit by an instruction, in order to send them to GDB as "extended stop information."
I'm considering an implementation for this.
* gdbstub: Change an ASSERT to DEBUG_ASSERT
I have never seen the (Reg[15] == last_bkpt.address) assert fail in practice, even after several weeks of (locally) developping various branches around GDB. Only leave it inside Debug builds.
udp_server might not be created due to error (occupied port etc.), in which case its destructor and thread-ending call chain will not be excuted in Server::Stop. However, the ending packet still need to be send no matter udp is on or not, so move it to Server::Stop
The necessity of this parameter is dubious at best, and in 2019 probably
offers completely negligible savings as opposed to just leaving this
enabled. This removes it and simplifies the overall interface.
The comment already invalidates itself: neither MMIO nor rasterizer cache belongsHLE kernel state. This mutex has a too large scope if MMIO or cache is included, which is prone to dead lock when multiple thread acquires these resource at the same time. If necessary, each MMIO component or rasterizer should have their own lock.
To eliminate System::GetInstance usage. Archive type like SelfNCCH and SaveData changes the actual reference path for different client, so archive backend interface should accept client information from the service interface. Currently we only pass the program ID as the client information.
Added a few interfaces for adding/deleting/replacing/saving cheats. The cheats list is guarded by a std::shared_mutex, and would only need a exclusive lock when it's being updated.
I marked the `Execute` function as `const` to avoid accidentally changing the internal state of the cheat on execution, so that execution can be considered a "read" operation which only needs a shared lock.
Whether a cheat is enabled or not is now saved by a special comment line `*citra_enabled`.
To break a circular reference formed by process->handle_table->shared_memory->process. Since SharedMemory uses its owner process in the destructor, which is not kept alive by SharedMemory any more, we need to make sure that the lifetime of process is longer than the shared memory. To partially resolve this, Process now explicitly releases shared memory first in its destructor. This is with the assumtion that there is no inter-process reference to shared memory on exit, which is not true when we introduce more multi-process emulation. A TODO is left there for this, as more RE needs to be done on how 3DS handles this situation
This commit it automatically generated by command in zsh:
sed -i -- 's/BitField<\(.*\)_le>/BitField<\1>/g' **/*(D.)
BitField is now aware to endianness and default to little endian. It expects a value representation type without storage specification for its template parameter.
The V-Sync option is fundamentally broken in Citra, so let's do the same as yuzu and remove it entirely for SDL2 and at least from the frontend for QT.
(It was also only used by 7.3% of users)
std::make_unique for arrays is equivalent to doing:
std::unique_ptr<T>(new typename std::remove_extent<T>::type[size]())
(note the ending () after the array size specifier). This means that the
default value within memory for the constructed types will be whatever
the default constructor for that type does. Given the built-in
type for std::uint8_t doesn't have a constructor, this is equivalent to
forcing zero-initialization, so the memory will already be zeroed out on
construction. Because of that, there's no need to zero it out again.
