| Age | Commit message (Collapse) | Author |
|---|
|
|
|
This avoids expensive copying when returning a value from the map in `SharedCache::GetExportListForHeader`. Additionally it ensures that the value stays alive and at the same location in memory if `m_exportInfos` is modified and requires its storage to be re-allocated.
I was unable to use a `unique_ptr` instead of a `shared_ptr` because of copy semantics with `m_exportInfos` in `ViewStateCacheStore`. I don't see things being any worse using `shared_ptr` instead of `unique_ptr` anyway and it means less code changes.
|
|
`SharedCache::FindSymbolAtAddrAndApplyToAddr` if a symbol is found
|
|
This commit changes 2 things;
1. `m_exportInfos` is now a map where its values are also a map rather than a vector of pairs. The reason for this is that `SharedCache::FindSymbolAtAddrAndApplyToAddr` is a hot path which does by far the most accesses to `m_exportInfos`. In that function it must find the correct symbol for a given address so a map lookup will be much quicker than iterating a vector. The other use cases of `m_exportInfos` would prefer a vector but they are executed very infrequently.
2. The symbols are stored in `m_exportInfos` as references to the `Symbol` type. This makes more sense because otherwise there is a lot of time spent converting to and from a `Symbol` type and a pair of `BNSymbolType` + a `std::string`.
|
|
`m_exportInfos` was modified
This probably makes more sense than the current solution of using execution of the callback parameter to determine if `m_exportInfos` was modified.
|
|
`SharedCache::ParseExportTrie` is getting called a lot during DSC library loading and analysis. In large part due to the hot path `SharedCache::FindSymbolAtAddrAndApplyToAddr`. Its unnecessary for it to be being called more than once per DSC header as the export list symbol information is stored in `SharedCache::m_exportInfos`.
This commit adds the function `SharedCache::GetExportListForHeader`, which will either return the header's list of symbol information cached in `SharedCache::m_exportInfos` or call `SharedCache::ParseExportTrie` and cache the results in `SharedCache::m_exportInfos`.
This should also improve the execution time of `SharedCache::LoadAllSymbolsAndWait`.
Further improvement here would be to add locking to `SharedCache::GetExportListForHeader` so that races don't result in redundant parsing of the export trie for the same header if multiple threads call `SharedCache::GetExportListForHeader` at the same time for the same header. This only really matters during initial loading because from what I can tell that parses all the export trie's anyway.
|
|
|
|
|
|
|
|
`ReadExportNode` is called a lot during the initial load of the shared
cache and thus impacts how long it takes for the UI to become
responsive. This is a collection of optimizations that cut the time
spent within `ReadExportNode` by 50%:
1. Pass iterators to `ReadExportNode` rather than a `DataBuffer` + offset.
The lack of inlining in `DataBuffer`'s `operator[]` kills performance.
Ideally this would have used `std::span`, but that would require
bumping the minimum C++ version to C++20.
2. Add `MMappedFileAccessor::ReadSpan` so that `ReadExportNode` can
operate directly on the mapped data without first copying it.
3. Removes a call to `GetAnalysisFunctionsForAddress` whose result was
unused.
4. Use `std::find` to find the nul at the end of strings rather than
assembling the string a character at a time. This avoids repeatedly
growing the string.
5. Avoid the usual `Symbol` constructor in favor of `BNCreateSymbol`.
The `Symbol` constructor has over head in two forms:
1. It has a `NameSpace` as an argument. Creating / destroying this
allocates and deallocates memory We could create a single
instance and reuse it for all calls, but...
2. The constructor copies all fields of the `NameSpace` to the heap
before calling `BNCreateSymbol` and then deallocates them
afterwards. This seems unnecessary, and adds a non-trivial amount
of overhead.
This can go back to directly constructing the `Symbol` once the
constructors are improved.
|
|
view workaround
|
|
Moves a bunch of stuff out of src/types.rs that did not belong:
- Confidence
- Variable
- Function specific stuff
- Refactored InstructionInfo, see the msp430 and riscv examples.
- Renamed Function::from_raw to Function::ref_from_raw and fixed places where the ref was incremented twice
- Fixed FunctionRecognizer leaking functions (see above)
- Fixed some apis incorrectly returning Result where Option is clearer
- Started to move destructured types to the From trait for converting to an from ffi types, see Location for an example
- Started to remove bad module level imports (importing std modules like mem all over the place)
- Moved some wrapper handle types to named handle field (this improves readability), see CoreArchitecture for an example
- Removed some unintuitive getters, this is bad practice for Rust code, just access the field directly, see DataVariable for an example
- General code cleanup, purposely did not run rustfmt, that will be a single seperate commit
More rust cleanup
- Fixed invalid views being able to invoke UB when dealing with databases
- Cleaned up some helper code in dwarf_import
- Fixed inverted is_null checks causing crashes! Oops!
More rust cleanup
Still a WIP, I think branch info is still invalid, need to figure out the issue there.
- Fixed some invalid Ref lifetimes when constructing indirectly, see Array<DataVariable> for example
- Added some more comments
- Removed some "magic" functions like MLIL Function::ssa_variables
There are still a bunch of invalid lifetimes that aren't crashing us due to the usage of those API's not living long enough. But they ARE an issue.
More rust cleanup
Trying to comment more TODO's as I go along.
- Renamed llil::Function to llil::LowLevelILFunction for clarity and consistency
- Take base structures by ref in StructureBuilder::set_base_structures to prevent premature drops
- Added more raw to wrapper conversions
- Removed UB prone apis
- Getting rid of more core module references, use std!
- Removed improper Guard usage in array impls for wrapper types with no context
- Untangling the UB of the Label api, still only half done :/
More rust cleanup
- Misc formatting
- Made Logger ref counted
- Fixed leaking name of logger every time something was logged
- Fixed the last (hopefully) of the unresolved labels
- Simplified some code
Fix leak in DebugInfo::AddType
componentArray was never freed
Add more HLIL related functions to rust
More rust cleanup
improve the CustomBinaryView init process
Canonicalize path in `create_delete_empty` test
Link core in rust
When linking you must depend on the -sys crate.
This is because linker arguments (what says where to find binaryninjacore) are NOT transitive. The top level application crate MUST provide it.
For more information see:
- https://github.com/rust-lang/cargo/issues/9554#issuecomment-857882964
- https://github.com/oxidecomputer/omicron/pull/225
Remove vendored pdb crate
Use cargo to manage the git repo ref instead
Fix misc rustdoc warnings
Move actual plugins out of `rust/examples` and into `plugins`
This is where all shipped public plugins that are not arch/view/platform/lang will be at from now on
Originally they were in the rust workspace, meaning they all shared a Cargo.lock which is ill-advised.
More rust cleanup
- More clarification on plugin/executable requirements
- Made examples actually rust examples
- Add Display impl for InstructionTextToken
- Renamed feature "noexports" to "no_exports"
Move under_construction.png to assets directory
This really did bother me
Remove unneeded `extern crate bindgen`
Replace nop'd log::info with println
We don't register a compatible log sink so they will just get sent into the void
Move inline tests into tests directory
This is being done in the hopes of curbing the multi-thousand line files that will occur once we flesh out the tests
Format rust code
Update rust ci
Still need to add support for running tests in ci
More rust cleanup
- Architecture id's are now typed accordingly
- Fix some clippy lints
- Make instruction index public in LLIL
- Removed useless helper functions
- LLIL expressions and instruction indexes are now typed accordingly
Generate binaryninjacore-sys documentation
This should show binaryninjacore-sys alongside binaryninja crate
More rust cleanup
- Remove lazy_static dependency
- Remove hacky impl Debug for Type and just use the display impl
- Add more debug impls
- Reorder some top level namespace items
- Add first type test
Remove unneeded script helper in rust api
More rust cleanup
- Added main thread handler api
- Register a headless main thread handler by default in initialization
- Refactor QualifiedName to be properly owned
- Loosened some type constraints on some apis involving QualifiedName
- Fixed some apis that were crashing due to incorrect param types
- Removed extern crate cruft for log crate
- Simplified headless initialization using more wrapper apis
- Fixed segments leaking because of no ref wrapper, see BinaryViewExt::segment_at
- Added rstest to manage headless init in unit tests
- Added some more unit tests
- Refactored demangler api to be more ergonomic
- Fixed minidump plugin not building
More rust cleanup
- Fixup usage of QualifiedName in plugins
- Make QualifiedName more usable
Implement rust TypeParser
fix Platform TypeParser related functions
separate User and System implementations of TypeParserResult
Implement rust TypeContainer
More rust cleanup
- Hopefully fixed the rust.yml CI
- Added TypePrinter API (this is still WIP and will crash)
- Added TypeParser API
- Added TypeContainer API
- More work around QualifiedName apis
Oh your suppose to do this
Add workflow_dispatch trigger to rust.yml
More rust fixes
- Swapped some usage of raw 255 to MAX_CONFIDENCE, no one likes magic numbers
- New InstructionTextToken API, complete with owned data, this still needs a lot of testing.
- InstructionTextTokenKind describes a destructured InstructionTextToken, this should make token usage much clearer, some docs pending
- Added some misc Default and Debug impls
- Updated TypePrinter and architectures to use new InstructionTextToken API
Misc formatting changes
More rust cleanup
- Fixed MANY memory leaks (most due to QualifiedName)
- Made StructureBuilder more builder and less structure
- Fixed CMakeLists.txt that were globbing entire api, resulting in 100 second slowdown on cmake generation
- Added more Debug impl's
- Added some more tests
- Fixed TypeParserResult UB
- Moved the From impls to blank impl for clarity, we have multiple different variants of core to rust for some structures, it should be explicit which one you are choosing.
- PossibleValueSet should now be able to allocate so we can go from rust to core with those variants that require allocation
- Misc doc code formatting
Misc clippy lints and clippy CI
Co-authored-by: Michael Krasnitski <michael.krasnitski@gmail.com>
Fix typo in rust CI
Misc rust formatting
Fix misc typos and add typos to rust CI
Add cargo workspace
This will help tooling and external contributors get a map of the rust crates within binaryninja-api
More rust cleanup
- Format all rust plugins
- Fix some tests that were out of date
- Simplify WARP tests to only binaries, building object files from source is a pain
- Link to core in all rust plugins
- Fix some memory leaks
- Update warp insta snapshots
- Fix some misc clippy lints
Run rust tests in CI
This commit also coincides with the creation of the "testing" environment which exposes a BN_SERIAL secret for pulling a headless Binary Ninja
Install missing wayland dependency in github CI
Apparently its needed for linux file picker for the WARP integration
Set the BINARYNINJADIR so rust can find binaryninjacore in CI
The chances of this working are low
Misc remove unused dependency
Rust misc formatting fixes
Improve initialization in rust headless scripts
Provide sensible errors and validation to rust headless scripts, solves https://github.com/Vector35/binaryninja-api/issues/5796
Add BN_LICENSE environment variable to rust CI
We pass the serial to download binary ninja, but we never provided the license for core initialization
Fix typo
More rust cleanup
- Improved binary view initialization (see init_with_opts)
- Allow floating license to free itself before initialization
- Add initialization unit test
- Add better Debug impls for some common types
- Use Path api for opening binary views, this is not breaking as it uses the AsRef impl
- Bump rayon dependency and constrain dependencies to x.x
Update readme and include that directly in the rustdocs
More rust documentation changes
Add format comment to InitializationOptions::with_license
Misc formatting and clippy lint allow
More rust cleanup
- Remove under_construction.png from the build.rs it has been removed
- Use InstructionIndex in more places
- Add missing PartialOrd and Ord impls for id types
More rust cleanup
- Make workflow cloning explicit
- Add workflow tests
- Add missing property string list getting for settings
- Remove IntoActivityName (see https://github.com/Vector35/binaryninja-api/pull/6257)
More rust cleanup
This commit is half done
Misc rust formatting
More rust cleanup
- Renamed common name conflictions (I will put my justification in the PR)
- Fixed invalid instruction retrieval for LLIL
- Added common aliases for llil function, instruction and expression types (see my comment in the PR)
- Refactored the instruction retrieval for LLIL, MLIL and HLIL
- Added instruction index types to MLIL and HLIL
- Moved llil module to lowlevelil module (mlil and hlil will be moved as well)
- Added preliminary LLIL unit testing
Fix typos
Misc clippy fixes
More rust cleanup
- Normalized modules
- Split some code out into own files
- Fixed some UB in type archive and projects
- Improved API around type archives and projects substantially
- Added ProgressExecutor abstraction for functions which have a progress callback
- Improved background task documentation and added unit tests
- Added worker thread api and unit tests
- Moved some owned types to ref types, this is still not complete, but this is the path forward.
- Add external location/library accessors to the binary view
- Added some misc documentation
- Replaced mod.rs with the module name source file
Still need to normalize some paths and also update some documentation surrounding that change.
Update some tests and examples
Fix background task tests colliding
We were creating multiple background tasks with the same progress text on multiple threads
More rust cleanup
- Fixed progress executor freeing itself after one iteration
- Updated the last of the doc imports
- Moved mainthread to main_thread
- Made project creation and opening failable
We could probably AsRef<ProgressExecutor> to get around the allocation and free inside the function bodies, not high priority as those functions are long running anyways.
Move binary view initialization into function body for LLIL test
Normalize test file names
More rust cleanup
- Updated README to clarify offline documentation
- Refactored settings api
- Added settings example to show dumping settings value and specific properties
- Use the workspace to depend on binaryninja and binaryninjacore-sys
- Remove binaryninjacore-sys as a workspace member (its not really required)
Update workflow test to new settings api
More rust cleanup
- Rename Label to LowLevelILLabel
- Update the functions label map automatically
This fixed a major api blunder where the label map is returned as a reference and originally resulted in UB prone lifetime semantics. It was temporarily "fixed" with explicit label updates in the architecture implementation code. But that was less than ideal and was easy to mess up. Now the label map will be updated automatically as the location of labels is now tracked.
Misc clippy lints
More rust cleanup
- Get rid of RawFunctionViewType
- Add better Debug impl for Function
More rust cleanup
- Fixed the documentation icon using local files (thank you @mkrasnitski)
- Fixed labels being updated and overwriting the label location used to update the label map
More rust cleanup
- Added unit tests for MLIL and HLIL
- "Fixed" MLIL, LLIL, and HLIL having issues regarding Instruction vs Expression indexes
- Renamed CallingConvention to CoreCallingConvention and removed architecture generic
- Renamed CallingConventionBase to CallingConvention
- Simplified calling convention code and fixed some bugs with implicit registers
- Added impl Debug to MLIL and HLIL instructions
Still need to at some point add an Expression to MLIL and HLIL. We also might want to look into having the Instruction kind just return the expression kind.
Misc clippy lint
More rust cleanup
- Allow calling conventions to be registered for multiple architectures
- Swapped a unreachable statement to an unimplemented statement
More rust cleanup
- Fixed the issue with PDB types, this has caused me an insane amount of grief
- Fixed LLIL visit_tree missing LLIL_LOAD expressions
- Added LLIL visitor test
- Made all WARP file pickers use the rfd crate
Use the dev branch of Binary Ninja in rust CI
Misc rust fmt
More rust cleanup
- Refactored BinaryReader and BinaryWriter
- Added some warnings to high_level_il and medium_level_il modules that they are unstable
- Add BinaryReader and BinaryWriter tests
- Changed BinaryView len to return u64 (that is what the core returns)
- Misc formatting changes
- Remove extern uses in lib.rs
Add impl Debug for BinaryReader and BinaryWriter
Turn off broken tests
Add more info to the rust README.md
More rust cleanup
- Make EdgeStyle type not wrap raw
- Regression tests for WARP will run on all bins in the out dir now
impl rust Collaboration and Remote API
Fix typo
Update collaboration API
Makes collaboration more in line with the refactor. Still a lot of work to do.
Namely still need:
- Proper errors
- _with_opts functions
- More ergonomic api
- Better connection procedure
- Updated documentation
- A LOT of unit tests
- An example
- Typed id's for everything (i dont want BnString as the id!!!)
- NEED to refactor the progress callbacks into the new progress api, but we should pull in some of the stuff the collab progress has
- Elimination of apis that are dumb helpers
Separate out the rust testing and use pull_request_target
pull_request_target allows PR's to access the headless license, for this to be safe we need to prevent people from running the job.
To prevent the job from being ran we add an environment requirement on testing that a reviewer must review the code and then manually approve it to run.
More rust cleanup
- Use GroupId instead of u64
- Use ProgressCallback in place of ProgressExecutor
- Misc cleanup of FileMetadata
- Add `save_to_path` and `save_to_accessor` to save modified binaries
- Added binary_view unit tests
- Added collaboration unit tests
- Fixed a few issues with the collaboration apis
- Renamed Command registration functions so that there is no import ambiguity
- Split out RemoteUndoEntry
- Collaboration apis now have a explicit `_with_progress` set of apis
- Misc clippy lint fixes
Fix some typos
More rust cleanup
- Add extra info to README.md
- Refactor components api
- Add components unit test
Add testing and documentation to contributing section in README.md
Fix misc doc comments
|
|
|
|
|
|
`PageMapping` was storing the path to the file it points within. This
was causing unnecessary work within `VM::MappingAtAddress` as the
`PageMapping`, and thus the path, is copied into the return value. This
copying was more expensive than the map lookup.
The file path is now stored within a `LazyMappedFileAccessor` class that
wraps the `SelfAllocatingWeakPtr`. This means the path is still
available via the `PageMapping`, but it does not need to be copied as
often.
This includes two additional improvements / optimizations while I was
touching the code in question:
1. `MMappedFileAccessor::Open` no longer performs two hash lookups when
a file accessor already exists.
2. `VM::MapPages` takes the path by const reference to avoid an
unnecessary copy.
|
|
fix compilation issue on linux
|
|
Find the relative selector base address in the Objective-C optimization
data pointed to by the shared cache header, rather than via
`__objc_scoffs`. This is only present on iOS, and not for every iOS
version that encodes selectors via direct offsets.
This also includes some related improvements:
1. Direct selectors get their own pointer type so they're rendered
correctly in the view.
2. Method lists encoded as lists of lists are now handled.
3. The `dyld_cache_header` type added to the view is truncated to the
length in the loaded cache. This ensures it is applied to the view.
4. A couple of methods that process method IMPs and selectors are
updated to check whether the address is valid before attempting to
process them. They would otherwise fail by throwing an exception if
they proceed, but checking for validity is quicker and makes
exception breakpoints usable.
|
|
`BackingCache` now tracks the `dyld_cache_mapping_info` for its mappings
so it has access to the memory protections for the region. This means it
can avoid marking some regions as containing code when they don't,
reducing the amount of analysis work that has to be done.
Using `dyld_cache_mapping_info` also makes references to mappings easier
to understand due to its named fields vs the nested `std::pair`s that
were previously in use.
|
|
They're surprisingly expensive to look up.
|
|
The existing view-specific state was stored in several global unordered
maps. Many of these were accessed without locking, including
`viewSpecificMutexes`, which is racy in the face of multiple threads.
View-specific state is stored in a new heap-allocated
`ViewSpecificState` struct that is reference counted via
`std::shared_ptr`. A static map holds a `std::weak_ptr` to each
view-specific state, keyed by session id. `SharedCache` retrieves its
view-specific state during its constructor.
Since `ViewSpecificState` is reference counted it will naturally be
deallocated when the last `SharedCache` instance that references it goes
away. Its corresponding entry will remain in the static map, though
since it only holds a `std::weak_ptr` rather than any state it will not
use much memory. The next time view-specific state is retrieved any
expired entries will be removed from the map.
|
|
`BNDSCViewGetAllImages`
|
|
A problem with processing Objective-C sections at the time a library is loaded is that some of the references within those sections may refer to unload sections. This results in things like selectors not being correctly typed and named.
This commit provides a way for users to manually trigger Objective-C parsing against sections for a specific library or all libraries, via the API. Combined with the previous commit a user can use the API to batch load a number of libraries and skip Objective-C processing for each one and then run it across the entire of the DSC once they are all loaded.
Further improvement would be to provide a way to trigger Objective-C processing through the UI.
|
|
library
This is useful when batch loading libraries to avoid extra processing (once the next commit has landed).
|
|
Prior to this commit the function `DSCObjCProcessor::PostProcessObjCSections` never does anything because it doesn't use the correct names to get the Objective-C sections of the recently loaded library. In fact it never does anything because the DSC never has sections with the names its searching for.
This commit passes the `baseName` (the name of the library that was loaded), which is what other Objective-C section processing code does. Combining the base name with the section names it will now find them and process them as intended. This was resulting in alot of Objective-C related stuff being missed.
There is however still an issue of the fact that the way this DSC plugin works means it only analyzes Objective-C sections once. This catches alot of things but there are a number of cases where other libraries need to be loaded first due to information being referenced in another library. For instance errors like `Failed to determine base classname for category` can be caused by the class reference in the category being to a class outside of the loaded library. Once the library containing the class has been loaded, the section containing the category should be re-proccessed.
|
|
From what I can tell the lock being taken in `SharedCache::~SharedCache` was purely for the decrement of `sharedCacheReferences`, however its an atomic so a lock isn't necessary. The lock being taken is extremely contentious and therefore often slow to be acquired. This resulted in a surprising amount of execution time spent in the `SharedCache` destructor. Nothing hugely significant but a quick and easy win to remove this single line of code.
|
|
|
|
symbol list
We should think through a better way of handling upgrades, as v3 can load v2 dbs just fine, but there is no clean way to upgrade the info currently. We need SharedCache ser/des functions in DSCView.cpp
|
|
`SharedCache::m_symbolInfos` isn't being serialized but there is an attempt to deserialize it. This commit adds in the code to serialize it.
I slightly modified the format because I didn't really understand how it was expected to be serialized based on the deserialization code. The deserialization code looked wrong to me but its likely a misunderstanding on my part. I kept it similar to how `m_exportInfos` is serialized.
|
|
Allows language bindings like rust to free conditions lists sanely
|
|
|
|
There seem to be a number of issues with slide info parsing. I decided the simplest fix was to largely mimick what `dyld` is doing.
Part of this process was to remove creating a vector of page starts and processing them in another loop below the one where they were read out. It wasn't clear to me the original design decision to separate it into 2 loops like that. I think this was part of the problem that was causing issues.
By adding rewrites in the same loop where page starts are being read out, it was much easier to mimick the code in `dyld` which I assume has to be correct. So as long as my copying was correct then I believe this should work as intended.
Not everything has been thoroughly tested but I'm pretty confident v3 and v5 are now working as intended. v2 should be but less testing of it has been done.
|
|
The issue this commit fixes was causing `SharedCache::ParseAndApplySlideInfoForFile` to completely fail to work with v5 slide info, which had a lot of knock on effects, i.e. lots of Objective-C analysis was failing due to invalid pointers which hadn't been fixed up.
`dyld_cache_slide_info5` has 4 bytes of padding before `value_add`. Whilst `value_add` is not actually being read from, `SharedCache::ParseAndApplySlideInfoForFile` will read at a location in the file based on the size of the structure `dyld_cache_slide_info5`. This being off by 4 bytes basically broke v5 slide info fixups.
With this fix many more Objective-C functions have names and a lot more `msgSend` calls are fixed up.
|
|
|
|
Copying the state from the cache into a new `SharedCache` object is done
with a global lock held and is so expensive that it results in much of
the shared cache analysis running on a single thread, with others
blocked waiting to acquire the lock.
The cache now holds a `std::shared_ptr` to the state. New `SharedCache`
objects take a reference to the cached state and only create their own
copy of it the first time they perform an operation that would mutate
it. The cached copy is never mutated, only replaced, so there is no
danger of modifying the state out from under a `SharedCache` object.
Since the copy happens at first mutation, it is performed without any
global locks held. This avoids blocking other threads.
This cuts the initial load time of a macOS shared cache from 3 minutes
to 70 seconds, and cuts the time taken to load and analyze AppKit from
multiple hours to around 14 minutes.
|
|
The flags field was being left uninitialized which could result in the
region being mishandled during later analysis.
|
|
There are typically only a few dozen mappings, while there can be
millions of pages. This reduces the amount of time spent populating the
mapping from region to file accessor along with the memory usage of the
same.
|
|
|
|
api/MetadataSerializable.hpp is removed in favor of including
core/MetadataSerializable.hpp. Both headers defined types with the same
name leading to One Definition Rule violations and surprising behavior.
The serialization and deserialization context are now created on-demand
during serialization rather than being a member of
`MetadataSerializable`. This reduces the size of every serializable
object by ~220 bytes.
The context is passed explicitly as an argument to `Serialize` /
`Deserialize`. As a result, `Serialize` / `Deserialize` can now be free
functions rather than member functions.
Since `MetadataSerializable` is not used for dynamic dispatch,
the virtual methods are removed and the class is updated to be a class
template using CRTP. This allows delegating to the derived class's
`Load` and `Store` methods without the additional size overhead of the
vtable pointer in every serializable object.
These changes reduce the memory footprint of Binary Ninja after loading
the macOS shared cache and loading a single dylib from it from 8.3GB to
4.6GB.
|
|
|
|
|
|
|
|
|
|
This is both a correctness fix and avoids leaking a massive amount of memory.
|
|
`MMappedFileAccessor::ReadBuffer` was returning a heap-allocated
`DataBuffer`, but no callers were ever deleting it. There does not
appear to be any reason to heap allocate the `DataBuffer` as the type is
effectively a smart pointer wrapper around `BNDataBuffer`. Switch to
returning it by value instead.
Additionally, `MMappedFileAccessor::ReadBuffer` was allocating a buffer,
copying data into it, and then handing that allocation to the `DataBuffer`
constructor. The constructor copies data into a new allocation it
owns so this allocation is unnecessary and was being leaked.
|
|
|
|
|
|
|
|
|
|
|
|
|
