This adds a check when creating a `TextEdit` to return `nil` if both the fixit and the replacement being used to create a `TextEdit` are empty. A `TextEdit` being created like this can cause a crash.
This allows sourcekit-lsp to make use of the path remappings recently added to
the index store and IndexStoreDB to remap remote paths into local paths
when loading index data locally.
These remappings can be provided via the `-index-prefix-map` command line flag to sourcekit-lsp or via the `BuildSystem` integration point.
(cherry picked from commit 472a06c88a)
- Use official textDocument/inlayHint request
- Rename InlayHintCategory to InlayHintKind
- Additionally, represent it using an Int, as in the proposed LSP API.
- Add inlay hint client capabilities
- Add inlay hint server capabilities
- Add dynamic registration of inlay hint request
- Rename InlayHintsRequest -> InlayHintRequest
This is to be consistent with the request itself being named in singular
in LSP and the other requests (e.g. DocumentSymbolRequest).
- Forward inlay hint requests to clangd
- Add colon before inlay hints
- Add other properties to InlayHint
- Add InlayHintLabel structures
- Conform InlayHintLabel to ExpressibleByStringX protocols
- Attach TextEdit to inlay hints for committing them
- Add InlayHint.data
- Fix InlayHintTests
We need to include text edits in the expected inlay hints.
We only open a document e.g. with the `FoldingRanges:` prefix to get the document structure as a one-off task, so we should be closing it afterwards.
Fixes#553 (rdar://93154201)
A didChangeWorkspaceFolders notification might cause documents that are already open to now be considered part of a different workspace. To reflect this, close documents whose workspace has changed in the old workspace and re-open them in the new workspace.
rdar://94026341 [#557]
Determining which workspace a document belongs to is potentially non-trivial (although it shouldn’t be too expensive either). But since it’s executed for every request, cache it.
This sets the stage for multi-workspace support. Everything should be handled internally, we are just missing the API to actually open multiple workspaces at this point.
This makes SourceKitServer keep track of multiple workspaces and their handling. It does not include the functionality to determine which workspace a file belongs to.
This makes SourceKitServer keep track of multiple workspaces and their handling. It does not include the functionality to determine which workspace a file belongs to.
When we introduce multiple workspaces, they need to share the same DocumentManager so we don’t accidentally open a Swift file twice in two workspaces that share the same `SwiftLanguageService`. Thus, `SourceKitServer` needs to start owning the document manager.
When we support multiple workspaces on one SourceKit-LSP instance, there won’t be a unique workspace anymore. Each request handler should determine the workspace it wants to use on its own.
Similar to how we reload the package if Package.swift is changed, we also watch `compilation_database.json` and `compile_flags.txt` and notify the build system delegate that build settings changed if these files are modified.
rdar://92388223 [#486]
#481 replaced SourceKit placeholders (like `<#code#>`) by LSP snippets for code actions, which are used to represent SourceKit refactoring and diagnostic Fix-Its. But LSP snippets are only supported for code completion, not for code actions. This results in LSP snippets like ${1:code} being inserted into the editor.
Instead of replacing SourceKit placeholders by LSP placeholders, we should just remove them altogether from the code action response.
rdar://92447079 [#488]
Previously, we would insert `<#code#>` placeholders when inserting missing protocol requirements (both via the Fix-It and the refactor command). Translate them to LSP placeholders before inserting them.
rdar://92161144
Workaround for SR-16097 to avoid hitting SR-16095: VSCode will send shutdown requests without a 'params' key. On macOS getting the superDecoder for a non-existent container returns an empty decoder, on Linux it throws `DecodingError.keyNotFound`, causing a crash.
Perform a targeted fix: If we don't have 'params' and the request is a ShutdownRequest, manually create a `ShutdownRequest` without any parameters.
rdar://91288093
This adjusts the sourcekit-lsp build to use static linking for the
internal libraries. It is not currently possible to build
SourceKitLSP.dll as that requires re-exporting the interfaces from the
consumed modules. However, this allows us to reduce the overall size of
the distribution of SourceKit-LSP by ~1 MiB and reduces the
distributed file set. The values here assume partial static linking of
swift-package-manager, which helps reduce the total size.
Before:
228,352 BuildServerProtocol.dll
1,773,056 LanguageServerProtocol.dll
114,688 LanguageServerProtocolJSONRPC.dll
49,152 LSPLogging.dll
262,656 SKCore.dll
54,784 SKSupport.dll
80,896 SKSwiftPMWorkspace.dll
150,528 SourceKitD.dll
645,632 SourceKitLSP.dll
70,144 sourcekit-lsp.exe
3,429,888 bytes
After:
2,416,640 sourcekit-lsp.exe
2,416,640 bytes
This will add an additional link request for dispatch and Foundation
libraries. These are really required on non-Darwin targets, and should
be satisfied either by the library search path or by explicitly
indicating where the dependencies can be found.
The swift project uses a minimum of CMake 3.19.6, and the current CMake
release is 3.22. Update the CMake requirement to a newer version and
remove the workaround for older releases of CMake.
Especially in test we would sometimes end up in situations where the SourceKitServer (i.e. the main sourcekit-lsp handler) is being deallocated before clangd has finished shutting down. This can result in clangd sending us notifications while SourceKitServer is halfway deallocated and thus cause crashes due to memory corruption.
To fix the issue, wait until all toolchain servers have shut down before sending a response to the ShutdownRequest.
* Pass in our custom toolchain
* Remove custom mirror config; swiftpm's default should now match our
specified location.
* Remove custom manifest loader; swiftpm's default uses our toolchain
and should work now.
This explicitly identifies the TSCUtility interfaces that SourceKit
depends on. This helps identify a "burn down list" of interfaces that
remain in TSC(Utility) which are in use. As these interfaces are
replaced, we can easily monitor the remaining interfaces that are in
use.
This reduces the dependency on the `TSCLibc` module which was not
actually used for anything but for importing the C library in
`SKSupport`. Although on Apple platforms and Linux this could largely
be dealt with by statically linking, Windows does not support static
linking of Swift content properly yet. This avoids the dependency on
the library instead.
If the diagnostic has a range associated with it that starts at the same location as the diagnostics position, we can use it to retrieve a proper range for the diagnostic, instead of just reporting a zero-length range.
In certain race condition scenarios, we might receive a document update to which the 0,0 replacetext request returned only syntactic information. In that case we shouldn’t update the sematnic syntax highlighting tokens.
The race condition rarely surfaced as a test failure of `testSemanticMultiEdit`.
