The URI standard RFC 3986 is ambiguous about whether percent encoding and their represented characters are considered equivalent. VS Code considers them equivalent and treats them the same:
```js
vscode.Uri.parse("x://a?b=xxxx%3Dyyyy").toString() -> 'x://a?b%3Dxxxx%3Dyyyy'
vscode.Uri.parse("x://a?b=xxxx%3Dyyyy").toString(/*skipEncoding=*/true) -> 'x://a?b=xxxx=yyyy'
```
This causes issues because SourceKit-LSP's macro expansion URLs encoded by URLComponents use `=` do denote the separation of a key and a value in the outer query. The value of the `parent` key may itself contain query items, which use the escaped form '%3D'. Simplified, such a URL may look like `scheme://host?parent=scheme://host?line%3D2`.
But after running this through VS Code's URI type `=` and `%3D` get canonicalized and are indistinguishable.
To avoid this ambiguity, always percent escape the characters we use to distinguish URL query parameters, producing the following URL: `scheme://host?parent%3Dscheme://host%3Fline%253D2`.
Now that all the types that model test projects are suffixed `Project` instead of `Workspace`, the `ws` abbreviation doesn‘t make sense. It also never really fit with the new style of avoiding abbreviations. Rename all occurrences to `project`.
rdar://124727401
Not entirely sure what introduced the race condition. Found by thread sanitizer. Probably wouldn’t have been an issue if we had migrated this code to strict concurrency checking.
Basically everything will be decoded as ClangWorkspaceSettings, even if it has nothing to
do wit hit, yielding for example:
```
clangd(LanguageServerProtocol.ClangWorkspaceSettings(compilationDatabasePath: nil, compilationDatabaseChanges: nil))
```
This breaks at least my case of using WorkspaceSettingsChange. This patch adds a - maybe bandaid - fix for that.
Furthermore I removed one Test, as it - AFAIU - broke the invariant that none of the attributes
of ClangWorkspaceSettings may be nil.
Add `.swift-format` to the repo and format the repo with `swift-format`.
This commit does not add any automation to enforce formatting of sourcekit-lsp in CI. The goal of this commit is to get the majority of source changes out of the way so that the diff of actually enforcing formatting will have fewer changes or conflicts.
Instead of having ad-hoc timeout durations in all the test cases, specify a default timeout duration that can be used by tests.
This allows us increase the timeout duration for all tests if we discover that e.g. sourcekitd is slower in CI setups.
rdar://91615376
This appears to have been accidentally removed in
17f656865d, the `children` field is
optional in the LSP spec.
Change-Id: I1055658c8f09873279d676fee0daff02fa59e3f8
* Add LSP types and server stubs for call hierarchy support
* Update CMakeLists.txt for new LSP types
* Call hierarchy misc fixes
* Minor PositionRangeArray fixes to simplify codable conformance
* Add CodingTests for PositionRangeArray and CallHierarchy encoding
* Additions to the LSP module for `workspace/didChangeWatchedFiles`
We can then use this functionality to allow a `BuildSystem` to watch
files (e.g. via the client or even in-process file watching),
which would allow us to do things like detect new files and provide
accurate build settings for them.
* Improve #file and #line for LSP test errors
Also regenerate Linux main
* Add `LSPAnyCodable` protocol requirement to RegistrationOptions
* Don't encode an optional documentSelector to null
* Skip running `testSourcekitdCrashRecovery` when not on macOS
While fallback arguments are being used (either from the fallback build system or fallback settings from the primary build system), withhold semantic diagnostics from sourcekitd and all diagnostics from clangd. This helps prevent user confusion from spurious errors.
- Also remove the DocumentURI standardization in favor of proper equality + hash checks to work around the %40 --> @ encoding issue seen on CI.
* Support for clangd's go-to-definition for header files
- By forwarding the request to clangd when it fails to give
symbol information, we are able to use its built in
go-to-definition support for headers (jump to header file)
* Add static tibs test for clangd go-to-#include
* Move #include test to SourceKitTests and regenerate linux main
* Fix improper escaping of %40 in file URLs
* Add URL escaping test
* Another attempt to fix broken BuildServerBuildSystemTests test on Linux
- URL's `standardizedFileURL` removes trailing slashes from file URLs,
but only on Linux
The LSP spec requires us to omit keys for nil values rather than using
the JSON `null` constant in most places. This change to CustomCodable
allows us to do it automatically using CustomCodable, removing one of
the limitations of the property wrapper.
The idea for the adding overloads of `encode` and `decode` came from
https://forums.swift.org/t/pre-pitch-codable-customization-using-propertywrappers/30244/
URL can in fact store URIs, it just doesn't have a very nice API to
interact with them. As long as we only operate on absoluteString, we
should be fine though. So instead of implementing the logic for
detecting file URLs ourselves, we can just use a URL as storage for
DocumentURI.
According to the LSP specification, arbitrary URIs can be used as
document identifiers. Instead of internally assuming that all URIs are
URLs, use a DocumentURI enum to represent URIs. These can either be file
URLs or other URIs whose value as treated as an opaque string.
We will be able to split the LSP modules off later. These LSP modules
will provide the ability to write custom LSP servers and clients in
Swift. The sourcekit-lsp repository will build on top of this new
package to provide an LSP implementation that creates a language server
for Swift and C-based-languages.
This lets us use the `Range<Position>` without manually converting back
and forth in the majority of cases. See the note on HoverResponse for
the behaviour with Optional that makes us have to provide a custom
Codable implementation in two places that used to be synthesized.
