The diagnostic group documentation now point to the swift.org URL rather
than the toolchain path, so it no longer needs to be passed all the way
through sourcekitd.
Resolves rdar://151500502.
(cherry picked from commit 381684a389)
We need this option for `collectVariableType` (aka inlay type hints) but since I’m at it, I’m adding an option to disable the implicit request cancellation for all requests that have it since we don’t want it in LSP at all.
Prerequisite to fixing https://github.com/swiftlang/sourcekit-lsp/issues/2021 / rdar://145871554, need to adopt this option in SourceKit-LSP.
This silences a large number of warnings due to
`-Winconsistent-dllimport`. While it is possible to support the library
to be built statically, that is not currently supported by the build
system, so simply leave that unsupported for decoration.
An option here for ELF targets would be to use
`__attribute__((__visibility__("default")))` and enable hidden
visibility by default enabling a small bit of optimization.
This allows us to load plugins into the sourcekitd service to handle requests (currently only used for code completion). This allows us to implement parts of sourcekitd in Swift and outside of the compiler repository, making it easier to iterated on them because the compiler doesn’t need to be rebuilt.
Introduces the new DeclarationsArrayBuilder and adds it to the
EditorConsumer. Declaration info always includes a kind, offset, and
length, and includes a USR where applicable.
As the USR is already available for editor.open.interface type requests,
this doesn't compute any new information, it just exposes more of what's
there already.
Building the generated interface for WinSDK can overflow the stack.
Treat it as a semantic request to run it on a large stack.
Fixesswiftlang/sourcekit-lsp#1115
rdar://123944504
`OptTable` was a source of consistent churn due to new arguments to the
`OPTION` macro. LLVM 3f092f37b7362447cbb13f5502dae4bdd5762afd extracted
the handling of the common option parts (eg. an ID and an info) out into
separate macros to reduce this - use those here (since unsurprisingly,
more arguments were added).
LLVM is presumably moving towards `std::string_view` -
`StringRef::startswith` is deprecated on tip. `SmallString::startswith`
was just renamed there (maybe with some small deprecation inbetween, but
if so, we've missed it).
The `SmallString::startswith` references were moved to
`.str().starts_with()`, rather than adding the `starts_with` on
`stable/20230725` as we only had a few of them. Open to switching that
over if anyone feels strongly though.
We were only setting `RequestBarriersEnabled` after dispatching the request to `requestQueue`. This left a brief gap where new requests could come in that wouldn’t see `RequestBarriersEnabled = true`.
SourceKit-LSP currently parses the XML comment to generate Markdown again but round-tripping a (probably markdown) doc comment to XML to Markdown is lossy in many cases and unnecessary work. Include the comment as it is spelled in source in the cursor info response so that sourcekit-lsp can display it.
Part of rdar://120685874
Sending a `SIGTERM` to sourcekitd XPC service lead to very surprising results: XPC thought that the process was about to die and sends a Connection Interrupted notification to the client. But other than that, we didn’t actually handle the signal and the service stayed alive. This caused us to send a `sema_disabled` notification to the client but since the service never got relaunched, we never sent the corresponding `sema_enabled` notification.
To fix this, just properly die on `SIGTERM` by calling `abort`.
Performing a related identifiers request on `self` returns an empty result because `self` is not an identifier. We thus can’t retrieve the first location to compute the old name.
rdar://121668042
While we don't want to return non-editable base names for Edit-All-In-Scope, we do want to return them if the related identifiers response is used as the input for find-syntactic-rename-ranges.
