Update gopls to handle the new form of "undeclared name: ..." error
messages, "undefined: ...", and update tests to be tolerant of both.
Also do some minor cleanup of test error messages related to mismatching
diagnostics.
With this change, TryBots should succeed at CL 432556.
Updates golang/go#54845
Change-Id: I9214d00c59110cd34470845b72d3e2f5e73291c1
Reviewed-on: https://go-review.googlesource.com/c/tools/+/434636
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
The documentSymbols handler joined syntax information with type
information, meaning that it was only able to satisfy requests for files
in valid workspace packages. However, the value added by type
information was limited, and in many cases could be derived from syntax
alone. For example, while generating symbols for a const declaration, we
don't need the type checker to tell us that the symbol kind is const.
Refactor the documentSymbols handler to derive symbols from syntax
alone. This leads to some simplifications from the code, in addition to
eliminating the dependency on package data. Also, simplify symbol
details to just use types.ExprString, which includes some missing
information such as function return values. Also, update handling to
support Go 1.18 type embedding in interfaces.
Notably, this reverts decisions like golang/go#31202, in which we went to
effort to make the symbol kind more accurate. In my opinion (and the
opinion expressed in golang/go#52797), the cost of requiring type
information is not worth the minor improvement in accuracy of the symbol
kind, which (as far as I know) is only used for UI elements.
To facilitate testing (and start to clean up the test framework), make
several simplifications / improvements to the marker tests:
- simplify the collection of symbol data
- eliminate unused marks
- just use cmp.Diff for comparing results
- allow for arbitrary nesting of symbols.
- remove unnecessary @symbol annotations from workspace_symbol tests --
their data is not used by workspace_symbol handlers
- remove Symbol and WorkspaceSymbol handlers from source_test.go. On
inspection, these handlers were redundant with lsp_test.go.
Notably, the collection and assembly of @symbol annotations is still way
too complicated. It would be much simpler to just have a single golden
file summarizing the entire output, rather than weaving it together from
annotations. However, I realized this too late, and so it will have to
wait for a separate CL.
Fixesgolang/go#52797Fixesgolang/vscode-go#2242
Updates golang/go#54845
Change-Id: I3a2c2d39f59f9d045a6cedf8023ff0c80a69d974
Reviewed-on: https://go-review.googlesource.com/c/tools/+/405254
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
Some of the json tags, especially for constructed structs, had improper
json tags (e.g., Range *Range `json:"Range"`, not `json:"range"). These
now start with lower-case letters.
Also, an optional field which is a struct type could be either
F FsType `json:"f,omitempty"`, or
F *FsType `json:"f,omitempty".
The second is generally better, as zero values should not be sent,
but there are many cases where the existing
code needs the first, to avoid allocations.
That is, if we're sending an empty optional field, there should be a *,
but if we're intializing a complicated structure (x.A.B.C.D = true, with
B and C optional) existing code wants no *. (These two cases could
conflict, and then the existing code would have to be changed.)
Fixesgolang/go#55099
Change-Id: I947a69322bcddf4c963d34662b72b464b738d17c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/431218
Run-TryBot: Peter Weinberger <pjw@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
The new LSP stubs were returning a "disabled" struct in CodeActions,
which is wrong, and this CL fixes.
Tested by hand.
Fixesgolang/go#55080
Change-Id: Id38df18f1e3f20ef2aa1cb1adfe96dacacad9ad6
Reviewed-on: https://go-review.googlesource.com/c/tools/+/431216
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Peter Weinberger <pjw@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
This CL updates the LSP to 3.17.0. It is a DANGEROUS CL as the stubs
are being generated by Go code reading vscode's language independent
description of the protocol (in metaMode.json in the vscode-languageserver-node
repository.)
Some of the union types in the protocol have Go types with names containing 'Or'.
These types have custom marshaling and unmarshaling code.
Embedded structures in the protocol are broken out as their own
types, with names constructed from the context in which they occur.
The natural output has been modified to minimize the number of changes
needed for gopls. (e.g., Workspace6Gn is preserved for compatibility.0
Thus, many types that are union types in the LSP description have been replaced by the
types gopls already uses.
Updates golang/go#52969
Change-Id: I16f6d877215155ac9e782b0f5bcbdab3f1aa2593
Reviewed-on: https://go-review.googlesource.com/c/tools/+/424214
TryBot-Result: Gopher Robot <gobot@golang.org>
gopls-CI: kokoro <noreply+kokoro@google.com>
Run-TryBot: Peter Weinberger <pjw@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
This CL was created using the following commands:
./gopls/internal/migrate.sh
git add .
git codereview gofmt
For golang/go#54509
Change-Id: Iceeec602748a5e6f609c3ceda8d19157e5c94009
Reviewed-on: https://go-review.googlesource.com/c/tools/+/426796
gopls-CI: kokoro <noreply+kokoro@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Peter Weinberger <pjw@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Robert Findley
Peter Weinbergr