When the corrupted `elc` file is detected
`cider-jack-in-lein-plugin` has a nil value
value added already which causes `cider` to fail.
This commit removes such a row in that case.
It seems to happen from time to time that sayid is not
properly compiled after the installation. In this case
the variable `sayid-version` is `nil` causing
nRepl to fail.
This commit tries to work around this problem by
reinstalling the package if this corruption is detected.
This normally fixes the issue described in #13906.
Microsoft has released pyright language server for python. It boasts better
performance than the old Microsoft language server for python. This commit adds
supports for pyright language server in python layer
* Switch from nabero's ron mode to my own custom ron-mode (hosted on MELPA)
* Replace depreciated perl6-mode withraku-mode
* Remember to update the init function
* Add RTags shortcuts for trees and dead functions
The change will add 3 shortcuts for dependency, references trees and for finding dead functions.
* Additions of some key bindings
We should always define the 2 flycheck initialization functions:
Spacemacs will call them as needed only when enabling syntax checking.
Previously flycheck wasn't getting activated at all.
Only activate `flycheck-ocaml` when a `.merlin` file is present.
It would work when there isn't but there is a high chance that you will
get a lot of errors displayed because it could not find any of the used
modules (`.merlin` specifies module paths).
OCaml projects usually do have a `.merlin` file
(either hand crafted or generated by the build system).
Signed-off-by: Edwin Török <edwin@etorok.net>
lang/ocaml: add merlin-eldoc and merlin-imenu support
This displays the type and documentation of thing under point
and highlights occurrences automatically when idle. Could be achieved by appropriate use of
keyboard shortcuts already set up by this module.
However `merlin-eldoc` can also display the expected type of function call parameters,
which merlin itself wouldn't.
`merlin-imenu` allows jumping to top level modules/functions/etc. bound
by default to `SPC j i`.
Signed-off-by: Edwin Török <edwin@etorok.net>
lang/ocaml: add merlin-iedit multiple cursor refactoring
bound to `, r e`.
Signed-off-by: Edwin Török <edwin@etorok.net>
fixup! lang/ocaml: add merlin-iedit multiple cursor refactoring
fixup! lang/ocaml: fix flycheck-ocaml activation
fixup! lang/ocaml: add merlin-eldoc and merlin-imenu support
biblio-/core is pulled in by multiple packages in this
layer already and their package declarations was empty.
Users can access the functionality biblio offers
from helm-bibtex.
The spacemacs analogue of cider-load-buffer-and-switch-to-repl-buffer
doesn't accept an argument to set the namespace while you are switching
to the repl buffer. This adds support for an argument, invokable with
the universal argument, for setting up the namespace as well. This will
eliminate an extra `(in-ns ...)` or `, s n` call as long as you invoke
`, s B` prefixed by the universal argument: `SPC u , s B`.
Missing parentheses around:
"\\.ron\\'" . ron-mode
in the line:
:mode "\\.ron\\'" . ron-mode
Updated the rust readme to mention the .ron support.
And added a changelog.develop entry.
* [ESS] Fix: remove obsolete variables in ESS
- use ess-offset-continued to replace the obsolete variables (from ESS 15.04):
ess-first-continued-statement-offset
ess-continued-statement-offset
- remove ess-expression-offset
no such variable now
* [ESS] fix: keep ess-default-style instead of ess-style
- keep the original ess-default-style instead of ess-style
since ess-default-style will be obsolete till ESS 19.04
while current ESS version is 18.10
`literate-haskell-mode` was renamed to `haskell-literate-mode` in this commit:
7ccb1ab0cc
Even though a deprecated alias was added, it broke Spacemacs' keybindings, which
were still all bound w.r.t. the old name.
README.org
Buffer formatting, first paragraph
The apostrophes before: yapf and black
caused the org emphasis verbatim equal signs to be visible.
config.el
python-formatter variable docstring
An apostrophe instead of a backtick, before the possible value: lsp
caused it to be shown with the default docstring color.
After using Java mode, Gradle keybindings infect or even clobber keybindings of
any subsequently used major-mode. For example, the critical Agda major mode
keybinding of `l` as `agda2-load` gets with a Gradle menu!
The cause, found by @duianto, is that `emacs-gradle-mode` sets itself to a
[global](e4d665d578/gradle-mode.el (L176-L183))
minor mode. The docs for `define-minor-mode` say about `:global:`
> If non-nil specifies that the minor mode is not meant to be
> buffer-local, so don't make the variable MODE buffer-local.
> By default, the mode is buffer-local.
I don't know why `gradle-mode` is doing this; presumably there is some reason or
need for it. But the author of that package hasn't been on GitHub since 2017,
and the last update of the package itself is from early 2015.
To seal the deal, `gradle-mode` hasn't been working anyway (at least, I've never
got it to work). I've been running Gradle by launching a terminal within Emacs.
Fixes#13750.
Add key bindings to refactor namespace forms for existing functions in CIDER.
"ran" 'clojure-insert-ns-form
"raN" 'clojure-insert-ns-form-at-point
"run" 'clojure-update-ns
A few of the layer's keybindings are shadowed when using the LSP backend:
- `F` - format buffer using stylish
- `hh` - lookup on Hoogle
- `gi` - navigate to imports
These are useful shortcuts to keep, so when the LSP backend is in use they are
rebound like this:
- `F` -> `S`
- `hh` -> `hg`
- `gi` -> `gl`
- `hH` -> `hG` (not one that is shadowed, but rebound to bind external and local
Hoogle lookups to similir shortcuts)
Popwin automatically configures pupo purposes
if properly configured. While I was checking
the existing configs I have found that the
settings will be ignored if the config is not
done via a post-config hook.
A quick search revealed a lot of popwin configs
which never had any effect.
This commit fixes these and with this also
restores the missing pupo configs.
spacemacs/go-run-test-current-function() previously failed when point was
anywhere but at the end of the line on the first line of the test function to
run.
Fuel mode is responsible to connect to a Factor instance. This change extends
the Factor layer to handle a graphical listener process, to which fuel can
connect afterwards.
A major motivation is also to make it easier to develop with different Factor
versions, which can be specified with (project-/directory-specific) variables.
When starting a Factor listener in a certain location this way, the elisp code
for fuel/factor mode is reloaded from that location.
AUCTeX is weird: It reports major-mode as latex-mode (since TeX-latex-mode is
applied as an :override advice on the basic built-in latex-mode), but its mode
hook is LaTeX-mode-hook, not latex-mode-hook (which is only run by the built-in
latex-mode). Since bind-map uses the value of major-mode, we must pass
latex-mode to spacemacs|define-jump-handlers. But then
spacemacs//init-jump-handlers-latex-mode gets added to latex-mode-hook, which
never gets run. So we must manualy add it to LaTeX-mode-hook.
[latex] Use dumb-jump as primary jump handler
Otherwise, the default is used, which prioritizes evil-goto-definition over
dumb-jump-go. Dumb Jump tends to Just Work, while evil-goto-definition doesn't
handle LaTeX very well, at least not without a TAGS table.
[latex] Update CHANGELOG.develop
Having an explicit dependency on lsp-treemacs for post-init-lsp-treemacs is no
longer necessary in the Scala layer, as lsp-metals superseeds it.
Removed lsp-metals-treeview-enable as it no longer exists.
This is a squash commit, it includes:
* Add go run and go test command variables
* Update go README for go-run-command and go-test-command
* Update CHANGELOG
As per the README, `SPC m s B' should put the REPL buffer in insert state, but
the keybinding as defined puts the racket file itself into insert state, not the
REPL. This means that on returning to the racket file buffer, the cursor is
left in insert state, which means it is easy to accidentally add unwanted text
into the file.
To fix this, a `with-current-buffer' wraps the `evil-insert-state', so that
insert state is enabled in the REPL buffer, not the file being edited.
The first time this is run, `racket-run-and-switch-to-repl' is asynchronous of
`evil-insert-state', so trying to get the Racket REPL buffer will error. To fix
this problem, we wrap the `with-current-buffer' with a check to determine that
the Racket REPL buffer is live. The first time the Racket REPL buffer is
created, the contents will not run, since the Racket REPL buffer will not yet be
live. This is fine, since we enter the REPL in insert state automagically.
This change does not need to be done to `spacemacs/racket-send-last-sexp-focus',
`spacemacs/racket-send-definition-focus', or
`spacemacs/racket-send-region-focus' since these functions follow all of their
racket-send functions with `(racket-repl)' before calling `evil-insert-state'.
flycheck-mix has been removed by the package author as being
obsolete with proper lsp and alchemist support. Therefore
the layer also has been cleared of it as it cannot longer
be found on Melpa.
As of Scala 2.13, Unicode arrows are deprecated:
* https://github.com/scala/scala/pull/7540
* https://github.com/scala/scala-dev/issues/585
* https://github.com/scala/bug/issues/11210
Using one will give a deprecation warnings like so:
> The unicode arrow `⇒` is deprecated, use `=>` instead. If you still wish to
> display it as one character, consider using a font with programming ligatures
> such as Fira Code.
As such the Scala layer's version slick capability to replace ASCII arrows with
Unicode ones is no longer useful, and I have removed it.
Based on my tests it doesn't seem that there is a need for a more graceful way
to deprecate this: i.e. nothing fails if there is extra junk in `:variables`.
Before the decision which linter to run before golangci-lint has been
taken by the layer. However this can also directly be done by the
checkers.
Also golangci-lint did run even when the previous checker still reported
errors. With this it is only called if the previous checker is satisfied.
All thanks goes to Tommi Komulainen.
Flycheck-golangci was not properly conigured which resulted in
not replacing disabled standard linters like go-test.
In addition there seems to be an issue in the current golangci
build causing the linter to fail. In this case spacemacs now
shows at least basic errors from go-build. When they are fixed
the linter should be able to work normally again.
A BEGIN_SRC block in the documents had no language specified
which would break CI.
I have also changed the eldoc setup to run each time the mode is
entered instead of only in the initially loaded buffer.
The previous way of adding dap to a layer did add the mode
unconditionally to `spacemacs--dap-supported-modes` causing
dap bindings to be added also when no lsp backend was used.
racket-xp-mode is an optional minor mode that enhances the racket-mode to
explain and explore Racket code. The racket-xp-mode is started with a mode-hook
on racket-mode. Deprecated racket-mode functions are replaced with their
racket-xp-mode versions. The remaining racket-xp-mode functions are added to
keybindings as per Spacemacs conventions.
Also add the racket-mode directory to .gitignore.
Also adjusted the tags to match the folders the layers
are in. I have also updated the binding for imenu mentioned
in protobuf layer and adapted the kubernetes tags to register
as tool instead of as os.
As discussed at length in #13392, the primary developer of EIN insists
on updating the spacemacs bindings to avoid several-iterations stale commands of
dubious functionality (in particular, multi-worksheet commands).
As the transient state docstring must be "mated" to the updated bindings, he prefers
not having to manually curate it, and exhorts the return of the
automated docstring. He doesn't insist on it, though, like the above.
bind-map.el counterfeits ein:notebook-mode
bind-map replaces on the `minor-mode-map-alist` the keymap for
ein:notebook-mode with the evilified map, thus making the original
ein:notebook-mode-map inaccessible.
This is might be intentional as one wonders whether having active both the
original keymap and the evilified keymap makes sense.
In EIN's case, yes, we really want the original keymap, e.g., C-c C-c
to execute, to work. Otherwise C-c C-c defaults to running a python
shell via the python-mode-map, which is really not what we want.
Go layer now supports dap, therefore when lsp is selected the layer
will load dap which in turn will load lsp to enable debugging capabilities
for all lsp users.
Show results of evaluating Clojure code in the same format they would be written
in the source code. This is especially useful for results that are
collections and nested collections.
Pretty print results as a comment.
1. Previously, the code had references to ein:notebook-mode which doesn't seem
to work (anymore), so I changed these references to ein:notebook.
2. The previous code had a big convoluted function that produced the
doc-string and that was difficult to read that depended on some other
smaller functions, though not enough of them to make the code
sufficiently modular, and therefore, it wasn't very maintainable.
Besides, when I tried the previous version with the name of the mode fixed,
spacemacs hung for a long time -- it's a mystery whether spacemacs would
have continued normally if I had waited long enough. Therefore I replaced
the doc-string with a hard coded one.
3. I added some keybindings from tkamat's fork to the code for I didn't see
a good reason not to
4. I removed the cl- prefix from "second" and "mapcan", because cl is a
deprecated package. However, I'm afraid the names without the prefix might
end up being unrecognized by previous versions of emacs.
Ensime seems to be finally dead, as ensime-mode is not longer
available on melpa. The same applies to ob-scala the package
which delivered scala support for org babel.
I have changed the layers default to metals and took care that
ensime is not tried to be installed until it is really selected
as a package.
In addition I have also fixed some smaller issues in the layer
which caused ensime specific settings to be forced even when
metals was selected as a backend.
I have also removed the not longer existing org-babel support
for scala as it requires ob-scala which in turn is based on ensime.
See https://github.com/hvesalai/emacs-scala-mode/issues/155 for details.
I have also updated the installation instructions as tsserver is not
longer available as single executable on npm. Also I've made sure that
modes register for tide backend only if it is really intended to.
Instructoins on the following external dependencies required by certain
functions are added/updated to the README file:
- `cargo-edit`
- `cargo-audit`
- `rustfmt`
- `clippy`
Signed-off-by: Lucius Hu <lebensterben@users.noreply.github.com>
Clojure (CIDER) is missing the incredibly useful function that jumps to the end
of the line and evaluates the last s-expression.
This function is in Emacs Lisp major mode under `SPC m e $` and `SPC m e l`
The function `spacemacs/cider-eval-sexp-end-of-line` is a copy of
`lisp-state-eval-end-of-line` with the last line replaced to call the equivalent
cider function, `cider-eval-last-sexp`.
The same keybindings are added to the Clojure layer as used for elisp.
The keybindings follow the Spacemacs convention for evaluation, as defined in
https://github.com/syl20bnr/spacemacs/blob/master/doc/CONVENTIONS.org#evaluationResolves#4124
This is a squashed commit it contains:
- Added new sections on `cargo-edit`, `cargo-audit`, and `Clippy`,
which are used for several key-bindings in this layer.
- Re-ordered the `Key bindings` section, added entries for new bidings,
and improved description.
- ~SPC m c a~ for =cargo add=
- ~SPC m c r~ for =cargo rm=
- ~SPC m c U~ for =cargo upgrade=
- ~SPC m c A~ for =cargo audit=
- ~SPC m c X~ is for =cargo run --bin= now, instead of =cargo run --example=,
which is bound to ~SPC m c E~
This layer supported the very old linum and the non working
auto-complete-rst package. The later is not longer maintained
and does not work with up-to-date releases of sphinx causing
To fix the issue I have disabled `linum` and `auto-complete-rst`.
Add installation instruction for `sqlfmt` executable
Adapt documentation to be more like the other layers
Replace wrong default sqlfmt parameter -u with --use-spaces
Make error buffers be reused by default to reduce memory consumption
Change prefix for repl keybindings to be conform with the other layers
Remove obsolete progn statement
Enhances `sqlfmt` support in SQL layer.
`sqlfmt` was only being applied against the whole buffer, and did not allow
invocation against a specific region in the buffer. `sqlfmt` customize options
were not included under the `sql` group. It was not possible to control if
`sqlfmt` reused an existing error buffer or generated new ones.
- Added `sqlfmt-region` interactive command and binding under `sql-mode` keymap
- `sqlfmt-buffer` calls into `sqlfmt-region` with `(point-min)` and
`(point-max`)
- Added `sqlfmt` customize group as child to `sql` customize group
- Added `sqlfmt-reuse-error-buffer` option
- Updated `sql` layer documentation to reflect changes
`lsp-mode` supports two LSP server backend for Rust language, i.e.
`rls` and `rust-analyzer`.
`rust-analyzer` is experimental and lacks certain functionalities
such as `DAP` support.
`lsp-mode` provides a function `lsp-rust-switch-server` that changes
the priority of LSP server backend for *new* LSP session.
This commit adds a key binding `SPC m s s` to `lsp-rust-switch-server`.
There's no keybinding for markdown-insert-kbd alongside all the other
markdown-insert-* keybindings, this patch adds one using the same
convention (specifically xk).
Added keybindings for the Sesman REPL session management commands to the Clojure
layer, following the Spacemacs keybinding conventions.
https://docs.cider.mx/cider/0.23/usage/managing_connections.html
Sesman has been a part of CIDER for many versions now and provides session
management for nREPL connections (in fact any connections).
Sesman keybinds are placed in the `SPC m c` menu, providing a cleaner separation
between sending code to the repl and session management.
See issue #12593 and PR #13140 which covers the background for the design of
these additional keybindings.
Formatting the region and defun keybinds have been added, along with formatting
for edn files. Moved to `=` menu to match Spacemacs convention.
Evaluation up to point keybinding has also been added.
Documentation in the README has been updated to complete the description of the
existing Clojure layer, as well as including details specific to the new
keybindings.
The snippets seem to have been accidentally changed in this commit:
a2de9a63af (diff-75e0388987cdf040d82762426e40e5a5R434)
where other instances of checking for a layer, were changed to check for a
package.
It doesn't work to check if the haskell package is used in the user config
section, because the haskell package hasn't loaded yet.
Reverting it to check if the haskell layer is used works.
Add Spacemacs keybindings to the Clojure layer for all the sesman functions.
Sesman has been a part of CIDER for many versions now and provides session
management for nREPL connections (in fact any connections).
Several keybindings relating to the use of a specific repl been migrated the
`e` evaluation menu and `es` sent-to-repl-buffer sub-menu, providing a cleaner
separation between repl connection and session management and interacting with a
specific repl.
Updated layer documentation and keybindings in README.org file.
Added changes to CHANGELOG.develop
Resolves#12593
Make rainbow-identifiers not colorize special operators and macros, so they
always visually stand out. Rationale behind this change is that special
operators and macros in Lisp may be considered "syntax" elements, so it makes
sense to have them visually distinguished at all times.
- Fixed setting dap-java shortcuts - ATM the shortcuts were not enabled due to
the fact that dap-mode configuration was overriding them since it was running
after this piece of code
- removed configuration code about lsp-java since it was either about
missing properties or about setting defaults already set by lsp-mode/lsp-java
(e. g. lsp-eldoc-render-all is now nil by default)
See #10825.
Try to match more closely the build process followed upstream:
- Force module build mode, which builds the latest stable release (not master),
and takes care of module replacements.
- Disable cgo.
- The -trimpath argument to go get requires go >= 1.13, which is also needed
because go <= 1.12 has a bug that causes problems with golangci-lint², as
explained in the golangci-lint installation instructions².
Note that the only thing that is missing to exactly match the upstream build
process is the addition of some -X variables, which only affect the data
reported by golangci-lint version.
¹: https://github.com/golang/go/issues/29612
²: https://github.com/golangci/golangci-lint#install
Thanks to @dbriemann and @seriousben for reporting and clarifying this issue!
removing the `lexical-let` caused the intended closure to be invalid. the
filename being compiled was nil when the compilation-finish-function was
run.
this adds a special var to hold the temp file being compiled instead, and
tries to narrow down the compilation-finish-function to only run when the
compilation buffer includes the "rustc -o /tmp" regexp
Before the ruby layer was only requiring standalone lsp layer,
however with the latest changes dap support has been added
therefore I have updated the layer file to require the full
dap layer instead which in turn requires the lsp layer.
I have:
- Revised the documentation to be more clear about the possible backends
- Simplified the lsp setup code
- Removed redundant loading of company-lsp (is done by lsp-mode already)
- Made lsp the default backend for dart and mentioned that the analyser is
deprecated
An older PR reintroduced a dependency on lexical let.
There is no proper replacement for this in cl-lib therefore
I have changed it to a normal let which works the same
way.
In future we should think of making the entire file lexically bound
and only mark the exceptions to be dynamically bound.
I have tried to integrate flow-js2-mode into flow-type in addition
I have fixed a couple of issues in the layer and replaced obsolete
react-mode bindings with rjsx-mode bindings.
This allows to use quick run when the default shell is not bash-compatible.
I also fixed the output name by using a digest of the filename rather
than always a new file as I noted that my /tmp got cluttered with 4MiB
files pretty quickly, and I was just using a hello-world like file.
Make LSP be used as default backend
Request DAP layer when LSP is used as backend
Update README.org with latest pyls package dependencies
Remove obsolete getter functions
Update the version of metals referenced in the metals installation
instructions, so that an old version of the metals binary doesn't get installed
by accident.
- lsp-mode supports defining remote versions of each clients(e. g. clangd which
works over tramp). Before this change the language server will be disabled
because it has different server-id.
Add a full/minified transient state toggle.
Grouped backend transient state key bindings by column.
Moved the html/init-web-mode :init section before the :config section.
The helpful layer delivers more sophisticated help buffers,
at least for elisp and emacs specific objects. Therefore
it makes sense to replace the existing help functions for
these bindings silently if this layer is loaded.
Make doc more clearly describe which tools belong to which backend.
Change keybindings to be consistent with spacemacs conventions.
Simplify LSP layer code.
This environment variable is used by python LSP to determine
the pyenv virtual environment to use. When switching to a different
project using projectile 'switch-to-project', it should only be
necessary to restart the LSP server but a restart will only
work if this VIRTUAL_ENV changes value too.
Improve documentation to be clear about the choices of backends.
Remove obsolete functions.
Make sure that company-lua is only loaded when the right backend is used.
Make keybindings conform to conventions.
To activate javascript-eslint all lsp layers where
running eslint during lsp-ui flycheck phase. This
caused a lot of side effects and has been deactivated
for now. Eslint shall be called by the lsp server instead.
Removed org verbatim emphasis around link descriptions.
Moved up a paragraph before the julia-mode-enable-lsp code snippet,
because it ended with "then install this layer with:".
Removed the link from the same paragraph, it linked to it's own section.
Loaded lsp-julia from melpa
And fixed#12852
spacemacs/go-setup-backend was incorrectly added to the hook:
julia-mode-local-vars-hook
The correct function is now added: spacemacs/julia-setup-lsp
correct a tick
The variable clojure-enable-linters can now be set to a list of linters:
'(clj-kondo joker)
or
'(joker squiggly)
or any other combination of the available linters.
In which case, all linters in the list will be active if they are
available to be. In that flycheck is configured so that each linter
in the list has as next-checker the next linter in the list.
The variable can also now be set to:
'joker
If only the newly added joker linter should be used.
This will pull down flycheck-joker and add Joker as a checker for
Clojure modes.
--
Implementation Details
The way it works is that for each Clojure linting mode, clj, cljc,
cljs and edn, a primary linter is chosen based on the order the
user specifies in clojure-enable-linters which support that mode.
It is made the primary linter for flycheck to use by being promoted
to the front of the flycheck-checkers list. All other linters specified
by the user which the mode also supports get added to it as a
next-checker.
Tested with clj-kondo, joker and squiggly. The only issue I saw was that
the way flycheck-clojure works, and the way I currently have setup the
logic for multiple checkers mean that we can't add to a mode multiple
checkers for the same linter. flycheck-clojure has eastwood, kibit and
core.typed. If you have it configured to use eastwood linter, then it
will all work, as eastwood gets added as the next-checker. But if you
have it disabled in flycheck-clojure, there won't be any
flycheck-clojure linters to activate. I'm not sure as of now how to best
support that. Ideally, we would not clubber linters into a single packge
like that, and so the layer would have instead of squiggly: eastwood,
kibit, core.typed. But that's not how flycheck-clojure works. Future
commit might be able to do something smarter about this.
Conda layer needed some additions to be spacemacs
conform. In detail I did:
* Add missing features block and TOC declaration in README
* Change bindings to extent the existing python bindings
rather than adding new general bindings.
* Make conda layer require python layer
There was a edge case with the declaration of the `lsp` layer in `layers.el`
files.
The `hy` layer depends on the `python` layer which in turn depends on the `lsp`
layer if and only if the `python-backend` layer variable is set to `lsp`.
When the `hy` layer was declared first then it declares the `python` layer
without its layer variables, thus the `lsp` layer was not declared because the
`python-backend` variable was not set.
The fix is to gather all the layer dependencies and resolve them only after all
the used layers have been declared.
* new function `configuration-layer/declare-layer-dependencies`
* replace all calls to `configuration-layer/declare-layer` by the new function
except for distribution layers (we declare layer dependencies right away in
distribution layers)
* use local-vars-hook coupled to setup function for dap
* define new private layer variable `spacemacs--dap-supported-modes` to
configure key bindings. This allows to move the key bindings definition from
`funcs.el` to `packages.el`
* remove duplication of DAP key bindings in READMEs by pointing to the dap layer
documentation
* alphabetically sort package configuration
The dart layer requires the official LSP server as the
emacs mode alternative is not longer properly maintained.
Therefore I have added some more description of how to
install the server properly and let the layer load
the LSP layer automatically.
In addition I have removed the unsupported variable
`dart-server-enable-analysis-server` from the
documentation.
See https://github.com/bradyt/dart-mode/wiki/LSP for
details.
