| Commit message (Collapse) | Author | Age |
|---|
| | |
|
| | |
|
| |
|
|
|
| |
Also requires yarn 3.2.0-rc.5 due to
https://github.com/yarnpkg/berry/issues/3722
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The root project (worktree) must have a valid package.json,
.yarnrc.yaml, and have the gradle precompiled script plugin
refinery-frontend-worktree applied to it.
All frontend modules (workspaces) must have a valid package.json, be
referenced in the worktree project.json, and be a valid gradle
subproject with the gradle precompiled script plugin
refinery-frontend-workspace applied to it.
To reduce unnecessary downloads, we do the following workarounds:
* The node version is pinned in gradle.properties and the installed node
version is stored in .node/frontend.properties. A new node version is
only downloaded if it differs from the installed one (or none is
installed). This improves on the default gradle behavior, which would
download a new node version if the classpath of the installNode task
was changed.
* The yarn classic version is pinned in gradle.properties and the
installed yarn classic version is stored in .node/frontend.properties.
This improves on the default gradle behavior, which would reinstall
the yarn npm package at every build.
* We disable the enableYarnBerry task, since a yarn berry version is
already committed into the repository at .yarn/versions as it is
recommended in http://yarnpkg.com/getting-started/qa/
* We use yarn set version --only-if-needed in the installYarn task.
* The installFrontend task takes package.json and yarn.lock as inputs
and produces .pnp.cjs and .pnp.loader.mjs as outputs. Whether this
task is up-to-date can be checked by gradle quickly, since it doesn't
have to traverse the node_module folder due to pnp resoltion:
https://yarnpkg.com/features/pnp/
All workspaces have the installNode, installYarnGlobally, installYarn
and installFrontend tasks disabled. However, these tasks depend on the
task with the same name in the worktree project, so installation is
still triggered globally. Moreover, the installFrontend task in the
worktree has the package.json of the workspaces as inputs, so
reinstallation is triggered when the workspace package.json changes.
|
| |
|
|
|
| |
It is not compatible with eslint 8 yet:
https://github.com/SonarSource/eslint-plugin-sonarjs/issues/286
|
| |
|
|
|
|
|
| |
Also upgrades various frontend dependencies.
We can't upgrade to typescript 4.5 yet, because
https://github.com/yarnpkg/berry/pull/3760 is not released yet.
|
| | |
|
| |
|
|
| |
All Java code should have tests, anyways.
|
| |
|
|
|
| |
Script plugins allow applying other plugins from plugins { } blocks,
which simplifies the build script.
|
| |
|
|
|
|
|
|
|
|
|
| |
Gradle version catalogs allow centralized management of dependency
artifact coordinates using a toml file.
Previously, we use the gradle.properties file, which cannot manage group
and name coordinates, only versions.
Node and yarn versions still have to be in gradle.properties, because
they do not correspond to any coordiates of a maven artifact.
|
| | |
|
| |
|
|
|
|
|
|
|
| |
Uses the xtext formatted on the server to format the document.
Also adds the capability to take (delta) changes from the server and
apply them before any pending local changes, then replay the changes.
This means that the server-side formatter is effectively acting as a
second user who is editing the document.
|
| |
|
|
|
|
| |
Use the zod library instead of manually written type assertions for
validating json messages from the server. This makes it easier to add
and handle new messages.
|
| | |
|
| |
|
|
| |
Suggestion by @vincemolnar to match SysML V2 terminology more closely.
|
| |
|
|
|
| |
Move information about which tokens should support implicit completions
into the Lezer grammar.
|
| |
|
|
|
| |
Contextual keywords make Xtext parsing more complicated and degrade
content assist.
|
| |
|
|
|
|
| |
We can't seem to be able to style the fold gutter in the current line to
set its background color, so we set the background of the whole gutter
instead.
|
| |
|
|
|
|
|
|
|
|
| |
Always try to complete the current token if it is a terminal
(e.g., true, false, unknown, and error after a : or an = sign).
Autocomplete still only starts without being explicitly invoked if there
is a QualifiedName to complete.
Discard completions with a shorter prefix than the current token,
because they would be filtered out by CodeMirror anyways.
|
| | |
|
| | |
|
| |
|
|
|
|
| |
Only return find occurrences results for NamedElement instances,
as other objects cannot be referenced by name and thus cannot have any
references at all.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
In some resource-constrained environments (e.g., my VPS with 2 vCPUs),
it was possible for validation and highlighting precomputation to be
finished before the server responded to a deltaText updating completion
request (updating completion take longer than a normal update, because
they also have to compute the completions). Therefore, the client
received push messages about a stateId it didn't know about yet.
To fix this, we delay any push messages originating during servicing a
call to be sent after the call is serviced. Thus the client first
receives the updating completion response with the new stateId, followed
by the push messages relating to that stateId.
|
| | |
|
| |
|
|
|
| |
Do not misuse ToggleButtons as ImageButtons when there's no toggled
state.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
extract utils/ConditionVariable from xtext/UpdateService as a generally
useful utility class for waiting for conditions
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
Kristóf Marussy