When the build files changed since the last sync, the cached effective-POM plugins and resolved dependencies replaced what the POMs declare now, so a just-added dependency never appeared until Reload. The cache is now applied only when the build files are unchanged; a non-verbose cache still marks the model stale without discarding its data.
Co-authored-by: a-rookie-of-C-language <186315048+a-rookie-of-C-language@users.noreply.github.com>
Names in Settings › Ignored files and folders were hidden only when untracked; their tracked or staged changes still showed in the change list. They are now filtered too, except conflicts, so a merge can't stall on a hidden file.
Co-authored-by: a-rookie-of-C-language <186315048+a-rookie-of-C-language@users.noreply.github.com>
Follow-up to #385. A single output event can carry thousands of lines, which became one console entry and escaped the line cap; output is now split at real line breaks (partial lines still merge across events) so the 50,000-line bound applies to physical lines. The host also coalesces debug output per session for up to 50 ms / 64 KB before it reaches desktop or remote IPC, flushing before any other session event so stops and exits never overtake earlier output.
Verified on macOS with a Java program printing ~480 lines/s for over a minute: the console stays responsive, follows the tail, and pause/step work.
Fixes#377
Kotlin line breakpoints never verified through java-debug because jdtls cannot resolve classes for .kt files. Kotlin Application configurations and JDWP attach for modules with Kotlin sources now use the official Kotlin Debug Adapter (installable from the debugger prompt, running on the language-server JDK); pure Java keeps java-debug.
The adapter resolves classes from source paths, so sources whose directories don't match their package are mirrored package-aligned outside the project (symlinks, copies where links fail) and paths are mapped back for breakpoints and stack frames; files that can't be mirrored are skipped rather than failing the session. Kotlin applications start with JDWP suspend=y and a session-owned JDWP relay holds the adapter's early resume until attach completes, so breakpoints are installed before the program runs.
Limitations (docs/kotlin-debugging.md): @file:JvmName facades are not resolved; package/source-root changes need a new session. Not verified on a real Windows machine.
Fixes#374
Adds Tomcat Server › Local run configurations for WAR applications: installation validation and version detection, Maven/Gradle WAR artifact discovery and build before launch, WAR file or exploded-directory deployment, context path and HTTP port, project JDK or a chosen JRE, VM options, environment variables and opening the browser.
Each configuration runs with its own CATALINA_BASE in the project cache (HTTP, shutdown and local JMX ports configured there); Stop requests a clean shutdown before the usual process timeout. Debug attaches the Java debugger over JDWP. Update Application (Cmd+F10) updates resources, classes and resources with HotSwap, redeploys or restarts the server. Configurations are grouped under Tomcat Server in the configuration tree and Services.
Not verified on a real Windows machine; command construction is unit tested.
Fixes#205
java-debug treated forward-slash Windows drive paths as URIs, so breakpoints never verified; source paths are now sent with native separators and mapped back for stack frames and breakpoint events. Debug sessions also flush every file's pending breakpoint updates before starting, so configurationDone never runs with an empty breakpoint snapshot (all adapters, launch and attach).
Not verified on a real Windows machine; the path conversion is covered by unit tests.
Fixes#380
Refs #374
- Git › Show History is offered for folders, multi-selection and the project root, not just single files.
- The Push dialog lets you pick any configured remote; pushing to another remote keeps the upstream's branch name and shows outgoing commits against that remote's branch.
- The Changes toolbar gains New Changelist (changelists already supported new/edit/delete/set active/move).
- Compare with Current gains Show Diff, opening the whole-project diff between the branch and HEAD.
- Merge view: F7 / Shift+F7 navigate conflicts, the three panes scroll together, and each side pane can be annotated with Git Blame of its own content.
- Windows-only test fixes in ide-lsp and ide-sdk.
Fixes#343Fixes#310Fixes#358Fixes#348Fixes#320
Refs #317
Co-authored-by: a-rookie-of-C-language <186315048+a-rookie-of-C-language@users.noreply.github.com>
Unversioned files and folders in the Commit tool window's changes menu offer Add to .gitignore / Add to .git/info/exclude, the same action as the Project view's Git submenu. vcs_add_to_ignore returns the ignore files it changed and the IDE opens the nearest one after writing. A selected symlink or Windows junction is ignored by name without a trailing slash; symlinked ancestors are still refused, and already-ignored paths are left unchanged.
Co-authored-by: guoyucode <29471644+guoyucode@users.noreply.github.com>
Recent Eclipse JDT LS releases need Java 25 to run, so projects on JDK 21 had no Java language support until the user downloaded a newer JDK by hand. When JDT LS is about to start, no installed JDK is new enough, and no language-server JDK is set explicitly, the IDE now downloads the Temurin JRE of the required version (about 50 MB, falling back to the JDK image where no JRE build exists) with a cancellable background task, then starts the server on it. The runtime lives in its own folder under the app data directory, outside the roots scanned for project JDKs, so the project JDK, run configurations and JDT LS's project runtimes are unchanged. One download runs at a time across windows and is reused; a failed attempt falls back to the existing manual Download JDK action and is not retried until the next start.
Fixes#258
Refs #271
未在 Windows / Linux 实机验证。
Three ide-agent tests occasionally hung forever (about 1 in 40 runs under parallel load). Their fake agent answers a turn by requesting permission from a blocking thread and waits for the reply; the tests ended while that request was pending, and dropping the Tokio runtime waits for its blocking pool, which never finished. The tests now cancel the pending turn and wait for the task to settle before shutting down. Test-only: in the app, shutdown kills agent processes rather than waiting on them.
Archived agent tasks could be opened but never returned to the active list, so their message composer stayed disabled. Unarchive is now available in the task header, in the archived-task notice above the composer, and in the task context menu (including mixed bulk selections). Restoring a task moves it back to the active list, keeps its conversation and worktree, and reconnects the existing agent session on the next message; the backend persists and broadcasts the restored status and refuses restoration while a stopped turn is still settling.
With a project open, the app became slow and then unresponsive after 30-60 minutes. Git status writes refreshed stat data back to .git/index; the watcher reported that write as a file change and a VCS change; the UI scheduled another status ~250 ms later, which wrote the index again. With any stat-dirty file this never settled: a full status, IPC traffic and decoration rebuilds in an endless loop.
Watcher batches now carry .git internals separately (FsBatch.vcs), so they no longer appear as worktree changes and no fsChanged is sent for .git-only batches. Status records the identity (size, mtime) of the index it writes, and the host skips VcsChanged only for that write; external index changes (git add / reset / stash from a terminal) and HEAD / refs / packed-refs changes still refresh as before. The UI's fsChanged handler also ignores .git-only batches.
Fixes#311
Co-authored-by: a-rookie-of-C-language <186315048+a-rookie-of-C-language@users.noreply.github.com>
After choosing a model and starting a new chat, the picker still showed that model but the session fell back to the agent's default. The remembered configuration was recorded from the agent's reported defaults before the user's choices were applied, and changing an option during a session did not update it. A new session now remembers its options after applying the user's choices, and changing an option updates the remembered configuration.
Co-authored-by: dimassfeb-09 <62858781+dimassfeb-09@users.noreply.github.com>
Zip symlink entries were only validated on Unix: on Windows they fell through to the regular-file path, so a link escaping the destination was written as a file instead of being rejected (and the escape test failed on Windows CI). The symlink branch now validates on every platform; Unix creates the in-archive link, while Windows skips it, since creating symlinks needs a privilege most users lack.
Co-authored-by: a-rookie-of-C-language <186315048+a-rookie-of-C-language@users.noreply.github.com>
The Angular language server (ngserver) was not found when Node and global npm packages came from a version manager (nvm, fnm, volta, asdf, mise) whose bin directories are not on the GUI app's PATH. Tool lookup now searches those directories too through a shared ide-core tool_path module used by the LSP launcher, SDK queries and tool detection; explicit relative commands resolve against the project root, empty PATH entries are ignored, and a non-executable file earlier on PATH no longer shadows the real command.
The process query test that flaked under load now uses a generous deadline for the success case, with the timeout covered by a separate test of a child that never exits.
Co-authored-by: nntndfrk <2458463+nntndfrk@users.noreply.github.com>
OpenCode handles /compact through its ACP session but does not list it in availableCommands, so the composer never offered it. Agent builtins the agent implements without advertising are now merged into slash-command completion without duplicating listed commands; the table holds only commands verified against the agent (OpenCode /compact). Claude Code already advertises /compact. Executable lookup keeps using the shared PATHEXT-aware resolution on main, with a test that the newest install across PATH entries is chosen.
Co-authored-by: a-rookie-of-C-language <186315048+a-rookie-of-C-language@users.noreply.github.com>
cargo deny check now passes, with a deny.toml at the repository root (advisories, licenses, bans, sources).
- zip 2.4.2 → 9 as one workspace dependency (RUSTSEC-2026-0367/0368/0369). Our own extraction is hardened too: tar/zip entries cannot escape the destination, symlinks are allowed only when they resolve inside it, entries cannot traverse a symlink, and setuid/setgid/sticky bits are dropped (also after system tar).
- tiberius 0.12 → 0.13 with rustls 0.23, which brings the fixed rustls / rustls-webpki (RUSTSEC-2026-0098/0099/0104/0346/0349/0382/0383). The bb8-tiberius adapter is replaced by a small bb8 connection manager in ide-db that keeps its single-redirect handling.
- crypto-bigint / ctutils updated (RUSTSEC-2026-0385/0387).
- rsa RUSTSEC-2023-0071 (no fixed release) is ignored with its reason: RSA is used only for MySQL password encryption and SSH signatures, never to decrypt attacker-controlled ciphertext.
- Unmaintained transitive crates (paste, proc-macro-error, rustls-pemfile, zbus/zvariant) remain warnings. MPL-2.0 is allowed (file-level copyleft, dependencies used unmodified).
未在真实 SQL Server 及 Windows / Linux 上验证解压。
A language server and the fallback servers after it in languages/*.toml now share one picker in the Language Servers details pane, for every language rather than Python only: Language server for <languages>: Automatic, each server (with its status) or None, with the Automatic order in the hint. Servers outside such a group keep their on/off checkbox. ToolInfo.alternatives lists each server's group in preference order; a regular server after a fallback starts no group with it. The Python-only picker at the top of the page is removed, and names drop a language suffix the label already states. Storage is unchanged (each server's enabled flag, global or project scope), so existing settings carry over.
Git no longer asks for the HTTPS user name and password on every fetch, pull and push. The password prompt offers "Remember for this host"; the credential is saved in the app's secret store (next to the saved SSH passwords, never in git's global configuration) only after the git command that used it succeeds, and a stored credential the server rejects is forgotten so the next command asks again. Settings › Version Control › Git lists saved credentials and forgets them.
Each git child gets its own id, passed to the askpass helper, so concurrent commands never save or forget each other's credentials. ide-vcs reports command completion through an observer installed at start-up instead of depending on ide-ssh. Credentials are keyed by scheme, host and port, and only git's authentication errors (Authentication failed, HTTP 401/403 responses, invalid username or password) forget a stored credential.
Fixes#272
Co-authored-by: a-rookie-of-C-language <186315048+a-rookie-of-C-language@users.noreply.github.com>
Accepting an LSP completion after typing more characters than the request saw (`System.out.` requested, then `Print` typed while the cached list stayed open) inserted the item after the typed text: `System.out.Printprint()`. The cached edit range started exactly at the request caret and was mapped forward past the newly typed characters. The range start now stays at the request caret, so the typed prefix is replaced, whatever its casing. Enter (insert) and Tab (replace) keep their behaviour for the text after the caret; this applies to every language server.
Fixes#316
Adds Settings › Build Tools › Gradle: Gradle JVM (defaults to the Project JDK), distribution (wrapper or local installation with its home) and Gradle user home, stored in the project or global settings.
Sync and Gradle tool-window tasks use the configured JVM, launcher and GRADLE_USER_HOME. Run configurations keep their own choices: one that uses the wrapper runs gradlew, one that runs plain `gradle` resolves it from the local installation, then the pinned toolchain, then PATH; an explicit JAVA_HOME in the configuration wins over Gradle JVM. Derived defaults are applied at launch and not saved into run configurations. New Gradle configurations start with the wrapper off when the project uses a local installation.
Fixes#232
Git blobs were always decoded as UTF-8, so files in GBK or other legacy encodings showed every line as changed in the gutter, diff, blame and commit views, and staging/committing could re-encode them wrongly.
Blob text is now decoded and encoded with the file's encoding: the open document's encoding, else detection on the working file with the project default as fallback, else detection on the blob. Covers line hunks, file-at-revision, blame, stage patch/hunks, commit, conflict versions, diff text, and the agent worktree's file_at, partial commit and patch helpers. Text that cannot be represented in the target encoding is reported as an error instead of being written lossily.
Fixes#243
未在 Windows 实机验证。
On Chinese Windows the Maven tool window showed sync errors as mojibake: tool output was decoded as UTF-8, but while JVMs are told to write UTF-8, cmd.exe, batch launchers (mvn.cmd, gradlew.bat) and native tools write their diagnostics in the ANSI code page (GBK). ide_core::output_encoding::decode_tool_output keeps valid UTF-8 and decodes other lines in the system code page (GetACP). It is used wherever OpenIDEA shows output of tools it launches itself: Maven/Gradle sync and tasks, tool installers, SDK probes, Docker, coverage, environment setup, Cargo stderr for debugging, agent version probes and notebook kernel stderr. Run consoles and protocol streams keep their own encodings. 未在 Windows 实机验证。
Refs #262
As in IntelliJ IDEA, a root Maven project can be unlinked from the Maven tool window's context menu (Unlink Maven Projects, with confirmation) and linked again with the toolbar's Add Maven Projects (pick a pom.xml). Unlinked root POMs are stored in the project's Maven settings (unlinkedProjects). ide-host computes the unlinked roots and their modules once with ide-build and applies the same exclusion to the build model, Maven sync, project source roots, detected run configurations and the Java language server import (java.import.exclusions). The project's files stay visible in the Project view as plain files.
Refs #271
Detected run configurations (Spring Boot and Java applications, npm scripts, Maven/Gradle tasks, pytest, Cargo, Go, …) could not be removed: the − button was disabled because a removed detection would be found again. As in IntelliJ IDEA, they can now be removed and stay removed: the removed id is stored in the workspace-local run.local.toml (suppressed-ids) and detection skips it; saving a configuration with that id again lifts the suppression. Applies to the Run/Debug Configurations − button, the Delete key in its tree, and Delete in the run configuration chooser. Removing a saved override still restores the detected configuration; temporary configurations are unchanged.
Refs #271
HTML files in Angular projects were highlighted as plain HTML. Languages can now declare framework UI modes with root markers (angular.json, node_modules/@angular/core); files under such a root get the framework's editor mode, re-detected when a marker appears or disappears, and the root-marker lookup is shared with language server roots. The Angular Template mode (vendored @codemirror/lang-angular, MIT) adds @if/@for/@switch/@defer/@let blocks, key and animation binding names, and colours {{ }} delimiters.
Lazily loaded editor modes are now loaded before a document's first editor opens, so a file never renders as plain text first (this also left Angular templates uncoloured).
Co-authored-by: nntndfrk <2458463+nntndfrk@users.noreply.github.com>
Opens .ipynb files in a native notebook editor, in a project or LightEdit.
- crates/ide-notebook: lossless nbformat documents (unchanged notebooks save byte-for-byte, unknown fields kept, external changes detected) and owned local kernels over ZMQ via kernelspec — the selected Python environment only needs ipykernel, no Jupyter server.
- ide-host owns notebook documents per project incarnation and admits one execution batch at a time; outputs stream over IPC channels and kernel status is an IdeEvent.
- src/features/notebook: virtualized Python/Markdown cells, insert/delete/move/change type, Run Cell / Run All (Shift+Enter), interrupt, restart, interpreter selection from ide-sdk discovery, shared split views. Raw/unknown cells stay read-only.
- Custom editor kinds take part in dirty state, save/save all, undo/redo and Save/Don't Save/Cancel; closing editors from other features awaits that decision.
- Outputs: text, tracebacks, PNG/JPEG and sanitized SVG inline; HTML sanitized in a sandboxed iframe without allow-same-origin; JavaScript output is shown as inert source.
- Explicit launches (opening a notebook from the OS) win over session restoration.
Closes#131
Co-authored-by: dvictor357 <187467850+dvictor357@users.noreply.github.com>
The database tool window can connect to Microsoft SQL Server (tiberius over rustls with a bb8 pool): browse schema objects, edit table data, run queries and view plans; SQL Server dialect, object dialogs and icons in the UI.
Values are rendered by column type: exact decimal/numeric, money with four decimals, tinyint, uniqueidentifier, date/time/datetime/datetime2/smalldatetime/datetimeoffset in ISO form, xml and binary. Multi-statement edits roll back on any failure and discard the connection rather than returning it to the pool mid-transaction; connect-time errors (login, database) are reported; result pagination reports more rows; cancelling a query stops reading and drops the stream. The sqlx drivers keep the shared with_pool! path.
Co-authored-by: dimassfeb-09 <62858781+dimassfeb-09@users.noreply.github.com>
The font pickers (editor, fallback, UI, terminal) become editable comboboxes: type to search installed families or enter any family name, and Refresh re-reads the system fonts. ide-fonts' cached family list is now refreshable (DirectWrite re-queries the system collection with checkForUpdates; CoreText and fontconfig are read anew), exposed through system_font_families(refresh). Before the first native result the pickers fall back to Canvas detection of common families; saved names and the last catalog are kept when a refresh fails. Fonts are enumerated on the local desktop in remote-development windows too.
Co-authored-by: bghtyu <3931591+bghtyu@users.noreply.github.com>
.bat and .cmd files were plain text: no highlighting, no comment toggling and no Run. Batch is now a built-in language (languages/bat.toml, line comment @REM as in VS Code) with a CodeMirror stream mode for keywords, labels, %VAR% / %%i / !VAR! variables, strings and :: / REM comments. Comment toggling always separates a word-like token (REM) from the code. On Windows hosts the editor and project-view Run use cmd.exe /c <file> in the file's directory; other hosts don't offer it. 未在 Windows 实机验证运行。
Fixes#236
On Windows std/tokio Command only tries .exe for a bare program name, so agents installed as npm shims (codex.cmd, claude.cmd) failed with "program not found". Bare names are now resolved through PATH x PATHEXT before spawning.
- ide-core::process gains find_executable / find_executable_in (moved from ide-run::which, which re-exports them).
- ide-jsonrpc Peer::spawn_framed resolves against the child's PATH (caller env, login shell, then ours): covers ACP agents and the native Codex driver.
- ide-agent newest_install builds candidates with executable_names, so the native Claude driver finds .cmd shims and multi-install selection works on Windows.
- ide-agent replay recorder resolves the same way.
Co-authored-by: a-rookie-of-C-language <186315048+a-rookie-of-C-language@users.noreply.github.com>
The OS clipboard's file list was only wired up for Windows (CF_HDROP), so on
macOS the project view never saw files copied in Finder: Paste stayed disabled
and ⌘V did nothing. Read and write the general NSPasteboard's file URLs
instead, so Finder's copies paste into the project view and the project view's
copies paste in Finder.
Closes#246
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Python file launches only exposed the script directory, so imports of sibling project packages failed. Add content, source and test roots to the effective PYTHONPATH after environment resolution, preserving existing paths. Apply the shared behavior to Python/test process launches and debugpy launches while leaving attach and remote adapter paths unchanged.
Refs #216
The Git tool window followed one globally selected repository even when multiple roots were registered. Aggregate repository logs and branches, preserve repository identity in selections, history tabs, stashes, diffs and patch commands, and capture repository targets for delayed operations and dialogs. Validate tag names against the captured repository.
Fixes#195
The font pickers never read the installed fonts: they probed a hard-coded list of ~30 names with canvas width checks, so anything outside the list never appeared and the monospaced filter changed nothing. A new ide-fonts crate asks the platform font system for families and their fixed-pitch flag (CoreText on macOS, DirectWrite on Windows, fc-list on Linux/BSD), exposed as system_font_families (spawn_blocking, cached). Every font picker — editor font and fallback, terminal font and fallback, UI custom font — uses one FontSelect with prefix type-ahead, a real monospaced filter, and keeps an uninstalled selection visible.
Not verified on Windows or Linux machines.
Fixes#199
Claude-Session: https://claude.ai/code/session_01Dq6jQQn8ZhRfkkyj3PRutA
Toolchain detection only knew fixed install locations (SDKMAN, nvm, pyenv…), so per-version managers were invisible and project-pinned versions were ignored; mise also only reaches PATH through an interactive-shell hook. JDK, Node.js, Python, Go and Maven installs under mise (MISE_DATA_DIR / XDG) and asdf (ASDF_DATA_DIR) are now detected, pinned versions from mise.toml variants and .tool-versions are resolved and used when the SDK setting is Automatic, and the managers' shims/installs are added to PATH when the login shell didn't.
Fixes#207
Claude-Session: https://claude.ai/code/session_01Dq6jQQn8ZhRfkkyj3PRutA
Only the JVM the IDE launched directly was told the console's output encoding, so forked JVMs (spring-boot:run, Surefire, bootRun, Gradle workers), debugger-launched JVMs and Build-window builds wrote GBK on Windows while the console decoded UTF-8. Every JVM the IDE starts on Windows now gets the console encoding: java gets -Dfile.encoding alongside the std* properties, build tools and their forks get JAVA_TOOL_OPTIONS (std* only) with the IDE's own 'Picked up JAVA_TOOL_OPTIONS' line filtered, shell command lines are recognised by their first word, and java-debug launches get the matching vmArgs. The Debug console now shares the Run console's ANSI parser. Encoding helpers moved to ide-core::output_encoding.
Not verified on a Windows machine.
Fixes#180Fixes#182
Claude-Session: https://claude.ai/code/session_01Dq6jQQn8ZhRfkkyj3PRutA
MySQL flags text columns with a binary collation as BINARY, so sqlx reports them as VARBINARY/BLOB and cell() formatted them as 0x… — including information_schema names (utf8mb3_bin under lower_case_table_names=0, or anything with character_set_results=binary). The hex schema name was then used in follow-up queries, which matched nothing. Metadata queries now always decode text as UTF-8; result grids show hex only for true binary data. Covers schemas, tables, columns, indexes, foreign keys, triggers, routines, DDL, completion and the table editor's _bin primary keys.
Fixes#196
Claude-Session: https://claude.ai/code/session_01Dq6jQQn8ZhRfkkyj3PRutA
When Claude Code loses its API connection or is refused, it retries on its own
for up to several minutes. The session showed only "Working…" meanwhile. It
now shows "API request failed, retrying in Ns (attempt 3/10)" with the error
in the tooltip, cleared by the agent's next output or the end of the turn.
jdtls was launched with -Xms1G and no heap cap or GC choice, so the JVM grew to a quarter of physical RAM during import and kept it. Default VM args are now ParallelGC with GCTimeRatio/AdaptiveSizePolicyWeight, -Xmx2G -Xms100m, disabled memory-mapped zips and VM-install detection, prepended as --jvm-arg so both the direct and launcher-script paths get them. User JVM options replace the matching default (heap, -D/-XX by name, any -Xlog, any GC flag drops the whole GC group).
Fixes#193
Claude-Session: https://claude.ai/code/session_01Dq6jQQn8ZhRfkkyj3PRutA
Adds Application and Spring Boot run configuration types with main class, VM options, program arguments, classpath and active profiles. Maven passes them through spring-boot.run.jvmArguments / profiles / argLine; Gradle through an init script.
Fixes#187
Claude-Session: https://claude.ai/code/session_01Dq6jQQn8ZhRfkkyj3PRutA
* feat(settings): HTTP Proxy settings for the IDE's outgoing connections
Add Appearance & Behavior › System Settings › HTTP Proxy with three modes:
No proxy, Auto-detect proxy settings (the default; environment proxy as
before) and Manual proxy configuration (HTTP or SOCKS, host and port, a
comma-separated "No proxy for" list with `*` wildcards, optional proxy
authentication). "Check connection…" GETs a URL with the settings as edited
and reports the status or the error.
- ide-core: `ProxySettings` plus routing (no-proxy globs, loopback always
direct, credentials percent-encoded), the current proxy for the process,
curl arguments and child-process variables.
- ide-http: `client::build` / `client::shared` is the one HTTP client
constructor; the HTTP Client requests and the update check go through it.
- ide-tools: curl downloads (tools, JDKs, debug adapters, agent registry,
remote server) follow the setting.
- A manual proxy reaches child processes (language servers, debug adapters,
terminals, git) as HTTP_PROXY / HTTPS_PROXY / ALL_PROXY / NO_PROXY.
- The proxy password is kept in the owner-only saved passwords file
(`proxy-password.json` in the app data directory), not in settings.toml.
Co-Authored-By: Claude <noreply@anthropic.com>
* fix(host): Allow the Unix-only post-start test helpers on Windows
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
ide-lsp returned the same `unavailable` error whether no server was running or a running server didn't implement the request, so the UI said no language server was available, and Invoke Inline Completion said no language server provides inline completion, while jdtls was running fine (it doesn't implement textDocument/inlineCompletion). When servers are running but none supports the feature, the error now names them ("jdtls doesn't support inline completion"), and every explicit action shares one message helper: Invoke Inline Completion, go to declaration / implementation / type declaration, quick documentation, parameter info, rename, reformat, intention actions, highlight usages, call / type hierarchy, find usages, safe delete and workspace symbols. With no server running it still says no language server is available. Background requests while typing stay silent.
Fixes#186
Claude-Session: https://claude.ai/code/session_01Dq6jQQn8ZhRfkkyj3PRutA
arookieofcanda-rookie-of-C-language
l0ng-ai
arookieofcanda-rookie-of-C-language
GuoYuandguoyucode
DIMAS FEBRIYANTO
Mykhailo Smolinskyiandnntndfrk
Dimas Victoranddvictor357
Walter Wangandbghtyu
JawandClaude Opus 5.5