Replace the automatic timed banner rotation with a random initial banner
selection and small, manually operated arrow buttons for navigation.
Previous behavior can be restored by setting BannerRotation.automaticRotationEnabled to true.
The original slides used position:absolute + opacity for transitions.
This was changed to display:none/block because absolutely positioned
children don't contribute to their parent container's dimensions, causing
the banner containers to collapse to zero width/height. The arrow buttons
(position:absolute, top:50%) need a parent with real dimensions to
center correctly. With display:none/block, the active slide determines
the container's natural size.
The fixed height:60px on .gold-banner was removed for the same reason —
it clipped banner images that are taller than 60px.
An explicit height was added to .header-banner to ensure the arrow
buttons can vertically center via top:50% within the platinum banner area.
The test wrote two values to a file and relied on a 3-second sleep for
the filesystem mtime to advance, which fails when both writes land in
the same second (ext4/tmpfs have 1-second mtime granularity).
Fix: use os.utime() to explicitly bump mtime after the second write.
This is deterministic, instant, and tests the actual cache invalidation
logic (mtime comparison) rather than hoping the clock advances.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add "Per-Project Configuration" section to JetBrains plugin docs
with cross-reference anchor for linking from other pages
- Add language_backend to project configuration feature list in workflow docs
- Add cross-reference from global config docs to per-project override
- Move language_backend field from ProjectConfig to SharedConfig,
following the same pattern as symbol_info_budget (review feedback)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The error is raised through `LanguageServerManager.get_language_server`, where it is done conservatively (only if multiple LS are available and we can't decide)
Determine the effective language backend at startup:
- If a startup project has a per-project override, use it
- Otherwise use the global config
Post-init activation of a project with a different backend returns
an error with workarounds (startup activation or separate MCP server
per backend). No dynamic tool swapping — the exposed toolset is fixed
for the session.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add SystemVerilog language server support using slang-langserver
- Add Language.SYSTEMVERILOG enum with .sv, .svh, .v, .vh file support
- Create SystemVerilogLanguageServer class with auto-download for Linux, macOS, Windows
- Add pytest marker and basic test cases for symbol retrieval
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Fix ruff lint errors by updating configuration
- Add UP042, PLW0108, PLC0207 to ruff ignore list (style preferences)
- Update test repo pyproject.toml to use lint.select and lint.isort sections
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Replace slang-langserver with verible-verilog-ls for SystemVerilog
slang v7.0 release URLs return 404 as the project stopped providing
pre-built binaries after v3.0. Migrate to Verible (CHIPS Alliance)
which actively maintains binary releases for all major platforms.
Changes:
- Update RuntimeDependency URLs to Verible v0.0-4051-g9fdb4057
- Add Linux ARM64 platform support
- Update fallback and error messages
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Add SystemVerilog language server support using slang-langserver
- Add Language.SYSTEMVERILOG enum with .sv, .svh, .v, .vh file support
- Create SystemVerilogLanguageServer class with auto-download for Linux, macOS, Windows
- Add pytest marker and basic test cases for symbol retrieval
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Replace slang-langserver with verible-verilog-ls for SystemVerilog
slang v7.0 release URLs return 404 as the project stopped providing
pre-built binaries after v3.0. Migrate to Verible (CHIPS Alliance)
which actively maintains binary releases for all major platforms.
Changes:
- Update RuntimeDependency URLs to Verible v0.0-4051-g9fdb4057
- Add Linux ARM64 platform support
- Update fallback and error messages
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Fix Verible binary paths for macOS and Windows
The archive directory names differ per platform:
- Linux: verible-v0.0-4051-g9fdb4057/bin/
- macOS: verible-v0.0-4051-g9fdb4057-macOS/bin/
- Windows: verible-v0.0-4051-g9fdb4057-win64/ (no bin/ subdirectory)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Update docs and comments: slang-langserver → verible-verilog-ls
- Update pyproject.toml marker description
- Update ls_config.py docstring
- Update test file docstrings
- Remove unused _slang_available() function
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Add verible-verilog-ls to CI and improve SystemVerilog LS
- Add verible-verilog-ls installation step to pytest.yml for
Linux/macOS/Windows (GitHub Releases, follows ZLS pattern)
- Remove unused threading import and server_ready Event
- Add capability validation (assert textDocumentSync, warn on
missing documentSymbolProvider/definitionProvider)
- Add window/logMessage notification handler
- Add codeAction, formatting, documentHighlight capabilities
- Improve error message with conda/Homebrew/GitHub install methods
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add verible detection tests and remove redundant ALU tests
Add comprehensive unit tests for verible-verilog-ls binary detection
logic (PATH lookup, version logging, graceful degradation, download
fallback). Remove duplicate ALU symbol tests that tested the same
LSP operations already covered by counter tests.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Fix SystemVerilog LSP convention deviations from code review
- Add client/registerCapability handler (all other 31 LSPs have it)
- Add locale: en to initialize params (project convention)
- Change log.debug to log.info for window/logMessage handler
- Fix dead code in test_find_references_cross_file skip/assert logic
- Move repeated imports to module top level in detection tests
- Remove unused IS_WINDOWS/IS_UNIX variables and sys import
- Fix hardcoded line number comment to semantic description
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Fix Windows CI verible path: flat zip structure has no bin/ subdir
The Windows verible zip extracts to verible-VERSION-win64/ with
executables directly in the root (no bin/ subdirectory), unlike
Linux/macOS tar.gz which have bin/ subdirectory.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Fix macOS CI verible path: tar.gz extracts with -macOS suffix
Each platform has different directory naming in verible releases:
- Linux: verible-VERSION/bin/ (no suffix)
- macOS: verible-VERSION-macOS/bin/ (has -macOS suffix)
- Windows: verible-VERSION-win64/ (flat, no bin/ subdir)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Refactor: read Verible version from settings instead of hardcoding
Extract hardcoded version string into _custom_settings.get("verible_version")
with the current version as default. This allows version override via
project.yml without changing source code.
Also introduce DEFAULT_VERIBLE_VERSION constant in detection tests
for consistency.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add rename tests and strengthen existing SystemVerilog tests
- Add TestSystemVerilogRename: within-file, cross-file, module rename
- Strengthen all existing tests with exact position/value verification
- Remove all pytest.skip, replace with assert (review feedback)
- Remove isinstance check in _find_symbol_by_name (review feedback)
- Extract _extract_hover_text and _extract_changes helpers
- Remove test_diagnostics (requires shared ls.py fix, separate PR)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Rationale: requesting the info is just not worth it. The risk of the tool being slow or the info bloating the answer unnecessarily are too high. If an agent needs more info about the found references, it can get this info by itself in subsequent tool calls,
request_full_symbol_tree was opening every file in the language server
(sending didOpen + didClose notifications) even when document symbols
were already cached. For a ~600 file Kotlin project, this caused 1198
unnecessary pipe writes to the LSP process, taking ~9 seconds.
By passing open_in_ls=False, files are only read locally for cache
validation. The LSP is only notified when a cache miss requires
re-parsing. This reduces full-project symbol tree time from ~9.5s
to ~0.5s (19x improvement) on cached runs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat(scala): Add stale lock detection for Metals H2 database
Add detection and automatic cleanup of stale Metals H2 database locks
to improve multi-instance support. When a Metals process crashes without
proper cleanup, it can leave a lock file that prevents proper AUTO_SERVER
coordination with new instances.
Changes:
- Add metals_db_utils.py with detection logic for stale locks
- Add StaleLockMode enum (auto-clean/warn/fail) for configurable behavior
- Add MetalsStaleLockError exception for fail mode
- Integrate stale lock checking into ScalaLanguageServer init
- Add configuration options: on_stale_lock, log_multi_instance_notice
- Add comprehensive unit tests for detection logic
This enhancement allows Serena's Metals to coexist with VS Code or other
editors running Metals on the same project, with proper handling of edge
cases like crashed processes.
* test(scala): Add integration tests for stale lock handling modes
Add 7 integration tests for ScalaLanguageServer stale lock detection:
- AUTO_CLEAN mode cleans stale locks and proceeds
- WARN mode logs warning without cleanup
- FAIL mode raises MetalsStaleLockError
- ACTIVE_INSTANCE logs info when notice enabled
- ACTIVE_INSTANCE silent when notice disabled
- NO_DATABASE proceeds silently
- NO_LOCK proceeds silently
Tests use mocking to avoid requiring Java/Coursier/Metals.
* docs(scala): Add multi-instance and stale lock documentation
Document Metals multi-instance support via H2 AUTO_SERVER:
- Explain how multiple Metals instances share database and Bloop
- Document memory considerations (~500MB per JVM)
- Add stale lock detection configuration options
- Document on_stale_lock modes (auto-clean, warn, fail)
- Add YAML configuration examples
* docs: Clean up docs
Removed memory considerations and additional configuration sections from the Scala setup guide.
* refactor(scala): Extract settings helper and add module-level constants
- Add DEFAULT_METALS_VERSION, DEFAULT_CLIENT_NAME, DEFAULT_ON_STALE_LOCK,
DEFAULT_LOG_MULTI_INSTANCE_NOTICE constants to eliminate duplication
- Extract _get_scala_settings() pure function for centralized settings parsing
- Remove redundant cleanup_stale_lock import
- Add warning log when AUTO_CLEAN mode fails to remove stale lock
- Simplify lock path handling with ternary expression
This refactoring improves code maintainability by centralizing Scala
settings extraction and applying functional programming principles.
* chore(scala): fix typo in docstring and comments
Michael Panchenko