* feat(jdtls): surface Lombok-generated methods in document symbols
JDTLS' DocumentSymbolHandler filtered out every SourceMethod where
JDTUtils.isGenerated() was true, so Lombok-synthesised methods (getX/setX,
builder(), equals/hashCode/toString, withX, fluent accessors, ...) never
reached the LSP client. As a result Serena's symbol-aware tools — find_symbol,
get_symbols_overview, replace_symbol_body, insert_before_symbol,
insert_after_symbol — could not see, address or edit any Lombok-generated
member; only search_for_pattern / replace_content (text-based) worked.
Upstream introduced an opt-in preference java.symbols.includeGeneratedCode
in eclipse-jdtls/eclipse.jdt.ls#3706 (merged 2026-02-24) and surfaced it
through redhat-developer/vscode-java#4335. First vscode-java release with
the new JDTLS build is v1.53.0 (2026-02-26). Serena was pinned to v1.42.0-561,
well before the fix existed.
Changes:
- New ls_specific_settings.java.lombok_show_generated setting (default true);
forwarded to JDTLS as java.symbols.includeGeneratedCode in initialize
params. Older servers ignore the unknown preference key silently.
- Bump bundled vscode-java DEFAULT 1.42.0-561 -> 1.54.0-923 (brings JRE
21.0.10, Lombok 1.18.39 and Equinox launcher 1.7.100). INITIAL stays at
1.42.0-561, so per the per-version install-dir convention from #1438 the
new build lands in vscode-java-1.54.0-923/ while legacy vscode-java/
caches and any explicit vscode_java_version: '1.42.0-561' override keep
working.
- New VsixResourcePaths dataclass + INITIAL/DEFAULT_VSCODE_JAVA_PATHS pin
the JRE / Lombok / launcher filenames *inside* the VSIX. Without this,
bumping the archive would break users on legacy 1.42.0-561 because
resource paths inside the archive change between releases. Resolution
mirrors SHA pinning: known versions get their pinned paths; any other
version raises SolidLSPException with a pointer to upstream-jdtls mode
(which derives paths dynamically and needs no pinning).
- Drop dead 'static/vscode-java/...' placeholder in initialize params; the
real runtimes[] is overwritten below from runtime_dependency_paths
.jre_home_path, which is correct in both default-VSIX and upstream-jdtls
modes.
- Test fixtures: Lombok provided dep in test_repo/pom.xml plus
LombokModel / FluentLombokModel / DefaultGreeter covering @Data, @Builder
(toBuilder=true), @With, @Delegate, @AllArgsConstructor, @NoArgsConstructor
and @Accessors(fluent=true). Regression test in test_java_basic.py
asserts every generated-method shape lands in document symbols.
Closes#1432
Adds tests for name resolution for each language and fixes the name resolution where tests have failed.
Also:
Extends python tests to cover ty
Extends skipping conditions for several LS (but only when not in CI)
Better test output in CI
Removed a wrong python test (started failing for ty and should have failed for pyright too)
Black dependency created a dependabot warning and couldn't be bumped since it collided with pathspec dependencies. Since ruff can also format, black is not needed. But we now have these small format changes, mostly in test files
a-simeshin