Skip to content
- #172: Optimize indexing for packages with many targets — Split indexes into reusable per-crate shards and made auxiliary-target analysis lazy, drastically reducing indexing time, memory use, and cache size for workspaces such as Tokio and Bevy.
- #176: Model extern items and the
proc_macro crate — Added indexing for extern blocks and functions and included the proc_macro sysroot crate, resolving more imports and symbols.
- #178: Support macro-generated modules — Added support for modules declared by macros, with stronger
#[path] handling and protection against expansion and lowering cycles.
- #180: Support build scripts — Added support for generated sources by discovering and reusing existing Cargo build-script outputs without running build scripts itself.
- #181: Use an older glibc for releases — Built Linux release binaries against an older glibc baseline for compatibility with the platforms supported by VS Code.
- #183: Add Windows support — Added Windows support across path handling, caching, formatting, editing, CRLF line endings, and LSP comparison tests.
- #186: Add a Zed extension — Added a Zed extension that downloads the prebuilt server and reports server health and indexing progress.
- #187: Add code actions and improve dirty-buffer analysis — Added initial code actions and taught live overlays to understand newly written impl items, trait context, associated items, and const/static initializers.
- #188: Fix primitive method resolution — Added resolution for inherent and associated methods on primitive types, including macro-generated impl items, while making trait lookup scope-aware and faster.
- #189: Add the VS Code extension icon — Added the rust-glancer project icon to the VS Code extension.
- #190: Add a VS Code extension menu — Replaced the status-bar click-to-restart behavior with a menu for starting or stopping the server, opening logs, and reindexing the workspace.
- #191: Shard def-map and semantic artifacts per crate — Split def-map and semantic artifacts into per-crate shards so queries in packages with many targets no longer need to load enormous shared artifacts.
- #192: Reduce memory fragmentation after initial indexing — Reduced memory retained through fragmentation after cold indexing, bringing post-index idle RSS closer to a fresh warm-cache process.
- #196: Report the server version correctly — Fixed LSP initialization so clients receive the actual rust-glancer server version.
- #193: Disable jemalloc on OpenBSD — Disabled the unsupported jemalloc integration on OpenBSD so rust-glancer can build and run there using the platform allocator.
- #197: Fix CI comments for forked pull requests — Fixed the CI workflow so pull requests from forks no longer fail while attempting to publish sticky comments.
- #199: Lower peak memory and optimize indexing — Added lower-peak-memory batch indexing, compacted def-map scope bindings, and replaced fixed-point scans with a worklist to reduce peak RSS and accelerate indexing.
- #201: Make mimalloc the default allocator — Switched the default allocator from jemalloc to arena-free mimalloc, substantially accelerating indexing and editor queries while preserving low idle memory.
- #203: Do not include generated sources to defmap artifacts - Optimized defmap artifacts
- #204: More candidate filtering and cancellation - Optimized filtering candidates for trait solving and improved query cancellation.
