Coder/deepseek-harness
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(lsp): make transport recovery ownership-safe

Browse Source
This commit is contained in:
Tianyi Cui
2026-07-22 15:56:29 +08:00
parent 2a8e7c661d
commit 769710cfb9
10 changed files with 94 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
packages/lsp/lsp-local/README.md
View File

@@ -7,10 +7,10 @@ Namespace plugin (`name` / `inject` / `Config` / `apply`, no default export).
## What it does
- Resolves every server-local setting before registration; an invalid mapping or registration conflict rolls back earlier entries, so a failed load leaves no provider routes.
- Lazily single-flights one server process per `(server id, canonical workspace realpath)`. A live server error is not replayed; if the transport becomes dead between the pool's liveness check and a read-only query, the provider evicts it and retries that query once on a fresh process.
- Lazily single-flights one server process per `(server id, canonical workspace realpath)`. A live server error is not replayed; if the selected pooled transport fails before or during a read-only query, the provider awaits its disposal and retries that query once on a fresh process.
- Uses a compatibility-first **transient-open** sequence per query: canonicalize and read the source with Node APIs, `textDocument/didOpen` (version 1, full text), the requested request, then `textDocument/didClose` in `finally`. A failed or canceled `didOpen` write terminates the instance before the pool can reuse it. Documents close after each call, so the first version needs no `didChange`, content cache, or document LRU.
- Serializes each source-read/open/query/close lifecycle through one abortable per-workspace queue so queued calls read current source only when their turn starts; distinct workspaces run in parallel.
- After protocol shutdown fails, terminates the server's descendant tree through POSIX process-group signaling or synchronous Windows `taskkill /T /F`, with a direct-child fallback for teardown races.
- After protocol shutdown fails, terminates the server's descendant tree through POSIX process-group signaling or synchronous Windows `taskkill /T /F`. Windows suppresses only taskkill's already-absent-tree result; command, permission, and other tree-kill failures remain visible.
- Reads sources through Node filesystem APIs in the subprocess's host namespace — NOT `ctx.fs`, and emits no `fs/observed`: only the LSP result is model-visible, so a query does not satisfy read-before-write policy.
## Configuration