mirror of
https://github.com/anomalyco/opencode.git
synced 2026-06-02 06:16:48 +02:00
Migrate MCP config tests to instance fixtures (#28338)
This commit is contained in:
@@ -80,6 +80,7 @@ Repeated setup work, long sleeps/timeouts, serial integration tests, filesystem/
|
||||
| Remaining simple config load cases can use Effect-aware instance fixtures | Migrated default config load and legacy TUI-key cases to `it.instance` | 7.78s | 6.39s | keep | Single baseline before edit; after median from three sequential reruns (5.76, 6.39, 6.53). Keep as cleanup with cautious timing. |
|
||||
| Managed settings config cases can use Effect-aware instance fixtures | Migrated managed override and missing-managed-file cases to `it.instance` | 2.40s | 1.76s | keep | Single baseline before edit; after median from three sequential reruns (1.75, 1.76, 1.80). |
|
||||
| Local plugin and subagent config fixtures can use Effect-aware instance fixtures | Migrated scoped npm plugin and custom subagent markdown cases to `it.instance` | 2.37s | 1.67s | keep | Single baseline before edit; after median from three sequential reruns (1.66, 1.67, 1.67). |
|
||||
| MCP merge config cases can use Effect-aware instance fixtures | Migrated three MCP merge/override cases to `it.instance` | 1.98s | 1.95s | keep | Neutral timing within noise; removes manual `tmpdir` + `withTestInstance` setup from isolated filesystem-only config cases. |
|
||||
|
||||
## Profiling Results
|
||||
|
||||
@@ -135,3 +136,4 @@ Full-suite sanity checks:
|
||||
| Socket reset retry test can shorten its idle-timeout path | Reduced Bun server idle timeout and tried forced server close | 16.46s | failed | discard | Shorter idle timeout changed the error shape; forced close hung. Keep the real socket reset. |
|
||||
| `tool/webfetch` can avoid per-test instance setup | Switched local HTTP tests from `it.instance` to `it.live` | 1.219s | failed | discard | Tool execution reads instance-local agent state, so the temp instance is required. |
|
||||
| LSP client interop tests can shorten coarse request-handling sleeps | Reduced fixed post-notification waits from 100ms to 10ms | 4.270s | 4.740s | discard | First run improved to 3.870s but verification was slower than baseline; not a clear win. |
|
||||
| Config content env cases can use Effect-aware instance fixtures | Migrated two `OPENCODE_CONFIG_CONTENT` token substitution cases to `it.instance` | 1.95s | 2.06s | discard | Passing but not neutral-or-better in focused reruns; keep existing explicit env cleanup. |
|
||||
|
||||
Reference in New Issue
Block a user