[GH-ISSUE #159] fix(native): retry failed Xcrun tool resolution instead of caching permanent failures #106

New issue

Closed

opened 2026-10-15 18:21:13 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-15 18:21:13 +00:00

Owner

Originally created by @morluto on GitHub (Jul 14, 2026).
Original GitHub issue: https://github.com/morluto/rea/issues/159

Symptom

XcrunCommandRunner caches resolution attempts in #resolved before the result is known to have succeeded.
If the first resolution attempt for a tool fails (e.g. transient xcrun failure, temporary IO error, or tool path not yet available), that rejected result is memoized and reused for the same tool on later calls.

Evidence

src/native/CommandRunner.ts:100-104 stores resolveTool(tool) in this.#resolved immediately and returns cached promise on subsequent calls.
resolveTool can return err(new NativeCommandFailure(...)) for transient failures (captureProcess failure, non-absolute path, realpath/hash failure).
Cached map entries are never evicted or refreshed in failure paths.

Impact

A temporary outage, path race, or transient xcrun issue can permanently downgrade native inspections for that tool until process restart, even after conditions recover.

Suggested fix

Cache only successful resolutions (and maybe successful tool metadata), or remove failed cache entries and allow controlled retries/backoff for later calls.

Originally created by @morluto on GitHub (Jul 14, 2026). Original GitHub issue: https://github.com/morluto/rea/issues/159 ## Symptom `XcrunCommandRunner` caches resolution attempts in `#resolved` before the result is known to have succeeded. If the first resolution attempt for a tool fails (e.g. transient `xcrun` failure, temporary IO error, or tool path not yet available), that rejected result is memoized and reused for the same tool on later calls. ## Evidence - `src/native/CommandRunner.ts:100-104` stores `resolveTool(tool)` in `this.#resolved` immediately and returns cached promise on subsequent calls. - `resolveTool` can return `err(new NativeCommandFailure(...))` for transient failures (`captureProcess` failure, non-absolute path, `realpath/hash` failure). - Cached map entries are never evicted or refreshed in failure paths. ## Impact A temporary outage, path race, or transient xcrun issue can permanently downgrade native inspections for that tool until process restart, even after conditions recover. ## Suggested fix Cache only successful resolutions (and maybe successful tool metadata), or remove failed cache entries and allow controlled retries/backoff for later calls.

bsakaguchi

2026-10-15 18:21:13 +00:00

closed this issue
added the
bug
label

bsakaguchi referenced this issue

2026-10-16 21:11:37 +00:00

[PR #106] [MERGED] chore(main): release rea-agents 1.3.0 #1837

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

morluto/rea#106

No description provided.

Rows
Columns