adapter: fix the XREAD wrong-type test's shared timing budget

[bootjp]

TestRedis_StreamXReadBlockChecksWrongTypeAtDeadline is timing-dependent and took down test (ubuntu-latest) on #1222, which touches only internal/snapshotoffload.

Mechanism

The test blocked for 2 s and waited up to 2 s for the stream waiter to register before overwriting the key:

Block: 2 * time.Second,
...
requireStreamWaiterRegistered(...)   // require.Eventually, 2s timeout
require.NoError(t, rdbWriter.Set(ctx, key, "now-a-string", 0).Err())

The wrong-type check runs when the block deadline fires, so a SET that lands after it gets the ordinary block-timeout nil instead:

Error: "redis: nil" does not contain "WRONGTYPE"

Registration and the remaining block window drew on the same 2 s, so a loaded machine could consume the whole window before the SET was sent.

Reproduced, not guessed

Delaying the SET past the deadline produces the identical failure:

--- FAIL: TestRedis_StreamXReadBlockChecksWrongTypeAtDeadline (3.89s)
    Error: "redis: nil" does not contain "WRONGTYPE"

Fix

Block window 6 s, registration wait 1 s — the two no longer share a budget. The registration timeout becomes a parameter, because any caller whose assertion depends on acting before the reader's deadline has to bound it well under that deadline, and it now carries its own failure message, so a slow registration fails as "the waiter never appeared" rather than surfacing later as a confusing nil read.

Risk

None to production: test-only, no production file touched. The test's subject — that the wrong-type check happens at the deadline — is unchanged; only the margin around it is.

Test evidence

• go test ./adapter/ -run TestRedis_StreamXReadBlockChecksWrongTypeAtDeadline -count=5 — ok
• With a 2.5 s delay injected before the SET (the condition that broke it): passes at the new 6 s window, and fails with the original 2 s window — so the widened window is what does the work, not the injected delay.
• golangci-lint --config=.golangci.yaml run ./adapter/... — 0 issues

Self-review

1. Data loss — none; test-only.
2. Concurrency / distributed failures — removes a timing dependency. The registration wait is still bounded, so a genuinely broken waiter registration still fails, now with a clearer message.
3. Performance — the test takes ~6 s instead of ~2 s in the deadline path. It is t.Parallel(), so the cost to the suite is minimal.
4. Data consistency — untouched; the wrong-type detection path is unchanged.
5. Test coverage — same assertions, now deterministic under load.

Fourth timing-dependent test flake in this batch (#1236 carries three others). All four share a shape: an assertion pinning one of several equally-correct outcomes, or a budget shared between setup and the thing being measured.

@codex review
@claude review

https://claude.ai/code/session_013rNHooj7NF3giihWVba8QE

[coderabbitai]

Warning

Review limit reached

Next included review available in 53 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 0bc9de5b-4fb5-423b-af37-8e5a2250acd4

📥 Commits

Reviewing files that changed from the base of the PR and between 4ca7e90 and f1e1027.

📒 Files selected for processing (1)

• adapter/redis_compat_commands_stream_test.go

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-14T08:44:02.801209Z	f1e1027	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f1e10274af

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Allow more time for waiter registration

When a loaded runner needs more than one second to perform the XREAD's initial work and register the otherwise healthy waiter, this newly shortened timeout fails the test before the widened six-second block window can provide any benefit. The original helper allowed two seconds, and the reported flake demonstrates that this setup can be delayed substantially; keep at least that tolerance (while still remaining safely below blockWindow) so the timing fix does not replace the confusing nil failure with a more frequent registration-timeout flake.

Useful? React with 👍 / 👎.

[bootjp]

CI の test (ubuntu-latest) 失敗はこの PR の変更とは無関係でした。TestGRPCS3BlobClusterAuthenticatedCapabilityPushAndFetch が codes.Unauthenticated を期待して codes.Unknown を受け取っていたもので、原因は main に元からあるクライアントストリームのバグです。

gRPC はサーバ側が既に終了させたストリームへの Send を素の io.EOF として返し、実際の status は受信側 (CloseAndRecv) からしか読めません。PushChunkBlob を含む 4 箇所がその io.EOF をそのまま呼び出し元へ返していたため、reset が Send に競り勝った実行でだけ status が失われます。テストの flake ではなく本番のバグで、s3_blob_replicator.go:136 のリトライ判断がこのエラーを見ています。

#1252 で 4 箇所すべてを修正しました。マージ後に main を取り込めばこの CI は緑になるはずです。