mirror of
https://github.com/pgsty/minio.git
synced 2026-10-06 17:55:59 +03:00
Feng Ruohang
055030ea53
Signed-off-by: Feng Ruohang <rh@vonng.com> (cherry picked from commit 0d48d32d7e038ae1ea5966f3d7e0cb86780a6311) Signed-off-by: Feng Ruohang <rh@vonng.com>
12 lines
3.0 KiB
Markdown
12 lines
3.0 KiB
Markdown
You are the actual independent Opus 5.0 reviewer. Read-only; do not attempt to write any file. The user requires explicit same-version agreement for the COMPLETE R8 scope before the new configuration change is implemented.
|
|
|
|
Review plan v2 at docs/investigations/r8/plan-v2.md, SHA256 426127ed9fb08aeddf8259ebdc4b1c24ebec8cda751a970ed99338a44b065f4c, on baseline 9ebe81c1b3611f9cc73e676b5b741c2be62c467a. V1 connection repair is already implemented after actual v1 CONSENSUS; see consensus.md and review/opus-v1.md. V1 does not cover the newly found CLI binding omission. The only new proposed production edit is ctxt.ReadHeaderTimeout = ctx.Duration("read-header-timeout") beside IdleTimeout in cmd/common-main.go. It has NOT been applied.
|
|
|
|
Evidence: evidence/config-baseline.log and config_baseline_test.go.txt exercise the REAL cli.App with serverCmd.Flags and buildServerCtxt. CLI parses correct default/flag/env/precedence/YAML/negative durations, but context is 0. Explicit zero remains 0. Also read evidence/runtime-v1.json and runtime_probe.py: a compiled v1 SILO binary was launched on disposable localhost-only storage; both flag and env set 100ms, idle 2s, but a 400ms health request header still returns HTTP 200 because missing binding leaves ReadHeaderTimeout=0 and Go falls back to idle. These are real tests by Codex, not claims that you ran them.
|
|
|
|
Read actual cmd/common-main.go buildServerCtxt; cmd/server-main.go ServerFlags, configCommonToSrvCtx, server setup; current DeadlineConn/listener/server files and new tests; review/implementation-opus.md if it now exists (v1 implementation review only); v2 validation matrix and compatibility notes. Independently confirm minimal sufficiency and side effects of binding defaults and custom values, zero/negative semantics, env precedence, YAML retention, header vs idle behavior, keepalive, TLS, h2, body/download renewal and shared callers. Do not trust a patch/plan claim as execution. Inspect Go1.27.1 source as needed.
|
|
|
|
Return Chinese verdict CONSENSUS or REVISE for COMPLETE plan v2, explicitly naming v2 and its SHA256. If consensus, say there are zero blocking disagreements and authorize the proposed configuration binding within that plan, with implementation/testing still separate. List numbered blocking/nonblocking findings with concrete evidence. Any substantive required design change must use REVISE. No rate-limit or missing reply counts as agreement.
|
|
|
|
Additional completed v1 implementation review: review/implementation-opus.md returned REQUEST_CHANGES for one TEST dependency issue (new direct x/net/http2 import while go.mod marked module indirect), explicitly no production behavior defect. Read review/implementation-dispositions.md. The test now uses the standard Go HTTP/2-only Protocols setting, so no external x/net import or go.mod change is needed; h2 assertions remain. Also added repeated-clamp unit coverage and widened only h2 keep-alive idle. Independently decide whether B1 is resolved and whether COMPLETE v2 has any remaining blockers.
|