doctor: false negative for v2ex (SSL EOF error) while API is reachable via curl

[BreezeDoo]

Summary

agent-reach doctor --json reports v2ex as unreachable with an SSL error, but the V2EX API is fully reachable via curl. The SSL failure is in agent-reach's bundled Python urllib, not in the network.

Environment

• agent-reach v1.5.0 (Windows 11)
• China network, no proxy

What doctor reports

status: warn
name: "V2EX 节点、主题与回复"
message: "V2EX API 连接失败（可能需要代理）：<urlopen error [SSL: UNEXPECTED_EOF_WHILE_READING] EOF occurred in violation of protocol (_ssl.c:1010)>"
active_backend: null

What's actually true

curl reaches the same endpoint successfully (HTTP 200 in ~2.5s, no proxy):

$ curl -sS -m 8 -o /dev/null -w "http_code=%{http_code} time=%{time_total}s\n" \
    "https://www.v2ex.com/api/topics/hot.json" -H "User-Agent: agent-reach/1.0"
http_code=200 time=2.519745s

Likely root cause

The check uses Python's urllib (the _ssl.c reference in the error confirms this), which hits SSL: UNEXPECTED_EOF_WHILE_READING. curl uses a different TLS stack and succeeds. This is a false negative from the Python SSL environment bundled in the agent-reach executable, not a real connectivity problem. This also contributes to the ~32s doctor runtime (the failing check times out slowly).

Suggested fix

• Use a more robust HTTP client / TLS config for the connectivity check (e.g. requests with proper cert handling, or shell out to curl).
• Or catch SSL: UNEXPECTED_EOF_WHILE_READING specifically and retry / fall back instead of reporting the channel as down.

Workaround

Ignore the warn — v2ex is reachable. Use curl directly for V2EX API calls.

Related

• fix: Windows 中文系统上 agent-reach doctor 误判微信公众号渠道不可用 #262 (closed: doctor false negative on Windows Chinese system for WeChat — same Windows + China-network false-negative pattern)

[Panniantong]

Merged #577 adds a narrowly constrained curl fallback only for the exact recoverable Python TLS EOF case, with regression tests and a successful live V2EX smoke.