fix: MQTT 连接生命周期竞态与事件处理
连接句柄（mqtt.rs）： - ConnHandle 增加代数 generation，事件循环退出只在仍是自己这一代时 才移除句柄——旧循环无条件按 id 删除会把重连后的新句柄删掉， 导致连接显示在线却无法从 UI 断开、发布/订阅报连接不存在 - connect() 同步插入占位句柄（client=None），建连期间也可被取消； 建连完成后仅当句柄仍是本代才升级为可用客户端 - 重连 2 秒退避改为可取消的 select：原先 close() 的取消信号在退避 期间不可见，会对用户已关闭的连接继续真实重连 - 重连成功后清零计数（max_reconnect_times 限流的是连续失败次数）； v4 CONNACK 拒绝上报真实返回码；test_handshake 断开后再 poll 冲出 DISCONNECT 报文；建连的证书文件读取移入 spawn_blocking 消息记录与事件（model.rs / app.rs / connection_view.rs）： - 非文本负载保留原始字节（MqttRecord.raw_bytes，仅非法 UTF-8 时 占用，纯文本零开销），详情 Hex/Base64 不再展示 lossy 失真字节 - EngineEvent::Published 补齐 response_topic/correlation_data/ message_expiry；correlation_data 两侧统一「优先文本、回退 hex」 - 事件入口加连接存在性守卫：删除连接后排队的引擎事件不再复活 状态/消息（内存泄漏）或为已删连接写孤儿订阅 - SUBACK 失败时移除乐观插入的订阅行；成功事件只刷新已有订阅， 不再用默认字段复活 SUBACK 前被删除的订阅 - Connecting 期间侧栏电源按钮防抖；分组名恰为 all/ungrouped 时 chip 元素 id 改前缀避免共享交互状态 - 「清空」按钮按消息总数而非过滤命中数判断禁用；搜索匹配的 大小写快速路径限定 ASCII（U+1E9B 等小写映射非自身的字符） - 应用发布预设后同步模板变量面板，残留行不再写回全局变量 - 日志文件套 BufWriter，高频消息流不再逐条落盘 syscall
This commit is contained in:
+83
-15
@@ -84,7 +84,7 @@ impl TransportKind {
|
||||
|
||||
|
||||
/// MQTT 5 遗嘱消息（3.1.1 下忽略 v5 专属属性）。
|
||||
#[derive(Debug, Clone, Default, Serialize, Deserialize)]
|
||||
#[derive(Debug, Clone, Default, PartialEq, Serialize, Deserialize)]
|
||||
pub struct LastWill {
|
||||
pub topic: String,
|
||||
pub payload: String,
|
||||
@@ -97,7 +97,7 @@ pub struct LastWill {
|
||||
}
|
||||
|
||||
/// SSL/TLS 自定义配置（CA、双向认证、跳过校验）。
|
||||
#[derive(Debug, Clone, Default, Serialize, Deserialize)]
|
||||
#[derive(Debug, Clone, Default, PartialEq, Serialize, Deserialize)]
|
||||
pub struct SslConfig {
|
||||
/// 自定义 CA 文件路径（PEM），作为系统根证书的补充
|
||||
#[serde(default)]
|
||||
@@ -250,15 +250,21 @@ impl ConnectionConfig {
|
||||
|
||||
/// 与另一份配置相比，影响 broker 会话建立的关键参数是否变化。
|
||||
/// 这些字段改动只有在重新建立连接后才会生效，用于提示用户重连。
|
||||
/// WS 路径、遗嘱、TLS 配置、keep_alive 同样随 CONNECT 报文一次性生效，
|
||||
/// 必须纳入比较。
|
||||
pub fn session_params_changed(&self, other: &ConnectionConfig) -> bool {
|
||||
self.host != other.host
|
||||
|| self.port != other.port
|
||||
|| self.path != other.path
|
||||
|| self.transport != other.transport
|
||||
|| self.protocol != other.protocol
|
||||
|| self.client_id != other.client_id
|
||||
|| self.username != other.username
|
||||
|| self.password != other.password
|
||||
|| self.clean_start != other.clean_start
|
||||
|| self.keep_alive != other.keep_alive
|
||||
|| self.last_will != other.last_will
|
||||
|| self.ssl != other.ssl
|
||||
}
|
||||
}
|
||||
|
||||
@@ -398,6 +404,11 @@ impl PayloadFormat {
|
||||
if !cleaned.len().is_multiple_of(2) {
|
||||
return Err("Hex 长度必须为偶数（每两个字符一个字节）".into());
|
||||
}
|
||||
// 必须先做全字节校验：长度检查按字节计数，多字节字符（如中文）
|
||||
// 可凑成偶数字节，直接按字节索引切片会落在字符边界内而 panic。
|
||||
if !cleaned.bytes().all(|b| b.is_ascii_hexdigit()) {
|
||||
return Err("Hex 解码失败: 含非十六进制字符".into());
|
||||
}
|
||||
(0..cleaned.len())
|
||||
.step_by(2)
|
||||
.map(|i| {
|
||||
@@ -435,6 +446,9 @@ pub struct MqttRecord {
|
||||
pub subscription_identifier: Option<u32>,
|
||||
/// payload 超过 [`MAX_PAYLOAD_RETAIN`] 被截断
|
||||
pub payload_truncated: bool,
|
||||
/// 非文本负载的原始字节（合法 UTF-8 时为 None）。文本化会失真
|
||||
/// （非法字节被替换为 U+FFFD），详情面板的 Hex/Base64 需要真实报文。
|
||||
pub raw_bytes: Option<Arc<[u8]>>,
|
||||
}
|
||||
|
||||
/// 消息历史中单条 payload 的保留上限。超限截断，避免个别大报文
|
||||
@@ -442,20 +456,27 @@ pub struct MqttRecord {
|
||||
pub const MAX_PAYLOAD_RETAIN: usize = 256 * 1024;
|
||||
|
||||
/// 把原始 payload 字节转成用于历史记录的共享文本，超限截断。
|
||||
/// 返回 (Arc<str>, 是否截断)。截断时回退到 UTF-8 字符边界。
|
||||
pub fn retained_payload(raw: &[u8]) -> (Arc<str>, bool) {
|
||||
if raw.len() <= MAX_PAYLOAD_RETAIN {
|
||||
(Arc::from(String::from_utf8_lossy(raw).as_ref()), false)
|
||||
} else {
|
||||
/// 返回 (文本, 非文本时的原始字节, 是否截断)。截断时回退到 UTF-8 字符边界。
|
||||
///
|
||||
/// 仅当文本化会失真（非合法 UTF-8，字节被替换为 U+FFFD）时才保留原始
|
||||
/// 字节（同样受 [`MAX_PAYLOAD_RETAIN`] 限制）；纯文本负载零额外开销。
|
||||
pub fn retained_payload(raw: &[u8]) -> (Arc<str>, Option<Arc<[u8]>>, bool) {
|
||||
let truncated = raw.len() > MAX_PAYLOAD_RETAIN;
|
||||
let text_part = if truncated {
|
||||
let mut end = MAX_PAYLOAD_RETAIN;
|
||||
while end > 0 && (raw[end] & 0xC0) == 0x80 {
|
||||
end -= 1;
|
||||
}
|
||||
(
|
||||
Arc::from(String::from_utf8_lossy(&raw[..end]).as_ref()),
|
||||
true,
|
||||
)
|
||||
}
|
||||
&raw[..end]
|
||||
} else {
|
||||
raw
|
||||
};
|
||||
let raw_bytes = match std::str::from_utf8(text_part) {
|
||||
Ok(_) => None,
|
||||
Err(_) => Some(Arc::from(text_part.to_vec())),
|
||||
};
|
||||
let text: Arc<str> = Arc::from(String::from_utf8_lossy(text_part).as_ref());
|
||||
(text, raw_bytes, truncated)
|
||||
}
|
||||
|
||||
// ─── 发布参数 / 发布预设 ──────────────────────────────────────────────────────
|
||||
@@ -897,13 +918,14 @@ mod tests {
|
||||
#[test]
|
||||
fn retained_payload_caps_and_preserves_utf8() {
|
||||
let small = b"hello";
|
||||
let (p, trunc) = retained_payload(small);
|
||||
let (p, raw, trunc) = retained_payload(small);
|
||||
assert_eq!(&*p, "hello");
|
||||
assert!(raw.is_none(), "合法 UTF-8 不保留原始字节");
|
||||
assert!(!trunc);
|
||||
|
||||
// 中文每字符 3 字节，让上限落在字符中间（256K 不是 3 的倍数）
|
||||
let big: Vec<u8> = "中".repeat(MAX_PAYLOAD_RETAIN).into_bytes();
|
||||
let (p, trunc) = retained_payload(&big);
|
||||
let (p, _, trunc) = retained_payload(&big);
|
||||
assert!(trunc, "超过上限必须截断");
|
||||
assert!(p.len() <= MAX_PAYLOAD_RETAIN);
|
||||
assert_eq!(p.chars().last(), Some('中'), "不能切坏 UTF-8 字符");
|
||||
@@ -914,7 +936,53 @@ mod tests {
|
||||
|
||||
// 恰好上限不截断
|
||||
let exact = vec![b'a'; MAX_PAYLOAD_RETAIN];
|
||||
let (_, trunc) = retained_payload(&exact);
|
||||
let (_, raw, trunc) = retained_payload(&exact);
|
||||
assert!(!trunc);
|
||||
assert!(raw.is_none());
|
||||
}
|
||||
|
||||
// 非文本负载：保留原始字节供 Hex/Base64 详情，文本侧 lossy 展示
|
||||
#[test]
|
||||
fn retained_payload_keeps_raw_bytes_for_binary() {
|
||||
let binary = vec![0x00, 0xff, 0xfe, 0x01];
|
||||
let (p, raw, trunc) = retained_payload(&binary);
|
||||
assert!(!trunc);
|
||||
assert_eq!(&*p, String::from_utf8_lossy(&binary));
|
||||
let raw = raw.expect("非法 UTF-8 必须保留原始字节");
|
||||
assert_eq!(&*raw, &binary);
|
||||
}
|
||||
|
||||
// Hex 编码：多字节字符不得 panic（曾按字节切片落入字符边界内而崩溃）
|
||||
#[test]
|
||||
fn hex_encode_rejects_non_hex_without_panic() {
|
||||
assert!(PayloadFormat::Hex.encode("deadBEEF").is_ok());
|
||||
assert!(PayloadFormat::Hex.encode("de ad be ef").is_ok());
|
||||
assert!(PayloadFormat::Hex.encode("abc").is_err(), "奇数长度报错");
|
||||
assert!(PayloadFormat::Hex.encode("zz").is_err(), "非法字符报错");
|
||||
// 偶数字节长度的多字节字符：修复前此处在字符边界内切片 panic
|
||||
assert!(PayloadFormat::Hex.encode("中文").is_err());
|
||||
assert!(PayloadFormat::Hex.encode("😀").is_err());
|
||||
}
|
||||
|
||||
// 会话参数比较必须覆盖所有随 CONNECT 报文一次性生效的字段
|
||||
#[test]
|
||||
fn session_params_changed_covers_connect_time_fields() {
|
||||
let a = ConnectionConfig::default();
|
||||
let mut b = a.clone();
|
||||
assert!(!a.session_params_changed(&b));
|
||||
|
||||
b.keep_alive = a.keep_alive + 1;
|
||||
assert!(a.session_params_changed(&b), "keep_alive 改动应触发重连提示");
|
||||
b = a.clone();
|
||||
|
||||
b.last_will = Some(LastWill {
|
||||
topic: "w/t".into(),
|
||||
..Default::default()
|
||||
});
|
||||
assert!(a.session_params_changed(&b), "遗嘱改动应触发重连提示");
|
||||
b = a.clone();
|
||||
|
||||
b.ssl.ca_file = "/tmp/ca.pem".into();
|
||||
assert!(a.session_params_changed(&b), "TLS 配置改动应触发重连提示");
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user