Repository navigation
v1.6.0 — 崩溃看门狗(Crash Watchdog)
崩溃看门狗(Crash Watchdog)
Added
adb_watch_crash工具:监控 crash buffer 里的新真实崩溃(前台轮询或后台 job)- 前台:轮询直到新崩溃出现(
FATAL EXCEPTION / Fatal signal / SIG*)或预算耗尽(默认 60s / 上限 600s),超时返回matched:false而非抛错 - 后台(
run_in_background):挂 ctx.jobs 持续轮询,job_output返回每次新崩溃文本摘要 - seed 去重:启动时记住已有崩溃,只报新出现的;启动标记(mtk-brm)不算崩溃(复用体检分类逻辑)
- 与现场采集衔接:检测到崩溃后配
adb_crash_report/adb_screenshot抓现场——「盯崩溃→采集→归因」链第一环
Verified:单测 92 全绿(新增 8 例:签名/增量/格式 4 + 轮询循环 4,含崩溃出现/不出现/瞬时失败容忍/seed 去重);headless 验证注册 + 无设备时结构化超时返回。崩溃触发正向路径依赖在线设备(模拟器已下线),轮询逻辑已单测全覆盖。
EN: adb_watch_crash — watch the crash buffer for NEW real crashes (foreground poll or background job), seed-deduped, boot markers ignored; matched:false on timeout. The first link of the monitor→capture→attribute chain. 92 unit tests green.