phargus-memory

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Phargus Memory

Phargus 内存

Identify the observed memory growth and test whether it persists under comparable work. Read memory and data semantics. Use tools for summary and heap analysis.
  1. Resolve this skill's absolute directory. Summarize the recording and establish workload, warmup, GC context and data quality. Request raw NDJSON when external/arrayBuffers are missing.
  2. Separate heap, RSS and external/buffer growth. Inspect comparable workload cycles and observed post-GC baselines before claiming a leak.
  3. Analyze supplied snapshots with the heap command and --project pointing to the app. Compare object-name counts and self sizes; these are not retaining paths or dominators.
  4. If needed, prepare an application-side capture for the user's requested operation. Snapshot/allocation APIs run in the monitored process, and snapshot capture pauses it. Read compatibility before choosing those APIs.
  5. Investigate supported sites/object groups and test the suspected correction over repeated equivalent workload cycles.
Return measured trajectories, capture/analysis limits, hypotheses and the next controlled experiment. Existing evidence may identify external pressure without proving a JS leak.
Example request: "RSS doubles after each batch; use these samples and snapshots to investigate."
识别观察到的内存增长,并在可比工作负载下测试其是否持续存在。 阅读内存和数据语义。 使用工具进行摘要和 heap 分析。
  1. 确定此 skill 的绝对目录。总结录制数据,并明确工作负载、预热情况、GC 上下文和数据质量。当缺少 external/arrayBuffers 时,请求原始 NDJSON。
  2. 区分 heap、RSS 和 external/buffer 的增长情况。在断言存在泄漏前,检查可比的工作负载周期以及观察到的 GC 后基线。
  3. 使用 heap 命令分析提供的 snapshot,并让 --project 指向应用。比较对象名称计数和 self size;这些数据不是 retaining paths 或 dominators。
  4. 如有必要,为用户请求的操作准备应用侧的捕获。Snapshot/allocation API 在被监控进程中运行,snapshot 捕获期间会暂停该进程。选择这些 API 前,请阅读兼容性。
  5. 调查受支持的 sites/object groups,并在重复的等效工作负载周期中测试疑似修正措施。
返回实测轨迹、捕获/分析限制、假设以及下一个受控实验。现有证据可能表明存在 external 压力,但不足以证明存在 JS 泄漏。
示例请求:“每处理一批数据后 RSS 都会翻倍;使用这些样本和 snapshot 进行调查。”