AGENT SKILL

confirm-crash-and-gather-logs

Guide for what to do when you get a timeout when flashing the board.

License 许可
Apache-2.0
最近更新
2026-10-12
Stars
★ 0
Forks
0

01 前言 Front Matter

name
confirm-crash-and-gather-logs
description
Guide for what to do when you get a timeout when flashing the board.

03 SKILL.md 正文

Confirming Board Crash and Gathering Logs

When to use the skill

Use the skill when you see a timeout when flashing the board. The cause is probably that the board is encountering a fatal error and crashing on boot after the last change.

INVOKE A SUBAGENT

The below troubleshooting procedure should be passed to a subagent. DO NOT ATTEMPT TO CARRY IT OUT YOURSELF. It is too easy to get distracted.

The Procedure

Follow the below procedure exactly.

  • Add a k_sleep(K_MSEC(3000)); before Os::init() in FprimeZephyrReference/ReferenceDeployment/Main.cpp so that there is enough time to capture serial output
  • Run make flash1 to reflash the board.
  • Run timeout 10 udevadm monitor --udev --subsystem=tty to observe additions and removals of tty devices. There should be two serial ports dropping in and out as the device resets.
  • Identify the serial port dropping in and out which has the lower index (e.g. ttyACM0 rather than ttyACM1).
  • Look for any other processes (e.g. running GDS) which might read the port as well. Kill any offending processes.
  • Capture the output on the lower index port with e.g. udevadm wait /dev/ttyACM0 && socat OPEN:/dev/ttyACM0 -
  • The output should have an indication of a fatal error. Record it.
  • Remove the k_sleep
  • Report the captured message. If any of the above steps doesn't go as expected, explain why.
  • Exit. Do not try to troubleshoot any further.

Finally

Continue what you were doing, now with the knowledge of the cause of the crash.