fix: Explicitly wait for the server shutdown before killing it forcefully #873
Conversation
| @@ -438,11 +426,35 @@ class UiAutomator2Server { | |||
| await this.adb.killProcessesByName('uiautomator'); | |||
| } catch {} | |||
| } | |||
| await waitStop(); | |||
There was a problem hiding this comment.
What I could remember is this random process end delay could occur after the adb.forceStop also because according to the device's adb log, the process kill related log itself delayed after sending a kill command. I still think we need to ensure the uia2 server has ended in cleanupAutomationLeftovers as the method's purpose.
There was a problem hiding this comment.
Let me search my past logs if the this.adb.processExists was reliable on such env. Maybe I have some research comment somewhere.
There was a problem hiding this comment.
I couldn't find a good log if this.adb.processExists is reliable, but I think it should be ok.
So far, I observed the process kill delay in adb logcat before, so we need to assume this.adb.forceStop itself could have delay to kill the process on the OS internal level. But I assume this.adb.processExists itself should be ok instead of /status endpoint check
There was a problem hiding this comment.
thanks for checking it.
Basically, this is what I figured out so far by analyzing the server code:
after receiving the shutdown request (which is sent as part of the DELETE session request) the server will asynchronously:
- shutdown http listener and other side servers, like mjpeg broadcaster
- exit the main Junit runner entry point
- exit the server process, which would also stop the instrumentation monitoring
The previous check (GET /status) was only verifying the first stage completion, while we actually need to wait for the last one. On my test phone the time difference between the first and the last stage completion is about 400ms.
There was a problem hiding this comment.
Yea, this PR idea looks good. Let me see HS's device pool, which could be good for this testing purpose. The original issue was Samsung Android 13+, so probably I could find a device
## [4.1.4](v4.1.3...v4.1.4) (2025-03-20) ### Bug Fixes * Explicitly wait for the server shutdown before killing it forcefully ([#873](#873)) ([d3f50ef](d3f50ef))
|
🎉 This PR is included in version 4.1.4 🎉 The release is available on: Your semantic-release bot 📦🚀 |
No description provided.