Reloading config bug in clash-rs

[xmh0511]

https://github.com/libnyanpasu/clash-nyanpasu/issues/3536#issuecomment-3314291700

[ibigbug]

@xmh0511 do you have the error logs?

I think the outbound proxy groups and providers get reloaded at https://github.com/watfaq/clash-rs/blob/75ca7aab46e57a50c156ba93e89b39c7074ab272/clash-lib/src/lib.rs#L360

how does the error in GUI get reproduced?

[xmh0511]

@xmh0511 do you have the error logs?

I think the outbound proxy groups and providers get reloaded at https://github.com/watfaq/clash-rs/blob/75ca7aab46e57a50c156ba93e89b39c7074ab272/clash-lib/src/lib.rs#L360

how does the error in GUI get reproduced?

clash-nyanpasu.2025-08-29.app.zip

This is the log file generated during the switching of different airports, with both the debug level of the log for the clash core and Nyanpasu

[Slinetrac]

@ibigbug Here’s what I noticed from the logs, hope this helps:

• GeoIP failure

[clash-rs]: WARN clash-lib\src\app\router\rules\geoip.rs:30: GeoIP lookup failed: MMDB not available.

Seems to fall back to DIRECT. Maybe related to a missing MMDB file? A simple existence check could make this clearer.

• Connection rejection

[clash-rs]: WARN clash-lib\src\app\dispatcher\dispatcher_impl.rs:201: failed to establish remote connection [TCP] 127.0.0.1:50419 -> cube.weixinbridge.com:443, error: "rejected"

Could be routing issues after the GeoIP failure. Extra debug logs in dispatcher_impl.rs might help confirm.

• Frequent DNS init

[clash-rs]: INFO clash-lib\src\app\dns\dns_client.rs:304: initializing dns client

Saw repeated initialization, which might hurt performance. Maybe caching with Once could avoid repeated init.

• Core termination

[clash-rs]: core terminated with status: TerminatedPayload { code: Some(1) }.

Possibly tied to the above issues. More detailed exit logs in core.rs could help diagnose.

[xmh0511]

@ibigbug

how does the error in GUI get reproduced?

You should have at least two subscribed airports. Using the clash-rs core in Nyanpasu, and just switching from one airport to another, as shown in this picture

[greenhat616]

I took a glance at current context. This Error message is not related, which is caused by a FetchError via fetching DELETE /connections endpoint.

• The core has stopped, so the endpoint is down

The core terminated with status code 1, in Windows, it possible it is a proactive exit from clash-rs.

• https://github.com/search?q=repo%3AWatfaq%2Fclash-rs%20exit&type=code

---

In this case,

Possbilely, we lost the eprintln! log from clash-rs.

Maybe we have closed the stdout/stderr logger too early before the stdio stream fullly closed.