Not working on M1 Macs / arm64

[HaswinVidanage]

I'm seeing this error thread 'main' panicked at 'failed to get local time offset: IndeterminateOffset', src/main.rs:101:51 followed by a [ERR] 2023-09-07 07:57:10.524 condure: Exited unexpectedly when running the docker image on Mac M1/ arm64.

Can we get support on M1 Macs ?
Thanks.

Full log:

[INFO] 2023-09-07 07:57:10.393 starting...
[INFO] 2023-09-07 07:57:10.402 using config: /etc/pushpin/pushpin.conf
[INFO] 2023-09-07 07:57:10.431 started
thread 'main' panicked at 'failed to get local time offset: IndeterminateOffset', src/main.rs:101:51
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace
[INFO] 2023-09-07 07:57:10.460 [proxy] starting...
[INFO] 2023-09-07 07:57:10.465 [zurl] starting...
[INFO] 2023-09-07 07:57:10.467 [handler] starting...
[ERR] 2023-09-07 07:57:10.524 condure: Exited unexpectedly
[INFO] 2023-09-07 07:57:10.546 [proxy] routes loaded with 1 entries
[INFO] 2023-09-07 07:57:10.552 [handler] zhttp in stream: ipc:///var/run/pushpin/pushpin-condure-out-stream
[INFO] 2023-09-07 07:57:10.554 [handler] zhttp out: ipc:///var/run/pushpin/pushpin-condure-in
[INFO] 2023-09-07 07:57:10.555 [handler] inspect server: ipc:///var/run/pushpin/pushpin-inspect
[INFO] 2023-09-07 07:57:10.555 [handler] accept server: ipc:///var/run/pushpin/pushpin-accept
[INFO] 2023-09-07 07:57:10.557 [handler] control server: tcp://:5563
[INFO] 2023-09-07 07:57:10.558 [handler] in pull: tcp://:5560
[INFO] 2023-09-07 07:57:10.558 [handler] in sub: tcp://*:5562
[INFO] 2023-09-07 07:57:10.559 [handler] retry: ipc:///var/run/pushpin/pushpin-retry
[INFO] 2023-09-07 07:57:10.559 [handler] ws control in: ipc:///var/run/pushpin/pushpin-ws-control-out
[INFO] 2023-09-07 07:57:10.559 [handler] ws control out: ipc:///var/run/pushpin/pushpin-ws-control-in
[INFO] 2023-09-07 07:57:10.559 [zurl] started
[INFO] 2023-09-07 07:57:10.562 [handler] stats: ipc:///var/run/pushpin/pushpin-stats
[INFO] 2023-09-07 07:57:10.563 [handler] proxy stats: ipc:///var/run/pushpin/pushpin-proxy-stats
[INFO] 2023-09-07 07:57:10.563 [handler] proxy control client: ipc:///var/run/pushpin/pushpin-proxy-command
[INFO] 2023-09-07 07:57:10.565 [zurl] stopping...
[INFO] 2023-09-07 07:57:10.566 [zurl] stopped
[INFO] 2023-09-07 07:57:10.567 [handler] http control server: 0.0.0.0:5561
[INFO] 2023-09-07 07:57:10.567 [handler] started
[INFO] 2023-09-07 07:57:10.615 [handler] stopping...
[INFO] 2023-09-07 07:57:10.622 [proxy] started
[INFO] 2023-09-07 07:57:10.623 [proxy] stopping...
[INFO] 2023-09-07 07:57:10.630 [handler] stopped
[INFO] 2023-09-07 07:57:10.646 [proxy] stopped
[INFO] 2023-09-07 07:57:10.655 stopped

[HaswinVidanage]

Also please note that this issue only occurs on the latest version and works fine on version 1.35.0-1

[jkarneges]

I wonder if you can work around this by setting the TZ environment variable?

[chrisbodhi]

The defect seems to stem from the line in Condure here.

There is an issue in the Rust library time that discusses the underlying problem: time-rs/time#296

ETA: setting the TZ env var to 'UTC' does not resolve the issue.

[jannic]

time-rs has a related bug ticket: time-rs/time#500

Unfortunately, it was closed without a solution. After looking at the time-rs source code, I agree with @jhpratt that it's probably some issue with num_threads, or that there are in fact multiple threads, for some reason.

We call current_local_offset rather early, so it's not clear where a second thread could come from. And I don't have a M1 Mac to reproduce the issue. So I can't investigate the root cause.

However, the time-rs docs suggest that callers should always be prepared to handle failure when calling current_local_offset. Do we really need to panic in that case? Wouldn't it be sensible to just log using UTC timestamps, instead?

[jannic]

Possible solution: fanout/condure#15

[chrisbodhi]

Wouldn't it be sensible to just log using UTC timestamps, instead?

FWIW, I think this is a great solution.

[jhpratt]

Please refrain from pinging a user when the sole purpose is to mention them, not alert them of something.

[jannic]

Please refrain from pinging a user when the sole purpose is to mention them, not alert them of something.

I pinged you intentionally, as in time-rs/time#500 you mentioned you were "unable to reproduce due to lack of code", and this might be a situation where code is available and reasonably simple, so it could be an opportunity to investigate why it doesn't work on Mac M1. But I also see that this is of limited use as you also mentioned you don't have that hardware. (I don't, either.)

Therefore, sorry if the ping was annoying to you!

[jkarneges]

This issue has been fixed in 1.38.0 and a multi-platform docker image has been posted!