does not work on macOS

[michaellilltokiwa]

Cleaned label: build docs
cat: /tmp/eco-ci/timer-step.txt: No such file or directory
/Users/runner/work/_actions/green-coding-solutions/eco-ci-energy-estimation/86f1b2ee12db687bca0d15160a529bb64a7b60d9/scripts/make_measurement.sh: line 44: 17308967776N: value too great for base (error token is "17308967776N")
Error: Process completed with exit code 1.

[michaellilltokiwa]

Seems like the date command behaves differently on macOS

[ArneTR]

Thanks for submitting an issue!

Jap, we sadly do not support macOS at the moment. For reference, this is also stated in: https://github.com/green-coding-solutions/eco-ci-energy-estimation?tab=readme-ov-file#limitations--compatibility

Did you spot a mismatch somehwere , where we may false-advertised that macOS is supported? If so please send us a link so we can update that.

2. In order to make macOS supported the following would need to be done:

• Create an energy model for the underly macOS hardware with the output format of https://github.com/green-coding-solutions/cloud-energy
• Patch the relevant files where the Linux only date +%s%N and date +%s%6N is used

If you are keen on supporting this project and taking on a PR we are happy to help. Let me know :)

[ribalba]

I think this can be closed

[ArneTR]

I would keep it open. I would see this as an open issue. Even if we currently have no timeframe when to support it

[ArneTR]

Hey @michaellilltokiwa ,

we add first beta support for macOS!

If you are still keen on trying out we would really like you as beta tester and feedback provier. Any interest ? :)

https://github.com/green-coding-solutions/eco-ci-energy-estimation/releases/tag/v4.3.0

[michaellilltokiwa]

@ArneTR Cool, I tested it:
https://metrics.green-coding.io/ci.html?repo=tokiwa-software/fuzion&branch=4658/merge&workflow=51514044&start_date=2025-01-16&end_date=2025-01-23

Results seem to be way off though?
138 million Joule would mean 38 kwh. That would be excessive for 30min of tests.

[ArneTR]

Thanks for trying it out so quickly!

Yeah, the numbers are way off :) Partially expected, as in the beta we have so far not set a reasonable power profile, so it should default to a full bare metal box.
But also there is at least one bug that I already spotted. Will dig deeper and update once we have the next patch ready.

Thanks for the support!

[ArneTR]

We just pushed a new version with:

• matching power profiles
• matching embodied carbon values

And I fixed the bug that was causing these absurd high values. Happy for feedback!

Commit hash for newest version on main is: 683e179

[ArneTR]

One addon: Sadly GitHub Actions has now way of determining the selected runner at runtime.

Thus, since you are not using the default linux runner you must tell Eco-Ci which runner you are using.

See how we do it here:

eco-ci-energy-estimation/.github/workflows/test.yml

Line 66 in 683e179

machine-power-data: "macos-14-mac-mini-m1.sh"

The flag is "machine-power-data" and in your case must be "macos-14-mac-mini-m1.sh"

[michaellilltokiwa]

@ArneTR The numbers seem to more realistic now, 40000J for one run. Thanks for investigating.

It would be cool to be able to get details on how the numbers where calculated, especially which assumptions are used.

Also, personally I have more of a feeling for kwh, Joule I have to first convert. Maybe add that info?

[ArneTR]

Jap, that's a good idea.

I will add this functionality in two places:

• badges
• PR/Table display

Find the Issues tracking it here:

• Add option to show Wh/kWh instead of Joules #113
• Badges should be able to show energy in kWh green-metrics-tool#1061

2. The assumptions made are all documented. Find them here:

• Parameter assumptions for GitHub Shared runners
• Assumptions and data sources of the underlying non-linear ML model

In case you feel that macOS support is sufficient now please close the issue. Or follow up with a request :)