sorawee
2020-12-10 11:34:37

I feel increasingly bad using any CI services. How much heat do they generate?


sorawee
2020-12-10 11:35:18

I think raco pkg new should ask users interactively if they want to generate a workflow file.


sorawee
2020-12-10 11:38:41

There are also duplicate work — the package server also runs tests in packages.


pocmatos
2020-12-10 12:07:32

@sorawee curious what you think the alternative is. Running your own servers or not regularly testing code?


notjack
2020-12-10 12:39:27

As someone working on the largest CI system on the planet, I am certain that the electricity consumption of all CI systems combined is still dwarfed by the collective electricity consumption of cryptocurrency mining systems.


sorawee
2020-12-10 12:50:53

One of the things I thought about is, if all changed files are scrbl (doc change), then don’t run tests. May still compile to make sure it doesn’t error.


sorawee
2020-12-10 12:52:27

But yeah, perhaps this is optimizing (well, reducing heat) at a wrong spot, and it won’t matter.


notjack
2020-12-10 12:55:57

avoiding compiling dependency packages yourself is an extremely effective way to reduce CI costs


samth
2020-12-10 12:59:34

I think avoiding our large quantities of CI when it’s not needed is the right thing here


samth
2020-12-10 13:00:13

As with documents or particular pkgs etc


notjack
2020-12-10 13:07:38

that reminds me: I keep getting emails for CI failures in my fork of the racket repo, is that because there’s custom runners for the repo that my fork can’t use? and can I turn it off somehow without actually adding any commits to my fork.


sorawee
2020-12-10 13:12:08

If a particular workflow only works for source repo and not forks, I think we should add a test against github.repository_owner (see https://docs.github.com/en/free-pro-team@latest/actions/reference/context-and-expression-syntax-for-github-actions#github-context) to stop the run


pocmatos
2020-12-10 13:39:14

Totally agree with @samth here that we need to minimize testing, building etc when not necessary. I just didn’t get there yet and won’t in 2020. It will be a 2021 task unless someone steps forward first. I hope to post a “Vision 2021” in racket/racketci and retire Vision 2020 soon.


pocmatos
2020-12-10 13:39:57

@notjack yes - the self-hosted runners are assigned to the racket org so you cannot use them, sorry. There should be a way to avoid the error, but I don’t know from the top of my head.


sorawee
2020-12-10 14:05:08

@pocmatos @notjack I’m testing https://github.com/sorawee/racket/tree/disable-fork-self-hosted. If it runs fine, I will submit a PR


pocmatos
2020-12-10 14:28:48

@sorawee thanks


samth
2020-12-10 18:15:06

I think this failing test should be disabled in this instance: https://github.com/racket/racket/runs/1532254497?check_suite_focus=true


sorawee
2020-12-10 18:56:20

What channel is this workflow supposed to send a notification to?


samth
2020-12-10 18:58:07