🔧 Check for mac updates only

[WangEdward]

Attempt to fix #491, #336 (and some other dup issue)

This idea is from https://github.com/mangerlahn/Latest(Thanks!). They also added "macSoftware" param, which I believe is unnecessary for us as it is not possible to upgrade an iOS-only app from command-line. (Not 100% sure, if that causes problem, we can add the param in the future)

[antifuchs]

Just tested this manually, using curl 'https://itunes.apple.com/lookup?id=1508732804&entity=desktopSoftware' | jq '.results[0].version' (Soulver 3 is one of the apps that have a different version on the desktop and on iOS), and that returns the correct version number. I think this might finally solve the issue of phantom-outdated versions for real!

[rmehner]

Can confirm that this fixes things for me as well. Thank you @WangEdward!

[iamrecursion]

Glad to see a fix is in the works! I ran into this one recently.

[rmehner]

@phatblat sorry for pinging you directly, but since the last commits all have been made / merged by you, I figured you might be the current maintainer of this.

What does it take to get this PR over the finish line? Happy to help with anything you need. Thank you!

[pawisoon]

Can we get this merged please? @phatblat

[nicerloop]

A custom formula for homebrew is available as nicerloop/nicerloop/mas in https://github.com/nicerloop/homebrew-nicerloop
Feel free to test.

[WafaAmr]

Can we get this merged please? @phatblat

[rgoldberg]

Looks good

[rgoldberg]

I haven't dealt with GitHub workflows before.

The workflow seems stuck at:

Build, Test, and Lint Expected — Waiting for status to be reported

Must we "Bump actions/checkout from 3 to 4" via #500 to unstick the workflow?

I've seen suggestions to amend the PR's commit, but I'd prefer to avoid messing with the git commits if It can / must be fixed via the GibHub interface / configuration.

[rmehner]

Hi @rgoldberg,

thanks for having a look, I'm excited to see this finally going forward again :)

Build, Test, and Lint Expected — Waiting for status to be reported
Must we "Bump actions/checkout from 3 to 4" via #500 to unstick the workflow?

This might solve the issue, if you rebase this branch then. If I'm not mistaken, it's merely an issue if retriggering the workflow that got stuck for whatever reason. The easiest way to do this is to add another commit to the PR / rebase the branch and then push again.

[GreyTeardrop]

According to GitHub docs, first-time contributors creating PRs from their pubic forks need maintainer approval before CI kicks in: https://docs.github.com/en/actions/managing-workflow-runs/approving-workflow-runs-from-public-forks.

[rgoldberg]

According to GitHub docs, first-time contributors creating PRs from their pubic forks need maintainer approval before CI kicks in: https://docs.github.com/en/actions/managing-workflow-runs/approving-workflow-runs-from-public-forks.

@GreyTeardrop Thanks for the info. I don't see "Workflow(s) awaiting approval" anywhere, so I assume that I'm not authorized to approve workflows. I'm only a Member; the only Maintainer is @phatblat.

[rgoldberg]

@WangEdward @rmehner @GreyTeardrop

Thanks for the suggestions. I think that either @WangEdward should follow @rmehner's suggestion (potentially doing what's suggested at https://stackoverflow.com/a/58190576/2030088) or that I should try to make my own PR duplicating @WangEdward's one. I might have more success because I've already contributed & am a project Member, so I might be authorized to start workflows from my PRs, and because I'll use a branch other than main.

@WangEdward You choose what you want to do, as it's your PR. If I don't hear back from you in 48 hours or so, I'll just make my own PR, in which I'll credit you, just to keep everything moving.

Thanks for the fix.

[WangEdward]

I force pushed it, and I can see the workflows are now in a state of awaiting approval. However, I don’t think there is any self-hosted CI machine attached to this repo, and the workflow will fail just like https://github.com/mas-cli/mas/actions/runs/6073669475 .

mas/.github/workflows/build-test.yml

Lines 35 to 37 in 4c044d8

	# https://docs.github.com/en/actions/hosting-your-own-runners/using-self-hosted-runners-in-a-workflow
	# https://github.com/mas-cli/mas/settings/actions/runners
	runs-on: [self-hosted, macOS]

[rgoldberg]

@MikeMcQuaid We're working on getting this merged & released. The regular maintainers are MIA, but I'm a project member (with no knowledge of the project's workflows, release procedures, etc.). I'll let you know when it's released so you'll be able to revert Homebrew/homebrew-bundle#1231.

[rgoldberg]

@WangEdward Do you mind if I make my own PR using a non-main branch to see if it works better for me?

[WangEdward]

@WangEdward Do you mind if I make my own PR using a non-main branch to see if it works better for me?

Please feel free to do so. I hope that will work.

[MikeMcQuaid]

@rgoldberg thanks for the heads up. keep me posted!

[rgoldberg]

I don't see any runners listed on https://github.com/mas-cli/mas/actions/runners?tab=self-hosted.

I have no idea why self-hosted was specified without any currently extant runners. Maybe they previously existed but are now gone.

If anyone knows GitHub workflows, actions, etc., help will be appreciated.

@phatblat Any ideas? Is/was @ldgmini somehow related to the self-hosted runner? What does @masclibot do?

[antifuchs]

If this repo works like most repos on github, running the CI jobs will be blocked until a repository admin / owner approves running the workflow. That's independent of any base branch to the PR.

This behavior is supposed to prevent exfiltration of github actions secrets (correction: apparently it's to prevent arbitrary "extra" code like crypto miners from being run as part of a test suite? There's lots of reasons not to let The Internet submit code that gets automatically started I guess). But as we have all witnessed, it results in an unknown in-between state where nobody knows what's wrong if the repo owner isn't around to press the button.

[GreyTeardrop]

@rgoldberg I was able to set up a self-hosted runner on my laptop with scripts from https://github.com/mas-cli/m1-github-actions-runner, attach it to my fork, recreate this PR in that fork, and run some CI for it. The CI run failed, probably because my laptop is on Sonoma and something has to be tweaked for it to properly build on Sonoma (oddly enough, the step that failed is "Lint", not "Build").

[rgoldberg]

@GreyTeardrop I was busy until now. Thanks again for your efforts.

I don't have an Apple Silicon Mac (I'm still running an Intel Mac), so I might not be able to setup a working runner.

Do you have any local logs for the CI run on your computer? Specifically for the Lint step?

It's possible that the problem is that swiftlint wasn't installed:

The Brewfile (which I assume is used by script/bootstrap) doesn't install swiftlint (which is run in script/lint), but it isn't included in macOS 12.7.1 (I don't know about macOS 13+). A comment mentions that it's "Already installed on GitHub Actions runner", but it's likely not already installed on your machine. If you uncomment brew "swiftlint", maybe the Lint step will succeed.

If that doesn't work, you can probably just make script/lint a no-op by deleting everything in the file. If that works, I think it would be fine for us to "skip" the Lint step for now since this PR's change is so small, and since everyone is waiting on it.

The other commands used in script/lint (besides swift and git, but they should both be included in macOS in /usr/bin/) are listed in Brewfile. swift & shellcheck are run, but are not listed in the linter list. I don't think that shellcheck is included with macOS (I didn't find it in macOS 12.7.1), but it's in Brewfile, so it should still be available to script/lint.

Also note that Makefile runs swiftenv, which must be installed from the kylef/formulae tap. I don't know why that would affect the Lint step, but not others, so I assume that's not the problem.

I saw some announcements about GitHub now providing Apple Silicon extra large runners, but I doubt those would be available for free usage. I haven't seen anything about smaller runners. If anyone knows if GitHub provides any free Apple Silicon runners, please let everyone know. I can probably look into that tomorrow.

[rmehner]

@rgoldberg if you need a macos runner, you might be able to use the one https://tart.run provides. Example on how to do that is right on their landing page (the diff).

[rgoldberg]

@rgoldberg if you need a macos runner, you might be able to use the one https://tart.run provides. Example on how to do that is right on their landing page (the diff).

@rmehner thanks for the info, but I don't have Apple Silicon; I'm still using Intel. It might useful for someone else who has Apple Silicon and who is willing to provide a runner.

[rmehner]

@rgoldberg tart provides the runner, so it doesn't run on your local machine.

[rgoldberg]

@rmehner It reads to me like Tart is virtualization software while Cirrus Runners is a service using Tart that provides GitHub Runners. While usage of Tart is free for fewer than 100 cores, Cirrus Runners cost $150 / month each. Maybe I'm mistaken, but that's what it reads like to me. Do you have any link to any information that shows otherwise?

[GreyTeardrop]

Do you have any local logs for the CI run on your computer? Specifically for the Lint step?

It doesn't seem to write any logs locally, apparently everything meaningful is streamed to the Actions console:

> ./start.sh

√ Connected to GitHub

Current runner version: '2.311.0'
2023-11-07 21:41:17Z: Listening for Jobs
2023-11-07 21:41:40Z: Running job: Build, Test, and Lint
2023-11-07 21:47:07Z: Job Build, Test, and Lint completed with result: Failed
^CExiting...
Runner listener exit with 0 return code, stop the service, no retry needed.
Exiting runner...

It's possible that the problem is that swiftlint wasn't installed:

Based on what I see in the script, it should install swiftlint automatically if one isn't present.

I'll see what else can be done a bit later.

[rgoldberg]

Probably simplest to just delete everything from, or do whatever else necessary to, script/lint to make it a no-op.

Based on what I see in the script, it should install swiftlint automatically if one isn't present.

Thanks. I didn't continue reading script/bootstrap after the call to brew bundle …, nor did I search for swiftlint after seeing the commented line in Brewfile. It's odd that swiftlint is handled differently instead of through Brewfile. You would think that brew bundle could handle a formula that is already installed…

[phatblat]

Sorry for the delay everyone. I've been swamped with work lately and haven't had time for open source contributions. I'm working on testing this now.

[GreyTeardrop]

GitHub made M1 runners available for free for open-source projects: https://github.blog/changelog/2024-01-30-github-actions-introducing-the-new-m1-macos-runner-available-to-open-source/ 🎉

[rgoldberg]

@GreyTeardrop See #508

[phatblat]

@rgoldberg if you need a macos runner, you might be able to use the one https://tart.run provides. Example on how to do that is right on their landing page (the diff).

@rmehner I love tart! Cirrus Labs' tart runners cost $150 per month. #509 switched back to GitHub-hosted runners, free for open source projects like this one.

[astrostl]

If I understand correctly, a fix for the long-standing fake upgrade problem already exists and is merged, but not released. What is standing in the way of that? Thanks!

[tonyarnold]

@phatblat can I help out with administrative load here? I'm happy to do PR review, releases, whatever helps to keep things moving while you're busy.

[phatblat]

@tonyarnold That would help immensely. I'm traveling in Japan for another couple weeks.

I worked up a release action in #510 but had trouble testing it with a pre release. #506 is the next action item but that was hard to delegate to someone without homebrew experience.

[tonyarnold]

No problems, I'll get a brew allowlist into place in the local tap, which should help with the workflow.