216 Commits

• More tooling updates & fixes (#37177)

  ## [Adjust CI test reporting to upstream action
  changes](https://github.com/ppy/osu/commit/f736337b1acf311bf48fb8c39cae67c3f172608d)
  
  It's been semi-broken since I bumped it a few weeks ago. Paper trail for
  this is at https://github.com/dorny/test-reporter/issues/750, but to
  TL;DR it: `dorny/test-reporter@>=v2.0.0` migrated from [creating new
  check runs via the GitHub
  API](https://docs.github.com/en/rest/checks/runs) to [job
  summaries](https://docs.github.com/en/actions/reference/workflows-and-actions/workflow-commands#adding-a-job-summary)
  by default.
  
  The main difference is that compared to check runs, job summaries do not
  *appear to* require extra `GITHUB_TOKEN` permissions, but in exchange
  are limited to the job that called it. This broke visibility of the test
  reports because due to `GITHUB_TOKEN` permissions foibles the test
  reporter was running in a separate workflow.
  
  I see migrating as a plus here, since:
  
  - The visibility of results is comparable-to-better (example available
  for preview at https://github.com/bdach/osu/actions/runs/23892493152)
  - No longer required to have a completely separate workflow for test
  result reporting
  - No longer required to give `checks: write` permissions to the action
  (I'd hope, we'll see, untested on a public repository with PRs involved)
  
  One downside is there'll be no in-code annotations for failing tests
  anymore but that's whatever IMO. Half the time they weren't even very
  helpful, test results pretty much require maintainer interpretation
  anyway.
  
  This needs to be applied to a few other repos but I'm starting here
  because this is the one where the traffic is highest and therefore
  unbreaking the report is of most value (and also the one where I'll see
  if it works with public PRs the fastest).
  
  Side note, I was hoping to remove the artifact upload/download games by
  just attaching the summary inside each individual test job in the
  matrix, but [it looks like
  crap](https://github.com/bdach/osu-framework/actions/runs/23888384309)
  because only the first three summaries are loaded by default, so if
  there are more, you have to click each remaining one to see its output.
  Wow. Awesome.
  
  Also updates the action to `v3.0.0` to resolve node deprecation
  warnings.
  
  ## [Update inspectcode version to resolve deprecation
  warnings](https://github.com/ppy/osu/commit/496cf6890ecb6571e6361731d46f53cceb7cb584)
  
  More node deprecation warning fixes.

  Bartłomiej Dach · 2026-04-03 00:19:46 +09:00
  5b6d215583
• Tooling updates (#37031)

  Most of this is as everywhere else, but there's also interesting code
  inspection fixes from the InspectCode bump, so I'll talk about that a
  little.
  
  ## [Fix suspicious equality in
  `Hotkey`](https://github.com/ppy/osu/commit/948136e49e88a721827d54e51c5759fe9aca811d)
  
  Inspection:
  https://www.jetbrains.com/help/resharper/TypeWithSuspiciousEqualityIsUsedInRecord.Global.html
  
  Pretty annoying to fix, nullable array types are a pain. Does look legit
  though.
  
  ## [Fix `StarDifficulty` using inefficient struct
  equality](https://github.com/ppy/osu/commit/2db775ebb0bb9f18de67677ef84b993465d26545)
  
  Inspection:
  https://www.jetbrains.com/help/resharper/DefaultStructEqualityIsUsed.Global.html
  
  This is a dodgy one because there's no real sane way to define equality
  on `StarDifficulty` now that it has difficulty and performance
  attributes jammed into it. So I just basically shut the inspection up
  with a `record` modifier and move on.
  
  Unclear where the equality is used precisely. It's from a global
  inspection. F12 is very unhelpful when trying to track down usages of
  `Equals()`. We definitely have `Bindable<StarDifficulty>` instances and
  those do use equality. Maybe more than that.
  
  ## [Use `nameof` expressions to reference enum member
  names](https://github.com/ppy/osu/commit/aa08175c803bc725f3b15a92174dfe6d1b812d91)
  
  Inspection:
  https://www.jetbrains.com/help/resharper/CanSimplifyDictionaryRemovingWithSingleCall.html
  
  Pretty quaint.
  
  ## [Prefer using concrete values over `default` or
  `new()`](https://github.com/ppy/osu/commit/b21ee08d7748be10d42268d5c2eb77369026545d)
  
  Inspection:
  https://www.jetbrains.com/help/resharper/PreferConcreteValueOverDefault.html
  
  I could see this one going both ways, but I'm kinda sold on this
  inspection. Explicit is always better. Saves some allocations in the
  `CancellationToken` cases as well.
  
  ## [Explicitly call `.AsEnumerable()` in some realm
  usages](https://github.com/ppy/osu/commit/c8ce1ecd42b9d8abb8b9e2ab93d471f463e80401)
  
  Inspection:
  https://www.jetbrains.com/help/resharper/PossibleUnintendedQueryableAsEnumerable.html
  
  Not fully sold on this one but it's quick and simple so might as well.
  
  ## [Simplify dictionary removal with single `.Remove()`
  call](https://github.com/ppy/osu/commit/5964ceccea900302df726b7a8ecbf6b74eb2e427)
  
  Inspection:
  https://www.jetbrains.com/help/resharper/CanSimplifyDictionaryRemovingWithSingleCall.html
  
  Not much to say.

  Bartłomiej Dach · 2026-03-19 00:05:52 +09:00
  18d4ba5874
• Only run deploy actions for tags without suffixes

  Suffixed tags are build releases, which are triggered manually for now
  (via private project).
  
  Tags without suffix are the ones that we want to build the nuget
  packages and such.

  Dean Herbert · 2026-03-18 00:08:10 +09:00
  f1f753f7c2
• Attempt to fix mod output including build output
  Dean Herbert · 2026-03-16 14:16:20 +09:00
  73644d87bc
• Remove "copy labels from linked issues" github workflow (#36854)

  [It still doesn't
  work.](https://github.com/ppy/osu/actions/runs/22759488243/job/66012293202)
  
  Looking at the [job
  output](https://github.com/ppy/osu/actions/runs/22759488243/job/66012293202#step:1:21)
  it appears that the permissions of the `GITHUB_TOKEN` are
  *automatically* constrained to `read` even if you request more scopes.
  Would be nice if that was *actually documented* somewhere!
  
  Also given supply-chain attacks that people are running on github [via
  *issue titles* these
  days](https://grith.ai/blog/clinejection-when-your-ai-tool-installs-another)
  I'm not sure we want any automation near where it can reach code. Sure,
  much of the fault in the aforementioned attack was the fault of meatbags
  trusting clankers *WAY* too much, which is a mistake we *would not* do,
  but given everpresent software degradation *from unknown sources and for
  unknown reasons* let's not ~~COPILOT~~ *ahem* tempt fate...

  Bartłomiej Dach · 2026-03-06 19:54:44 +09:00
  f9e863af01
• Final attempt to make copy labels work
  Dean Herbert · 2026-03-03 17:24:29 +09:00
  317be21569
• Attempt to fix "copy labels from issues" workflow permissions again

  It's still broken:
  
  https://github.com/ppy/osu/actions/runs/22310482792/job/64540909415
  
  and I'm not sure what else it could be other than this, so I'm giving it
  one more honest try before I throw up hands.
  
  For better or worse,
  https://docs.github.com/en/actions/reference/workflows-and-actions/workflow-syntax#permissions
  explicitly lists that the `pull-requests` permission allows tagging PRs,
  so I'm holding out hope that this is it. Still terrible marks for
  documentation here, both to the github action, as well as everything to
  do with `GITHUB_TOKEN`.

  Bartłomiej Dach · 2026-02-24 13:34:06 +01:00
  ccb0224a1b
• Fix "copy labels from issues" workflow not having the appropriate permissions to do what it claims to do (#36602)
  Bartłomiej Dach · 2026-02-06 20:22:53 +09:00
  2efe0c95e6
• Add automation to copy labels from issue when PR is opened

  https://github.com/marketplace/actions/copy-issue-labels

  Dean Herbert · 2026-02-04 14:37:35 +09:00
  892e34b1f8
• No longer download iOS simulator

  https://github.com/actions/runner-images/issues/12862#issuecomment-3209787203

  Dan Balasescu · 2025-08-22 13:29:46 +09:00
  0756c45d70
• Force Xcode 16.4 on CI
  Dan Balasescu · 2025-08-16 15:52:27 +09:00
  1306f56fba
• Use macOS-15 for iOS builds
  Dan Balasescu · 2025-07-22 18:14:09 +09:00
  4daa900a19
• Remove iOS workload rollbacks
  Dan Balasescu · 2025-07-22 17:27:26 +09:00
  d3701f4659
• Enable nuget uploads and delete appveyor workflows
  Dan Balasescu · 2025-06-05 15:52:38 +09:00
  7145acc73f
• No symbols for templates package
  Dan Balasescu · 2025-06-05 15:46:05 +09:00
  4b92e338d5
• Add GHA deploy workflow
  Dan Balasescu · 2025-06-05 15:41:10 +09:00
  b6c4a713d3
• Increase diff calc processor timeout minutes
  tsunyoku · 2025-05-20 11:29:46 +01:00
  a1bdbd8ec3
• Fix osu.Game.Tests tests running twice
  Dan Balasescu · 2025-03-22 02:28:06 +09:00
  aacb5d86c8
• Build iOS tests project
  Dan Balasescu · 2025-03-18 16:19:52 +09:00
  2630d9437c
• Fix android build again

  Another month, another freak android build failure.
  
  From what I can tell, this time the build is broken because the second-
  -to-last workaround applied to fix it, namely explicitly specifying
  the version of workloads to install, stopped working, presumably because
  github pushed a new .NET SDK version to runners, and microsoft didn't
  actually put up a set of workload packages whose version matches the
  .NET SDK version 1:1.
  
  Thankfully, the fix to the *last* android build breakage (which caused
  the workload installation to completely and irrecoverably fail), appears
  to be rolling out this week, and *also* fix that same second-last issue,
  so both workarounds of specifying the workload version and pinning the
  image to `windows-2019` appear to no longer be required.
  
  Note that the newest image version, 20250209.1.0, is still not fully
  rolled out yet, thus rather than just fix all builds, this will fix like
  20% of builds until the newest image is fully rolled out. But I guess
  a 20% passing build is better than a 0% passing build, in a sense...?

  Bartłomiej Dach · 2025-02-12 14:18:03 +01:00
  ee6dcbd808
• Roll back windows build image to 2019 on android build job

  Per workaround suggested in
  https://github.com/actions/runner-images/issues/11402#issuecomment-2596473501.
  
  Applying this now as my hopes for a swift resolution without changes on
  our side are slim to none (read thread linked above in full to learn
  why).

  Bartłomiej Dach · 2025-01-23 10:12:30 +01:00
  b4e8a17f03
• Allow tsunyoku and stanriders to trigger diffcalc spreadsheet generator
  Dan Balasescu · 2024-12-11 09:59:18 +09:00
  315a9dba9b
• Fix android workflow not installing .NET 8 version
  Salman Alshamrani · 2024-11-27 23:46:21 -05:00
  19e396f878
• Make android build on non-maui workflow
  Dan Balasescu · 2024-11-18 17:20:11 +09:00
  02deb39335
• Adjust job name
  Dan Balasescu · 2024-11-11 13:44:11 +09:00
  91d9c0a7e8
• Enforce concurrency by using single job

  I've yet again re-confirmed by doubts about using concurrency groups.
  It's just not flexible enough. In this case, it cancels any _future_
  jobs.

  Dan Balasescu · 2024-11-11 13:40:17 +09:00
  0b570c4e15
• Fix empty JSON sent on non-workflow_dispatch events
  Dan Balasescu · 2024-11-11 13:11:21 +09:00
  394ff88a62
• Don't fail grep if no matches
  Dan Balasescu · 2024-11-11 13:02:26 +09:00
  c1686fb687
• Fix wrong cleaned up directory
  Dan Balasescu · 2024-11-11 12:49:23 +09:00
  a3b8c4d127
• Set -euo pipefail in diffcalc workflows
  Dan Balasescu · 2024-11-11 12:49:23 +09:00
  2767054286
• Rename workflow
  Dan Balasescu · 2024-11-09 23:49:30 +09:00
  5e8df623d4
• Split diffcalc workflow to add concurrency group
  Dan Balasescu · 2024-11-09 19:59:20 +09:00
  6183daa95f
• Merge pull request #30543 from smoogipoo/fix-ci-report

  Fix up + improve CI test reporting workflow

  Bartłomiej Dach · 2024-11-08 08:37:22 +01:00
  ce9c74af85
• Fix CI test report workflow
  Dan Balasescu · 2024-11-08 11:55:24 +09:00
  233560b867
• Skip diffcalc job if target is not valid

  Same as all other jobs... Just so that it doesn't appear as "completed"
  in the workflow list.

  Dan Balasescu · 2024-11-07 19:15:58 +09:00
  cb2e5ac2c4
• Update GHA
  Dan Balasescu · 2024-11-06 14:02:01 +09:00
  4b9f9b9be5
• Fix argument order in diffcalc workflow
  Dan Balasescu · 2024-10-17 16:57:00 +09:00
  1536a9886c
• Improve diffcalc workflow with explicit wait + logs
  Dan Balasescu · 2024-10-17 13:44:00 +09:00
  135b85a55a
• Overwrite downloaded data packages

  In https://github.com/ppy/osu/actions/runs/11311858931/job/31458581002, I cancelled the run during the download from `data.ppy.sh`.
  
  In https://github.com/ppy/osu/actions/runs/11313128285/job/31461534857, `wget` skipped downloading the file due to the `-nc` option (no-clobber), i.e.: if the file exists, don't re-download.
  
  The only way I'm aware of to resolve this with wget is to either use `-c` (continue), which may lead to broken files, or to explicitly specify the output file via `-O`. Thought I'd clean up a few pieces in the process.
  
  Why not curl? Mostly historical - some distros don't come with curl. It may be okay now but there's probably no point changing this at the moment...

  Dan Balasescu · 2024-10-13 18:43:02 +09:00
  9681e3ac46
• Save master state in workflow
  Dan Balasescu · 2024-10-07 18:52:13 +09:00
  639caf167d
• Make diffcalc workflow recreate comment on completion
  Dan Balasescu · 2024-10-07 15:50:59 +09:00
  c7f2564c0a
• Rollback iOS workload to last known working version
  Dan Balasescu · 2024-10-05 23:00:04 +09:00
  232381c9fb
• Fix iOS CI build
  Dan Balasescu · 2024-09-27 16:26:11 +09:00
  cb51e12d13
• Remove macOS/Xcode version pinning in iOS workflow
  Salman Ahmed · 2024-08-08 13:03:09 +03:00
  e12dba24ae
• Disable macOS test runs for now

  We are seeing update frames run as little as [once per
  second](https://github.com/ppy/osu/blob/aa4d16bdb873d7296b899cee7b7491ffdf5cd6ab/osu.Game/Overlays/BeatmapListingOverlay.cs#L141).
  Until we can ascertain why this is happening, let's reduce developer
  stress by not running macOS tests for now.

  Dean Herbert · 2024-07-25 09:05:58 +09:00
  4cc07badbd
• Raise workflow timeout time

  https://github.com/ppy/osu/actions/runs/9985890747/job/27597500883

  Dan Balasescu · 2024-07-18 16:26:40 +09:00
  7bb680a8a4
• Add frenzibyte to users that can use diffcalc workflow
  Dan Balasescu · 2024-07-08 13:58:42 +09:00
  9375f79879
• Update action versions in diffcalc workflow
  Dan Balasescu · 2024-03-29 00:25:32 +09:00
  cbbb46cad8
• Update diffcalc workflow
  Dan Balasescu · 2024-03-27 13:37:47 +09:00
  4490dbf896
• Update github actions to resolve most node deprecation warnings

  As is github tradition, workflows started yelling about running on a
  node version that was getting sunset, so here we go again.
  
  Relevant bumps:
  
  - https://github.com/actions/checkout/releases/tag/v4.0.0
  - https://github.com/actions/setup-dotnet/releases/tag/v4.0.0
  - https://github.com/actions/cache/releases/tag/v4.0.0
  - https://github.com/actions/setup-java/releases/tag/v4.0.0
  - https://github.com/peter-evans/create-pull-request/releases/tag/v6.0.0
  - https://github.com/dorny/test-reporter/releases/tag/v1.8.0
  
  Notably, `actions/upload-artifact` is _not_ bumped to v4, although it
  should be to resolve the node deprecation warnings, because it has more
  breaking changes and bumping would break `dorny/test-reporter`
  (see https://github.com/dorny/test-reporter/issues/363).

  Bartłomiej Dach · 2024-02-22 17:18:10 +01:00
  99bbbf810b