[ric-oliv]

This follows up on #5746. On #5752, the first version7 PR after it, the build workflow took 63.3 minutes, and the macOS .NET job took 59.6 of those. This PR makes two more cuts, and no test is dropped. The device tests now also run for a few changes they used to skip.

I first opened it against version7 and measured it there, on top of #5746. Neither change depends on #5746, so it now targets main.

• Sentry-CI-Build-macOS.slnf no longer includes Sentry.Maui.Device.TestApp. The macOS build job compiled it for net9.0-android, net10.0-android and net10.0-ios (net10.0-android, net11.0-android and net11.0-ios on version7), and nothing in build.yml runs it. In chore(samples): move the Azure Functions sample to Azure.Functions.Sdk and net11.0 #5752's run on version7, packaging that app was all the Build step did after minute 18.4 of 24.7, and in a recent main run it took the last 8 minutes of 35. The Linux and Windows filters already exclude it, and both device test workflows build it themselves. On main, the macOS leg of verify-api builds the same filter, so it stops building the app too.
  • Nothing in the build workflow ever used what it built. dotnet test skips it because platform TFMs aren't test projects here, dotnet pack skips it because it isn't packable, and the sparse checkout deletes its bin and obj before the integration tests. So all it gave us was a compile check, and both device test workflows already run one with scripts/device-test.ps1 -Build. Every project it depends on still builds for its mobile TFMs, and the job packs the same packages as before.
  • To keep that safe, the device test path filters now also list what the app compiles and they didn't: Sentry.Analyzers, Sentry.Testing.CrashableApp, the nested Directory.Build.* files and .editorconfig, which can fail the build because warnings are errors.
  • On version7, both filters also need src/Sentry.Extensions.Hosting/**, since feat: add Sentry.Extensions.Hosting with UseSentry for generic host apps #5739 made Sentry.Maui.Tests reference it. That project isn't on main, so I left it out here. It needs adding when main next merges into version7.
• verify-api and format-code shared the pr-auto-commit concurrency group at workflow level, so each waited for the other to finish. On one PR, format-code started 35 minutes after its run was created, and on another verify-api waited 13 minutes. The group exists to keep their git pushes apart, so it now covers only accept-api-changes and the format-code job.

Here's the CI run on main, compared with the medians of main PR runs since October 15, which are 26 build runs and 14 verify-api runs:

	main PRs, median	This PR on main
Build workflow	78.9 min	66.7 min
macOS .NET job	67.0 min	53.2 min
macOS Build step	32.7 min	22.9 min
.NET (win-x64) job	51.9 min	55.8 min
verify-api macOS job	35.7 min	29.6 min

• That's about 12 minutes off the build workflow. It would have been about 8 more, but Unsupported AOT (win-x86), which waits on every leg, spent 8 minutes in setup on this run.
• The macOS job ran the same 59 test runs with the same 11,451 passing tests as a recent main PR and packed the same 53 packages. format-code was done after 12.5 minutes instead of waiting for the API tests, and the device tests passed.

Here's what the build workflow took on three CI runs of this PR while it targeted version7, compared with #5752:

	#5752	Run 1	Run 2	Run 3
Build workflow	63.3 min	62.9 min	56.4 min	56.7 min
macOS .NET job	59.6 min	57.8 min	48.1 min	49.0 min
macOS Build step	24.7 min	19.8 min	15.6 min	15.7 min
.NET (win-x64) job	48.8 min	53.4 min	53.3 min	53.9 min

So on version7 this PR saves about 7 minutes on every build workflow run. Runs 2 and 3 took 56.4 and 56.7 minutes, 6.6 to 6.9 minutes under #5752. Run 1 barely moved because that runner spent 4.5 minutes downloading one workload pack during setup and ran the tests 2.4 minutes slower.

• The macOS Build step no longer builds the test app, and the macOS job is now about 11 minutes shorter.
• verify-api and format-code ran side by side, and both were done within 12.6 to 15.6 minutes of the push in all three runs. The API commit job still waits for the format job, as intended.
• The test counts match chore(samples): move the Azure Functions sample to Azure.Functions.Sdk and net11.0 #5752. The macOS job ran the same 79 test runs with the same counts, except for 4 AssemblyReader tests that ref: remove Android assembly store v4 support (never shipped) #5750 removed on version7. All 2997 iOS device tests and 8 iOS integration cases passed as well.

With macOS down to about 49 minutes on version7 and 53 on main, .NET (win-x64) is now the slowest leg on both, at 53 to 56 minutes. So the next round of CI work should focus on the Windows jobs. Their Test step alone takes 16 to 21 minutes, and in #5752's run only about 8 of those were spent inside the tests themselves.

The first two runs also included an iOS change that I've since taken out: booting the simulator while the device test app builds. It cut the app install from 6 min 46 s to under 20 seconds, but the app build got slower by about the same amount, so the iOS job didn't get faster. The build workflow doesn't use either of the files it touched, and run 3 ran without it.

Craft releases aren't affected. The device tests still run on every push to main and release/*, the macOS job still uploads the commit-SHA artifact with the same packages, and no check names change.

#skip-changelog

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.13%. Comparing base (04cac0d) to head (aa12ec7).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5758   +/-   ##
=======================================
  Coverage   76.13%   76.13%           
=======================================
  Files         528      528           
  Lines       19579    19579           
  Branches     3829     3829           
=======================================
  Hits        14906    14906           
  Misses       3791     3791           
  Partials      882      882           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jamescrosswell]

Nice - thanks @ric-oliv !

Any specific reason to target version7 with this change? If not, maybe we could change the target to main before merging?

[ric-oliv]

Good point, no reason. Changed to main!

[jamescrosswell]

Nice - thanks @ric-oliv !