Use osx.13.arm64.open queue for osx testing

[MichalStrehovsky]

The osx 12 queue only has a couple machines servicing it. Looks like osx 12 is already out of support by Apple so dnceng scaled it down?

This queue is the main reason why runtime-nativeaot-outerloop almost never succeeds despite my weeks long effort to get it green. E.g (https://helix.dot.net/api/jobs/226c27df-153e-4f75-9003-5dd5f5a355a2/workitems/System.Diagnostics.Tools.Tests?api-version=2019-06-17):

  "Queued": "2025-02-17T23:02:28.573+00:00",
  "Started": "2025-02-18T05:34:09.794+00:00",
  "Finished": "2025-02-18T05:34:12.732+00:00",
  "Delay": "06:31:41.2210000",
  "Duration": "00:00:02.9380000",

Six and a half hour wait time so we can run a three second job.

Per dotnet/dnceng#3008 (comment) the recommendation is to move to osx 14 queues but I'm trying with osx 13 since I assume our preference is as old as possible.

Cc @dotnet/runtime-infrastructure

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

[MichalStrehovsky]

/azp run runtime-nativeaot-outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[ivanpovazan]

Should we bump the x64 variant as well?

[MichalStrehovsky]

Should we bump the x64 variant as well?

I've not seen problems with that queue so my gut says "don't touch what's not broken". I don't know what schedule we follow in upgrading these. I just know that we had problems with this queue since december when the number of machines dropped from 30 to 10 (dotnet/dnceng#4734) and I don't see anyone doing anything about this so I'm just trying to do something.

[ivanpovazan]

Ok, thanks.
We are experiencing timeouts on osx.1200.amd64 as well, for iOS platforms testing, so was wondering wether we should bump everything to osx.13 instead of gradually moving to the newer queue.

[am11]

x64 variant

Perhaps we should promote osx-arm64 to how much we are currently caring for (the declining) osx-x64 #108570 (Apple stopped producing x64 macs in 2023: second paragraph).

[akoeplinger]

bump this one too?

Suggested change

	- OSX.1200.ARM64
	- osx.13.arm64

[MichalStrehovsky]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[MichalStrehovsky]

System.Security.Cryptography.OpenSsl.Tests is running into SIGABRT on native AOT; that doesn't look very promising. I fired off libraries outerloop since by default we don't seem to test anything on arm64 macs and no other leg ran there.

[filipnavara]

System.Security.Cryptography.OpenSsl.Tests is running into SIGABRT on native AOT; that doesn't look very promising.

That's not entirely unexpected. macOS doesn't ship OpenSSL by default and the test prints No usable version of libssl was found which is followed by abort.

[am11]

It should be running this script:

runtime/eng/pipelines/common/global-build-job.yml

Lines 197 to 198 in 6e06aef

	- script: $(Build.SourcesDirectory)/eng/common/native/install-dependencies.sh ${{ parameters.osGroup }}
	displayName: Install Build Dependencies

[akoeplinger]

@am11 that is for the build machine, it doesn't apply to Helix

IIRC openssl was manually installed on these machines by dnceng?

[am11]

Yup, I meant test machine can use the same script to install runtime dependencies. We are using it in multiple places in addition to build machines.

[steveisok]

@am11 that is for the build machine, it doesn't apply to Helix

IIRC openssl was manually installed on these machines by dnceng?

I pinged @ilyas1974 and he'll have it looked into.

[steveisok]

Any reason why we can't bump to osx.15.arm64 and osx.15.arm64.open? We appear to have even more capacity there.

[MichalStrehovsky]

Any reason why we can't bump to osx.15.arm64 and osx.15.arm64.open? We appear to have even more capacity there.

I assumed we have a preference for oldest possible macOS and then cross fingers about Apple's backwards compatibility. But I don't know our testing strategy for mac.