outerloop-nightly is broken because "native test components" is unavailable

[kunalspathak]

It seems that we skip to upload the "native test components" because of which all the test jobs in outerloop-nightly fails because it tries to download those artifacts but doesn't find it. The conditions to skip uploading seems legitimate, so not sure if we should skip downloading them as well with the same conditions check.

runtime/eng/pipelines/coreclr/templates/build-job.yml

Lines 294 to 305 in c3a2bce

	- ${{ if and(ne(parameters.compilerName, 'gcc'), ne(parameters.testGroup, ''), ne(parameters.testGroup, 'clrTools'), ne(parameters.disableClrTest, true)) }}:
	# Publish test native components for consumption by test execution.
	- ${{ if and(ne(parameters.isOfficialBuild, true), eq(parameters.pgoType, '')) }}:
	- template: /eng/pipelines/common/upload-artifact-step.yml
	parameters:
	rootFolder: $(nativeTestArtifactRootFolderPath)
	includeRootFolder: false
	archiveType: $(archiveType)
	tarCompression: $(tarCompression)
	archiveExtension: $(archiveExtension)
	artifactName: $(nativeTestArtifactName)
	displayName: 'native test components'

runtime pipeline

vs.

outerloop-nightly pipeline

and the download job for native test components keep failing.

Tagging subscribers to this area: @dotnet/runtime-infrastructure
See info in area-owners.md if you want to be subscribed.

Issue Details

---

It seems that we skip to upload the "native test components" because of which all the test jobs in outerloop-nightly fails because it tries to download those artifacts but doesn't find it. The conditions to skip uploading seems legitimate, so not sure if we should skip downloading them as well with the same conditions check.

runtime/eng/pipelines/coreclr/templates/build-job.yml

Lines 294 to 305 in c3a2bce

	- ${{ if and(ne(parameters.compilerName, 'gcc'), ne(parameters.testGroup, ''), ne(parameters.testGroup, 'clrTools'), ne(parameters.disableClrTest, true)) }}:
	# Publish test native components for consumption by test execution.
	- ${{ if and(ne(parameters.isOfficialBuild, true), eq(parameters.pgoType, '')) }}:
	- template: /eng/pipelines/common/upload-artifact-step.yml
	parameters:
	rootFolder: $(nativeTestArtifactRootFolderPath)
	includeRootFolder: false
	archiveType: $(archiveType)
	tarCompression: $(tarCompression)
	archiveExtension: $(archiveExtension)
	artifactName: $(nativeTestArtifactName)
	displayName: 'native test components'

runtime pipeline

vs.

outerloop-nightly pipeline

and the download job for native test components keep failing.

Author:	kunalspathak
Assignees:	-
Labels:	area-Infrastructure
Milestone:	-

[kunalspathak]

@agocke @hoyosjs @jkoritzinsky

[trylek]

It seems to me that the problem is that we're not setting the parameter testGroup when calling build-coreclr-and-libraries-job so that the condition ne(parameters.testGroup, '') doesn't hold. Other than that, I'm somewhat surprised why we have two different flavors of coreclr outerloop runs,

https://dev.azure.com/dnceng-public/public/_build?definitionId=135&_a=summary

vs.

https://dev.azure.com/dnceng-public/public/_build?definitionId=108&_a=summary

In particular, I believe that in our Monday backlog sync-ups JeffSchw looks at the latter so if the intent was to improve outerloop coverage by implementing some checked vs. release build variations, it would seem logical to me to add them to the existing outerloop run instead of creating a new one.

By the way, even though it doesn't seem to be the root problem here, the pipeline template build-coreclr-and-libraries-job doesn't pass the isOfficialBuild parameter to build-job so the value passed to the template at line 31 is silently dropped on the floor.

[kunalspathak]

In particular, I believe that in our Monday backlog sync-ups JeffSchw looks at the latter so if the intent was to improve outerloop coverage by implementing some checked vs. release build variations, it would seem logical to me to add them to the existing outerloop run instead of creating a new one.

Agree.

[trylek]

I have looked at the GitHub history and it looks like the coreclr-release-outerloop-nightly pipeline was either created or ported as part of our repo merging in early 2020. I must admit I have no detailed memory of the pipeline despite the fact that I apparently contributed to it by one or two fixes. While I believe it should be possible to fix the pipeline by adding the testGroup jobParameter when calling build-coreclr-and-libraries-job, longer term I would find it preferable to fold this pipeline into the runtime-coreclr outerloop pipeline provided the various infra stakeholders agree (unless there's some crucial counterargument I'm missing right now).

[trylek]

I have triggered an experimental run with my proposed fix for the pipeline:

https://dev.azure.com/dnceng-public/public/_build/results?buildId=250522&view=logs&j=af986455-df3a-5ed6-b067-9144b6c8d0c6

It's only just started but I do see the "Zip / publish native test components" steps in the CoreCLR build jobs. I can send it out for PR if the test finishes fine; we can then have a secondary discussion regarding its potential consolidation with the other outerloop pipeline.

[kunalspathak]

I can send it out for PR if the test finishes fine

Sounds good to me.

[trylek]

OK, so it's far from green (I haven't found any previous green run of the pipeline) but at least it's now actually sending the tests to Helix and just a handful of them is failing, a few test execution jobs have passed. I'm going to go ahead and publish the PR now as a general goodness, I think that some of the test failures are really interesting in terms of test coverage, we may have been missing some of them previously due to using checked builds in the runtime-coreclr outerloop pipeline.