Runtime: `/azp run` can't start a pipeline if it is marked as `pr: none`

[sandreenko]

Azure Pipelines could not run because the pipeline triggers exclude this branch/path.

_Originally posted by @azure-pipelines in https://github.com/dotnet/runtime/pull/32362#issuecomment-590618865_

with /azp list you can get the list of the available pipeline, but if you try to trigger them you will see:
Azure Pipelines could not run because the pipeline triggers exclude this branch/path.
because they have that setting:
https://github.com/dotnet/runtime/blob/321632436c98365191d1b6175503553e82535f20/eng/pipelines/coreclr/ci.yml#L26

So if we want to be able to trigger something with /azp run pipeline name we need to do the workaround from https://github.com/MicrosoftDocs/vsts-docs/issues/4206

[karelz]

@sandreenko only 1 area label please - see why

[ViktorHofer]

@trylek @safern do we know if this will change on the AzDO side? If not, I don't think this item is actionable and should be closed.

[trylek]

Well, so far I'm not aware of any indication that AzDO would be planning to roll out a fix / new functionality addressing this issue in the short term especially as this is kind of a feature request rather than a mere bugfix.

[ViktorHofer]

Can you please remind me why some pipelines like the one mentioned in the top aren't triggerable? How is this pipeline different from the libraries outerloop which is triggerable?

[sandreenko]

Can you please remind me why some pipelines like the one mentioned in the top aren't triggerable? How is this pipeline different from the libraries outerloop which is triggerable?

I think the task is:

1. trigger only a small amount of pipelines are when you run /azp run;
2. make all optional heavy pipelines triggerable with /azp run NAME;

right now if we delete pr: none from all pipelines when you do /azp run it will run them all, that is undesirable for jit GCStress jobs etc. that are taking hours and many machines. But with pr: none you can't get the second.

[ViktorHofer]

Isn't /azp run already a no-op in dotnet/runtime as we set the pipeline limit to 10 and we have more than that?

[safern]

I think if we just remove the PR triggers and then set the options on the UI that the pipeline requires a comment to run, it would solve this issue.

Isn't /azp run already a no-op in dotnet/runtime as we set the pipeline limit to 10 and we have more than that?

Yeah, this would happen and no pipelines would run, so I'm OK with removing the PR trigger section from the yml file.

[ViktorHofer]

@trylek can you help with that as you know which coreclr pipelines make sense to be triggered by an /azp command? I believe we just need to remove the pr: none entries and then mimic the pipelines UI settings with what we already have for libraries outerloop:

[trylek]

If I'm right to understand that we still operate under the limitation that "/azp run" runs all the eligible pipelines, I guess we should only support a relatively small number of pipelines to avoid clogging the system by use of the parameterless command. I guess this rules out all the stress tests and mostly leaves in pipelines representing build variations - R2R, Crossgen2, maybe jit-experimental. Adding @BruceForstall who has the best knowledge of our pipeline portfolio and its usefulness to chime in.

[ViktorHofer]

/azp run doesn't do anything as we exceed the specified pipeline limit (10): https://github.com/dotnet/runtime/pull/41620#issuecomment-701691751

[trylek]

OK, in such case I guess I'll just modify the UI settings for all pipelines as you described above and then we can arbitrarily start removing the pr: none specifiers.

[ViktorHofer]

Thanks Tomas

[sandreenko]

As I saw we now can trigger many pipelines with /azp run, it was fixed by #42933, @trylek should this issue be closed?

[ViktorHofer]

Yes, we can close the issue now. Thanks