[psrvere]

Changes

Closes #415.

shp build upload created its own BuildRun and exposed only a fixed set of BuildRun fields as flags. A Local-source build could not ask for per-step compute resources, and a local folder could not be streamed into a BuildRun created some other way. This came up while migrating OpenShift BuildConfigs that set spec.resources to Shipwright with the crane plugin.

This PR adds both flags proposed in the issue.

• --step-resources is repeatable and sets BuildRunSpec.StepResources, in the form <step>=<limits|requests>.<resource>=<quantity>, for example build-and-push=limits.memory=2Gi. Entries for the same step merge. It goes through BuildRunSpecFromFlags, so shp build run and shp buildrun create get it too.
• --buildrun-name streams the local source into an existing BuildRun instead of creating one. That covers any BuildRun field the CLI has no flag for.

--buildrun-name fails early, with an error, in these cases.

• It is combined with a flag that only shapes a new BuildRun, such as --sa-name. That flag would have no effect. Global flags such as --namespace are still accepted.
• The Build uses a source bundle image. Its BuildRun pulls the bundle from a registry and never waits for the CLI.
• The BuildRun references another Build. The upload finds the build pod by the Build name, so it would never match and would wait forever.
• The BuildRun has no Local source. Its pod has no container waiting for the upload.

The pod of an existing BuildRun is often already running when the upload starts. The upload reacts to the first event it sees for the pod, so it streams at once instead of waiting until the BuildRun times out.

The PR has two commits. 07e9276 adds the two flags. aa15a22 fixes four ways --buildrun-name failed when tested on a cluster: it rejected --namespace, it sat idle when the BuildRun's pod was already running, it waited forever on a BuildRun of another Build, and it gave an unclear error on a BuildRun without a Local source.

Checks run locally: go build ./..., make test-unit and make verify-docs pass. The new unit tests fail on 07e9276 and pass on aa15a22.

Each case above was also run on two clusters, a kind cluster running Shipwright Build and an OpenShift 4.20 cluster with Builds for Red Hat OpenShift 1.9.0. On both, 07e9276 showed all four failures. With aa15a22 the upload streams through --namespace and into a BuildRun whose pod is already running, and both builds succeed. The other cases stop at once with an error that names the problem. A plain upload without --buildrun-name still builds.

Submitter Checklist

• Includes tests if functionality changed/was added
• Includes docs if changes are user-facing
• Set a kind label on this PR
• Release notes block has been filled in, or marked NONE

See the contributor guide for details on coding conventions, github and prow interactions, and the code review process.

Release Notes

`shp build upload` gains `--step-resources` to override per-step compute resources and `--buildrun-name` to stream local source into an existing BuildRun.

Co-Authored-By: Claude

[openshift-ci]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please assign saschaschwarze0 for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[psrvere]

/kind feature

[psrvere]

Thanks for the review, @sayan-biswas. Testing your two questions turned up four bugs in --buildrun-name, all fixed in aa15a22.

1. It rejected --namespace and the other global flags.
2. When the BuildRun's pod was already running, the upload sat idle until the pod stopped waiting, about 60 seconds, and then failed with "container not found". It now reacts to the pod's first event as well.
3. A BuildRun of a different Build made the command wait forever. It now stops with an error.
4. A BuildRun without a Local source gave an unclear error. It now says the BuildRun needs a Local source.

I ran every case on a kind cluster and on OpenShift 4.20 with Builds for Red Hat OpenShift 1.9.0. Each case fails on 07e9276 and works on aa15a22. The new unit tests fail on the old code and pass on the new.

[sayan-biswas]

/lgtm