Errors with CUE for Tekton

[cueckoo]

Originally opened by @ekarlso in cuelang/cue#435

Hi guys, I have the following code that's setup. But when I do a cue dump it just hangs and seems to be memory leaking + hogging all cpu. Should I file a bug possibly?

export GO111MODULE=on
go get github.com/tektoncd/pipeline/pkg/apis/...@v0.13.2
cue get go github.com/tektoncd/pipeline/pkg/apis/...

package kube

objects: [ for v in objectSets for x in v {x}]

objectSets: [
  tknTask
]

dump_toom.cue

package kube

import (
	"tool/cli"
	"encoding/yaml"
)

command: dump: {
	task: print: cli.Print & {
		text: yaml.MarshalStream(objects)
	}
}

tkn.cue

package cicd

import (
  tkn_v1_beta "github.com/tektoncd/pipeline/pkg/apis/pipeline/v1beta1"
)

tknTask: [ID=_]: tkn_v1_beta.#Task & {
  apiVersion: "tekton.dev/v1beta1"
  kind:       "Task"
  metadata: name: ID
}

tknTask: "terraform-cli":  {
  spec: {
    workspaces: [{
      name: "source"
    }]
    params: [{
      name:        "ARGS"
      description: "The terraform cli commands to tun"
      type:        "string"
    }, {
      name:        "subdirectory"
      description: "What subdirectory to execute in"
      type:        "string"
      default:     "infra-terraform"
    }, {
      name:    "tf-workspace"
      type:    "string"
      default: "default"
    }]

    steps: [{
      name:       "init-and-run"
      image:      "x"
      workingDir: "$(workspaces.source.path)"
      script: """
      #!/usr/bin/env sh

      export VAULT_ADDR=\"https://x\"
      metadata=$(curl -H Metadata:true \"http://169.254.169.254/metadata/instance?api-version=2019-08-15\")
      subscription_id=$(echo $metadata | jq -r .compute.subscriptionId)
      vm_name=$(echo $metadata | jq -r .compute.name)
      vmss_name=$(echo $metadata | jq -r .compute.vmScaleSetName)
      resource_group_name=$(echo $metadata | jq -r .compute.resourceGroupName)
      cat <<EOF > auth_payload_complete.json
      {
          \"role\": \"aks-role\",
          \"jwt\": \"$jwt\",
          \"subscription_id\": \"$subscription_id\",
          \"resource_group_name\": \"$resource_group_name\",
          \"vm_name\": \"$vm_name\",
          \"vmss_name\": \"$vmss_name\"
      }
      EOF

      export ARM_USE_MSI=true
      echo $(workspaces.source.path)/$(params.subdirectory)
      cd $(workspaces.source.path)/$(params.subdirectory)

      terraform init
      terraform workspace select $(params.tf-workspace)
      terraform $(params.ARGS)
      """
    }]
  }
}

package kube

import (
	"tool/cli"
	"encoding/yaml"
)

command: dump: {
	task: print: cli.Print & {
		text: yaml.MarshalStream(objects)
	}
}

[cueckoo]

Original reply by @seh in cuelang/cue#435 (comment)

This sounds reminiscent of #391, though I can't see where you're using the CUE source imported via the cue get go command at the top.

[cueckoo]

Original reply by @ekarlso in cuelang/cue#435 (comment)

I do have the import at the top yes.

[cueckoo]

Original reply by @mpvl in cuelang/cue#435 (comment)

Yeah, it is a known problem. The new evaluator, which is getting really close now, addresses a myriad of performance issues. So I expect that this will be fixed as of v0.3.0.

It doesn't make much sense to try to fix this in the current implementation at this point.

[cueckoo]

Original reply by @ekarlso in cuelang/cue#435 (comment)

@mpvl any idea on a "beta" or something that can work ?

[DavidGamba]

For the record, this example worked for me with cue version v0.6.0-alpha.1

Steps:

cue mod init

go mod init tekton-cue-demo

go get github.com/tektoncd/pipeline/pkg/apis/...@latest

cue get go github.com/tektoncd/pipeline/pkg/apis/...

dump_tool.cue

package kube

import (
	"tool/cli"
	"encoding/yaml"
)

command: dump: {
	task: print: cli.Print & {
		text: yaml.MarshalStream(objects)
	}
}

objects.cue

package kube

objects: [ for v in objectSets for x in v {x}]

objectSets: [
  tknTask
]

task.cue

package kube

import (
  tkn_v1_beta "github.com/tektoncd/pipeline/pkg/apis/pipeline/v1beta1"
)

tknTask: [ID=_]: tkn_v1_beta.#Task & {
  apiVersion: "tekton.dev/v1beta1"
  kind:       "Task"
  metadata: name: ID
}

tknTask: "terraform-cli":  {
  spec: {
    workspaces: [{
      name: "source"
    }]
    params: [{
      name:        "ARGS"
      description: "The terraform cli commands to tun"
      type:        "string"
    }, {
      name:        "subdirectory"
      description: "What subdirectory to execute in"
      type:        "string"
      default:     "infra-terraform"
    }, {
      name:    "tf-workspace"
      type:    "string"
      default: "default"
    }]

    steps: [{
      name:       "init-and-run"
      image:      "x"
      workingDir: "$(workspaces.source.path)"
      script: """
      #!/usr/bin/env sh

      export VAULT_ADDR=\"https://x\"
      metadata=$(curl -H Metadata:true \"http://169.254.169.254/metadata/instance?api-version=2019-08-15\")
      subscription_id=$(echo $metadata | jq -r .compute.subscriptionId)
      vm_name=$(echo $metadata | jq -r .compute.name)
      vmss_name=$(echo $metadata | jq -r .compute.vmScaleSetName)
      resource_group_name=$(echo $metadata | jq -r .compute.resourceGroupName)
      cat <<EOF > auth_payload_complete.json
      {
          \"role\": \"aks-role\",
          \"jwt\": \"$jwt\",
          \"subscription_id\": \"$subscription_id\",
          \"resource_group_name\": \"$resource_group_name\",
          \"vm_name\": \"$vm_name\",
          \"vmss_name\": \"$vmss_name\"
      }
      EOF

      export ARM_USE_MSI=true
      echo $(workspaces.source.path)/$(params.subdirectory)
      cd $(workspaces.source.path)/$(params.subdirectory)

      terraform init
      terraform workspace select $(params.tf-workspace)
      terraform $(params.ARGS)
      """
    }]
  }
}

$ tree -L 1
.
├── README.adoc
├── cue.mod
├── dump_tool.cue
├── go.mod
├── go.sum
├── objects.cue
└── task.cue

Then run cue dump