buildInfo isn't properly collected for jobs run within a docker container

[five-iron]

Describe the bug

I'm trying to run an npm build, publish build info, and execute xray scan while running inside a withDockerContainer block. However, after I try publishing the build info, a command to kick off an xray scan fails.

Current behavior

The jf 'npm i' works, then jf 'rt bp' runs without error, then jf 'bs' blows up saying it can't find the build info. When I check our artifactory, there was a build pushed but the build info doesn't have any of the metadata/dependencies from the npm install. I have a feeling from inspecting the job logs that the jfrog plugin/CLI might use environment variables tmp dirs to store build info, but they don't carry across the different commands

Reproduction steps

Run a pipeline job with steps similar to this scripted pipeline syntax:

node {
    withDockerContainer(
      image: 'node:18.17.1',
      args: "-v <tools_dir>:<tools_dir>:rw,z" // this was necessary so that the container could use the globally configured jfrog-cli tool
    ) {
        withEnv(["JFROG_BINARY_PATH=${tool 'jfrog-cli'}"]) {
            stage('Build and scan') {
                // checkout some npm project here
                jf 'npm-config --repo-resolve <resolve_repo> --repo-deploy <resolve_repo>'
                jf 'npm i'
                jf 'rt bp'
                jf 'bs' // 404 returned here
            }
        }
    }
}

Expected behavior

The build info should be populated and pushed correctly, allowing an xray scan to run against that build's dependencies.

JFrog plugin version

1.4.0

JFrog CLI version

2.45.0

Operating system type and version

Amazon Linux / UBI

JFrog Artifactory version

7.49.6

JFrog Xray version

3.66.6

[nickajacks1]

Hi @five-iron , I was encountering something slightly different but related.

tl;dr try setting this environment variable:

JFROG_CLI_TEMP_DIR="${WORKSPACE_TMP}/jfrog"

The default location where build info is kept track of is in /tmp, so you lose any build-info status when a build container is destroyed. You can override the default location with the JFROG_CLI_TEMP_DIR variable. The situation I'm describing doesn't quite match the single-node example you provide, but hopefully this information will give you the ability to debug a bit further.