twisterrob / github-workflows Goto Github PK

Account for error:

Publishing build scan...

Publishing failed.

The response from https://scans-in.gradle.com/scans/publish/gradle/3.15/token was not from the build scan server.
Your network environment may be interfering, or the service may be unavailable.

in which case fromJSON(steps.gradle.outputs.build-scan-url) will be fromJSON('') which fails as

The template is not valid. TWiStErRob/net.twisterrob.sun/.github/workflows/CI-screenshot-record.yml@1c747a488ff4a223fb5cee1df0a3f718a0fc0aa4 (Line: 63, Col: 19): Error reading JToken from JsonReader. Path '', line 0, position 0.

This issue lists Renovate updates and detected dependencies. Read the Dependency Dashboard docs to learn more.

This repository currently has no open or pending branches.

Detected dependencies

• Check this box to trigger a request for Renovate to run again on this repository

Instrumentation crashed on Inventory reused workflow, but the Gradle scan step was skipped.

TL;DR: publish reports as gh-pages.

Location:

• Separate shadow branch in the same repo
• Separate gh-pages branch in same repo
• Dedicated repo for specific repo, like org/repo-reports
• Dedicated repo for whole org, like org/gha-reports

Hosting:

• gh-pages or main branch, doesn't matter match, as long as Pages is set up.
• custom domain possible in Pages
• in case of shadow-branch only specific files are visible, I don't think HTML would be viable on raw.github, but have to check

Structure:

• depending on hosting/scope, mostly for conflict avoidance.
• possible tokens in path:
  • org
  • repo
  • branch (probably not necessary)
  • sha
  • workflow ID
  • attempt #
  • module
  • report-name (lint, detekt, unit-test, android-test, etc.)
• as long as enough discriminators exist, order doesn't matter
• workflow input with template?

For each report name we might need separate processing rules, or a separate job.

Steps:

• download artifact
• clone target repo
• move artifacts to right folder
• post-process if necessary (sed paths)
• report resulting (deployed) path in workflow call output.

Making it visible:

• deployment location for each report
• update comment in PR
• status checks with Details link
• create annotations on a root file (of modules), probably noisy
  • can be reduced if there's an uber-report collating all available reports for a workflow run
  • can be reduced with report browser based on paths

Implementation details to account for parallel CI jobs:

• either restrict parallelism, which is probably not possible per workflow call;
• or create PRs, because their merges can be serialized without conflicts.
  This has the potential to validate the branches, but that might be an overkill. Anyway possible if necessary.
  • steps
    • push branch (with discriminator name or hash of if length restricted)
    • auto-open PR in target repo
      • enable auto-merge
      • there should be no conflicts because of distinct paths
  • can we delete PRs?

Clean-up:

• wipe repo on the weekend/night?
• retention could be configured per path in pages repo
• steps either way
  • Clone pages repo
  • figure out which folders to delete
    • rm -rf
    • git filter-repo to remove specific from history.
  • force push

https://www.thisdot.co/blog/how-to-retry-failed-steps-in-github-action-workflows/

TWiStErRob/net.twisterrob.libraries#12

They should be delayed for evaluation until used. (i.e. after checkout)

      cache-mixin:
        required: false
        default: ''
        description: |
          Additional string to add to the cache key.
          Useful to invalidate the cache when the setup-script changes.
          Note: setup-script contents will be automatically added to the cache key,
          but if it calls an external script, that needs to be included here.
          Usage: `cache-mixin: hashFiles('...')`

https://stackoverflow.com/a/74586595/253468

actions/upload-artifact#50 (comment), but no direct URL access yet.

Nor it is possible to link to file within a ZIP file.

if-no-files-found: error