chore: tidy the legacy content out of .github - #36
Conversation
![radiusred-checky[bot]](https://avatars.githubusercontent.com/in/4719924?s=60&v=4){kind=small}
There was a problem hiding this comment.
M1-R2 requires the rewritten CONTRIBUTING.md to reflect the current org, but two absolute statements in it are not accurate against the resulting tree/org:
- Under “What this repository holds”, “Nothing else lives here” is contradicted by the retained
.codecrew.yml,.gitignore, andCONTRIBUTING.mditself (the PR body andgit ls-filesboth show them). These are legitimate repository-local/coordination files, but the text needs to acknowledge that exception rather than say the listed shared assets are literally all that lives here. - Under “Contributing to a project”, “Every public repository carries its own README.md” is false for the public
.githubrepository: GitHub’s repository README endpoint returns none;profile/README.mdis the special organization profile, not this repository’s root README. Scope this claim to the project repositories (or otherwise phrase it so.githubis not included).
Please correct these inventory/contribution claims so the guide is honest about the org as it is. The functional cleanup itself checks out: the two caller repositories still reference the untouched reusable workflows at @main, the surviving tracked-file set matches the task’s explicit keep list, and the draw.io payload in ha-sinkhole#117 is byte-for-byte identical to the source removed here.
Review on PR #36 found two absolute statements that were untrue against the resulting tree: "Nothing else lives here" ignored .codecrew.yml, .gitignore and CONTRIBUTING.md itself, and "Every public repository carries its own README.md" was false for .github, whose profile/README.md is the organisation profile rather than a repository README. Name the housekeeping files and scope the README claim to project repositories. Refs M1-R2.
|
Addressed both findings from review 5096582894 in
No other files changed. The PR description now lists the fix commit and notes that ha-sinkhole#117 has merged ( |
There was a problem hiding this comment.
Both findings from review 5096582894 are resolved at this head: the repository housekeeping is now named explicitly, and the README claim is scoped to project repositories with .github accurately explained.
Re-verified M1-R2 and the task plan: the tracked tree is exactly the claimed retained set; the diff changes only CONTRIBUTING.md and the six intended removals; both reusable workflows, the calculate_version action, and profile assets are untouched. The receiving ha-sinkhole PR is merged as recorded.
Direct checks of the callers on their main branches confirm tradedesk and tradedesk-dukascopy still call both reusable workflows at @main, and reusable-release still calls calculate_version at @main. No new defect or unresolved decision gate found.
Nothing was served from these files: the repo has Pages set to build by workflow with no Pages workflow present, index.html only redirected to a /blog/ that no longer exists, and docs/index.md pointed at the removed radiusred/blog repository. _config.yml, mkdocs.yml and requirements.txt existed only to build that site. Refs M1-R2.
The old file described .github as holding cross-repo documentation and listed only the three tradedesk repositories. Say what the repository actually holds (the two reusable release workflows and their callers, the calculate_version action, the org profile), point contributors at each project's own CONTRIBUTING.md instead of a list that goes stale, and spell out how changes to this repository land: PR, linear history, conventional commits, and that callers pin the workflows at @main. Refs M1-R2.
The file only ever held ha-sinkhole's architecture diagrams. It now lives in that repository as .files/ha-sinkhole-architecture.drawio, next to the SVGs rendered from it, per the operator's decision on the task issue; the receiving PR is radiusred/ha-sinkhole#117. Refs M1-R2.
Review on PR #36 found two absolute statements that were untrue against the resulting tree: "Nothing else lives here" ignored .codecrew.yml, .gitignore and CONTRIBUTING.md itself, and "Every public repository carries its own README.md" was false for .github, whose profile/README.md is the organisation profile rather than a repository README. Name the housekeeping files and scope the README claim to project repositories. Refs M1-R2.
4f22ad5 to
66c7301
Compare
|
Rebased onto |
Closes #35
Task: #35 — milestone radiusred/ops#1 (M1), requirement M1-R2. Implemented by
radiusred-codyunder the implementer contract.What was done
chore: remove the dead blog scaffolding— deletes_config.yml,mkdocs.yml,index.html,docs/index.mdandrequirements.txt. Nothing is served from them: the repo's Pages setting isbuild_type: workflowwith no Pages workflow present,index.htmlonly redirected to a/blog/that no longer exists, anddocs/index.mdpointed at the removedradiusred/blogrepository. An org-wide code search finds no consumer ofradiusred.github.io/.github/.chore: move doc/radius-red.drawio to ha-sinkhole— removes the file here; it now lives in ha-sinkhole as.files/ha-sinkhole-architecture.drawio, overwriting the copy there and keeping that filename, per the operator's decision on #35. Receiving PR: docs: take the newer architecture draw.io source from radiusred/.github ha-sinkhole#117 (approved and merged as349328d). The SVGs there are not regenerated (not asked for); the ha-sinkhole PR notes that they are now behind the source.docs: make the CONTRIBUTING.md inventory claims exact— after review: names the housekeeping files (CONTRIBUTING.md,.gitignore,.codecrew.yml) instead of claiming the shared assets are everything, and scopes the "carries its own README.md" claim to project repositories, sinceprofile/README.mdis the org profile, not this repo's README.docs: rewrite CONTRIBUTING.md for the org as it is— the file now says what.githubholds (the two reusable release workflows and their callerstradedeskandtradedesk-dukascopy, thecalculate_versionaction, the org profile), points contributors at each project's ownCONTRIBUTING.mdinstead of a three-repo list that had gone stale, and spells out how changes here land: PR, rebase-only linear history, conventional commits, and that callers pin the workflows at@mainso a merge here is live for every caller at once.Untouched, by instruction:
.github/workflows/*,.github/actions/calculate_version/*,profile/*.After this PR
git ls-filesonmainwill be:.codecrew.yml, the two workflows, the action,.gitignore,CONTRIBUTING.md,profile/README.md,profile/banner-whitetext.webp.The draw.io gate, resolved
The task said to move
doc/radius-red.drawio"next to the SVGs it generates", but ha-sinkhole already held.files/ha-sinkhole-architecture.drawio, from which the shipped SVGs were rendered; this copy had diverged from it. Raised as a human gate on #35; the operator answered that this copy is the more up-to-date one and it should overwrite the ha-sinkhole file under that filename. Done in radiusred/ha-sinkhole#117, since merged.Requirements
Tests
None ride with this PR: no code changes. Everything removed or rewritten is prose or dead site configuration; the workflows and the action are untouched.
Notes for the operator
require-lintruleset makes "Lint commit messages" a required check on this repo'smain, but this repo has no commitlint workflow, so that check will never report on a.githubPR. Adding one is a workflow change and out of scope here. Both commits are conventional.task startcould not assign@radiusred-cody[bot]to Tidy the legacy content out of .github #35 (GitHub returned 403 on the assignment; Apps cannot be issue assignees). The task is otherwise started as the App.Deviations
None from the plan. The one gate is recorded on #35 (question and the operator's decision); the plan's change 3 landed as written once answered.