Create the release last so its tag never gets ahead of its binaries, and run the app job on Ubuntu only for pushes

release-app.yml now starts by hand with a version, title and notes (plan2code-publish does it). It builds the four systems, makes a draft release with every file attached and then publishes it, so the vX.Y.Z tag the update check watches appears only when the downloads exist. Publishing a release by hand or pushing a v* tag still starts it as a fallback.

plan2code-publish dispatches the workflow and watches it instead of running gh release create.

app.yml: pushes to main run the app job on Ubuntu only; the full three-system run stays on manual runs and the full-matrix label.
This commit is contained in:
jparkerweb committed 2026-11-06 10:43:22 -08:00
1 parent f28d4d592a
commit 76e0c44e0b
3 files changed
+88 -44

No files matched your search

+31 -29
View File
@@ -5,7 +5,7 @@ description: "Publish a GitHub Release for jparkerweb/plan2code whenever CHANGEL
# Plan2Code Release Publisher
Publish a GitHub Release for `jparkerweb/plan2code` whenever `CHANGELOG.md`'s top version is ahead of the latest published release on GitHub. This turns the merged CHANGELOG entry on `main` into an actual GitHub Release (which also creates the `vX.Y.Z` git tag).
Publish a GitHub Release for `jparkerweb/plan2code` whenever `CHANGELOG.md`'s top version is ahead of the latest published release on GitHub. This turns the merged CHANGELOG entry on `main` into an actual GitHub Release (the `release-app` workflow builds the app first and creates the release, with its `vX.Y.Z` git tag, last).
**Repo-local by design.** This skill lives in the repo's `.claude/skills/` and is intentionally NOT wired into `install.js` — it is a maintainer dev tool, not part of the shipped product, so it is never installed to `~/.claude/skills/`. Do **not** copy it there: the uninstaller (`uninstallFiles()` in `install.js`) and every re-install's pre-copy cleanup in `install()` both delete every entry matching `/^plan2code-/` under `~/.claude/skills/`, so a copy placed there would be silently removed.
@@ -103,57 +103,59 @@ Present all three to the user and wait for an explicit answer before any write.
If the user supplies a suffix, set `title = "v$CHANGELOG_VERSION " + <suffix>` (single space join) and proceed to publish. The tag and notes are unaffected by the suffix.
### Step 9 — Publish (on approval)
### Step 9 - Publish (on approval)
On approval, write `$notes` to a temp file — never pass multiline text inline via `--notes`, that regresses into a quoting bug — then create the release targeting `main`. Two requirements for the temp file: `$notes` must already hold **real newlines** (per Step 8) because `printf '%s'` / `WriteAllText` write it byte-for-byte — a literal `\n` in the string lands literally in the release body; and it MUST be **UTF-8** because the notes contain emoji (`🎉`, `🐛`, `✨`, `🔧`).
The release is created by the `release-app` workflow, **last**, not by this skill. The workflow builds the app on four systems, attaches every file to a draft release, and only then publishes it, so the `vX.Y.Z` tag (which the app's update check watches) never exists before its binaries do. Creating the release here with `gh release create` would put the tag up first and leave users offered an upgrade with nothing to download. Do not do that.
On approval, write `$notes` to a temp file, never inline (multiline text on the command line is a quoting bug). `$notes` must already hold **real newlines** (per Step 8) because `printf '%s'` / `WriteAllText` write it byte-for-byte, and the file MUST be **UTF-8** because the notes contain emoji. Then start the workflow on `main`, passing the file as the `notes` input (`-F key=@file` reads the file's contents):
**bash (preferred in this environment):**
```bash
NOTES_FILE=$(mktemp)
printf '%s' "$notes" > "$NOTES_FILE"
gh release create "$tag" --repo jparkerweb/plan2code --title "$title" --notes-file "$NOTES_FILE" --target main
gh workflow run release-app.yml --repo jparkerweb/plan2code --ref main -f version="$CHANGELOG_VERSION" -f title="$title" -F notes=@"$NOTES_FILE"
rm -f "$NOTES_FILE"
```
**PowerShell:** do NOT use `Set-Content` — under Windows PowerShell 5.1 it writes ANSI/UTF-16 by default and mangles the emoji into `??`. Write UTF-8 **without BOM** (a BOM would leak into the release body):
**PowerShell:** do NOT use `Set-Content` (under Windows PowerShell 5.1 it writes ANSI/UTF-16 and mangles the emoji into `??`). Write UTF-8 **without BOM**:
```powershell
$NotesFile = [System.IO.Path]::GetTempFileName()
[System.IO.File]::WriteAllText($NotesFile, $notes, [System.Text.UTF8Encoding]::new($false))
gh release create "$tag" --repo jparkerweb/plan2code --title "$title" --notes-file "$NotesFile" --target main
gh workflow run release-app.yml --repo jparkerweb/plan2code --ref main -f version="$CHANGELOG_VERSION" -f title="$title" -F notes=@"$NotesFile"
Remove-Item -Path $NotesFile
```
Always delete the temp file afterward, regardless of whether `gh release create` succeeded or failed.
Always delete the temp file afterward, whether or not the dispatch succeeded. If `gh workflow run` exits non-zero, show its error verbatim.
**Failure handling — already-exists classification:** if `gh release create` exits non-zero, inspect the error text.
- If and only if it contains the substring `already exists` (real output: `HTTP 422: Validation Failed` / `Release.tag_name already exists`), report this to the user as already published, not as a raw CLI error:
> This version ($CHANGELOG_VERSION) was already published as a release — nothing more to do.
- Every other failure (auth, network, permissions, etc.) must be surfaced to the user verbatim. Never silently reclassify a genuine failure as "already published."
### Step 10 — Verify
Confirm the release now exists and report its URL:
Then find the run it started and watch it to the end (the builds take a while; tell the user so):
```bash
gh release view "$tag" --repo jparkerweb/plan2code
gh run list --repo jparkerweb/plan2code --workflow release-app.yml --event workflow_dispatch --limit 1 --json databaseId,status,createdAt
gh run watch <databaseId> --repo jparkerweb/plan2code --exit-status
```
Report the release URL to the user.
If the run list shows a run that started before the dispatch, wait a few seconds and list again. If the run fails, show which job failed and stop: no release or tag exists yet (a failed run leaves at most an unpublished draft, which a re-run reuses). Never publish a release by hand to get around a failed build.
### Step 10 - Verify
Confirm the release now exists, is published (not a draft) and carries the files, and report its URL:
```bash
gh release view "$tag" --repo jparkerweb/plan2code --json url,isDraft,assets -q '.url, .isDraft, (.assets | length)'
```
Expect `false` for the draft flag and five assets (four binaries and the SHA256SUMS file). Report the release URL to the user.
## Rules
- **Repo-local only** — this skill is not part of the installed product; never add it to `install.js`, and never copy it to `~/.claude/skills/` (the uninstaller deletes `plan2code-*` entries there).
- **Read-only on version files** — never modify `CHANGELOG.md`, `version.json`, or `package.json`; this skill only reads and compares them.
- **Version-sync gate is hard** (Step 4) — if the three version sources disagree, stop and report; do not publish a release from an inconsistent repo.
- **Never run a local `git tag` or `git push`** — tag creation is delegated entirely to `gh release create --target main`.
- **Always use `--notes-file`**, never inline multiline `--notes`, and always write the notes file as UTF-8 (no BOM) so emoji survive.
- **Repo-local only**: this skill is not part of the installed product; never add it to `install.js`, and never copy it to `~/.claude/skills/` (the uninstaller deletes `plan2code-*` entries there).
- **Read-only on version files**: never modify `CHANGELOG.md`, `version.json`, or `package.json`; this skill only reads and compares them.
- **Version-sync gate is hard** (Step 4): if the three version sources disagree, stop and report; do not publish a release from an inconsistent repo.
- **Never run a local `git tag` or `git push`, and never run `gh release create` yourself**: the `release-app` workflow creates the release and its tag after the builds, so the tag never gets ahead of its binaries.
- **Always pass the notes through a UTF-8 (no BOM) temp file**, never inline, so emoji survive.
- **Always clean up the temp notes file**, even on a mid-run error.
- **Always get explicit approval before any write** (Step 8) — no release is created without a yes (or a yes-with-suffix).
- **Derive the current version from the highest published semver tag** (Step 5), never from GitHub's manually-pinnable "Latest" flag. On an empty or failed release list, treat it as "no prior release" (baseline `0.0.0`) — no error-text matching needed.
- **On a `gh release create` failure** (Step 9), only reclassify as already-published when the error contains `already exists` — every other failure must be shown verbatim, never swallowed.
- **Idempotent and safe to re-run** at any time — re-running after a successful publish hits the Step 7 no-op; re-running after a race-lost publish hits the Step 9 already-exists handling.
- **Always get explicit approval before any write** (Step 8): no workflow run is started without a yes (or a yes-with-suffix).
- **Derive the current version from the highest published semver tag** (Step 5), never from GitHub's manually-pinnable "Latest" flag. On an empty or failed release list, treat it as "no prior release" (baseline `0.0.0`); no error-text matching needed.
- **Idempotent and safe to re-run** at any time: re-running after a successful publish hits the Step 7 no-op; re-running after a failed build starts a fresh run, which reuses any leftover draft.
+4 -3
View File
@@ -17,8 +17,9 @@ concurrency:
group: app-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
# Speed: a pull request runs on Ubuntu only (the cheapest runner) unless it carries the
# `full-matrix` label; pushes to main and manual runs use all three systems. A change that
# Speed: pushes and pull requests run on Ubuntu only (the cheapest runner) unless a pull request
# carries the `full-matrix` label; manual runs use all three systems. The release build
# (release-app.yml) compiles on Windows and macOS too. A change that
# touches only the website, the README, the changelog or similar skips the heavy `app` job. The
# fast `prose` job runs on every push and pull request. The root `npm test` suite does not run on
# pushes or pull requests at all: run it locally, or start `node-tests` by hand.
@@ -76,7 +77,7 @@ jobs:
strategy:
fail-fast: false
matrix:
os: ${{ fromJSON(github.event_name == 'pull_request' && !contains(github.event.pull_request.labels.*.name, 'full-matrix') && '["ubuntu-latest"]' || '["windows-latest","macos-latest","ubuntu-latest"]') }}
os: ${{ fromJSON((github.event_name == 'workflow_dispatch' || contains(github.event.pull_request.labels.*.name, 'full-matrix')) && '["windows-latest","macos-latest","ubuntu-latest"]' || '["ubuntu-latest"]') }}
defaults:
run:
working-directory: plan2code-app
+53 -12
View File
@@ -1,7 +1,12 @@
name: release-app
# Builds the desktop app on four systems when a GitHub release is published, or a v* tag is
# pushed ahead of one, and attaches one file per system, named
# The normal way to release is to start this workflow by hand (`plan2code-publish` does it) with
# the version and the release notes. It builds first and creates the release LAST, so the vX.Y.Z
# tag, which is what the app's update check looks for, only appears once every file is attached
# and nobody is offered an upgrade that has no binary yet. A release published by hand, or a v*
# tag pushed ahead of one, still starts it as a fallback; the tag then exists before the files.
#
# Builds the desktop app on four systems and attaches one file per system, named
# plan2code-app-<version>-<os>-<arch>[.exe], plus one checksum file
# (plan2code-app-<version>-SHA256SUMS.txt). install.js fetches the matching file and checks it
# against that list. Linux ships the plain binary, not an AppImage: docs/desktop-app.md documents
@@ -9,6 +14,20 @@ name: release-app
# how to allow it.
on:
workflow_dispatch:
inputs:
version:
description: 'Version to release, without the v (it must match CHANGELOG.md, version.json and package.json)'
required: true
type: string
title:
description: 'Release title (defaults to the tag)'
required: false
type: string
notes:
description: 'Release notes (Markdown)'
required: false
type: string
release:
types: [published]
push:
@@ -22,9 +41,13 @@ permissions:
# A release published after its tag push starts a second run for the same tag: queue it behind
# the first so two builds never upload over each other.
concurrency:
group: release-app-${{ github.event.release.tag_name || github.ref_name }}
group: release-app-${{ github.event.release.tag_name || (inputs.version && format('v{0}', inputs.version)) || github.ref_name }}
cancel-in-progress: false
env:
# The release tag: the published release's, the version a manual run was given, or the pushed tag.
TAG: ${{ github.event.release.tag_name || (inputs.version && format('v{0}', inputs.version)) || github.ref_name }}
jobs:
build:
name: build (${{ matrix.label }})
@@ -50,8 +73,6 @@ jobs:
cache: npm
cache-dependency-path: plan2code-app/package-lock.json
- name: Check the versions agree with the release tag
env:
TAG: ${{ github.event.release.tag_name || github.ref_name }}
run: node scripts/check-versions.mjs --tag "$TAG"
- name: Install the pinned Rust toolchain
working-directory: plan2code-app
@@ -97,7 +118,6 @@ jobs:
contents: write
env:
GH_TOKEN: ${{ github.token }}
TAG: ${{ github.event.release.tag_name || github.ref_name }}
steps:
- uses: actions/checkout@v4.2.2
- uses: actions/download-artifact@v4.3.0
@@ -123,9 +143,30 @@ jobs:
cd dist
sha256sum plan2code-app-* > "plan2code-app-${VERSION}-SHA256SUMS.txt"
cat "plan2code-app-${VERSION}-SHA256SUMS.txt"
- name: Make sure the release exists
# A tag push alone names no release; create one (same tag, generated notes) so the
# upload below always has somewhere to go, whichever trigger ran it.
run: gh release view "$TAG" || gh release create "$TAG" --title "$TAG" --generate-notes
- name: Attach to the release
run: gh release upload "$TAG" dist/* --clobber
- name: Attach to the release, creating it last
env:
RELEASE_TITLE: ${{ inputs.title }}
RELEASE_NOTES: ${{ inputs.notes }}
# No release yet (a manual run, or a bare tag push): create it as a DRAFT with every file
# attached, then publish it, so the tag appears only when the files are already there.
# A release that already exists (published by hand) just gets the files. A draft left
# by an earlier failed run is reused. Releases made with the workflow's own token do not
# start another run.
run: |
state="$(gh release list --limit 100 --json tagName,isDraft -q ".[] | select(.tagName==\"$TAG\") | .isDraft")"
if [ -z "$state" ]; then
notes_file="$(mktemp)"
if [ -n "$RELEASE_NOTES" ]; then
printf '%s' "$RELEASE_NOTES" > "$notes_file"
notes_args=(--notes-file "$notes_file")
else
notes_args=(--generate-notes)
fi
gh release create "$TAG" dist/* --draft --target "$GITHUB_SHA" --title "${RELEASE_TITLE:-$TAG}" "${notes_args[@]}"
state=true
else
gh release upload "$TAG" dist/* --clobber
fi
if [ "$state" = "true" ]; then
gh release edit "$TAG" --draft=false
fi