fix(release): CI owns the release stamp; :latest moves only from a release

Two beads, one root: four sites stamped a version and only two of them were CI.

`Taskfile.yml`'s build and `install.sh` derived GATEWAY_VERSION/RUNTIME_VERSION
from `git describe`, so a from-source build claimed a release it was not
running, and the stamp became a function of the commit graph -- the one input a
content-checksum build cache cannot see. An empty commit moved `git describe`
while `task build` reported "up to date" and shipped the previous stamp
(archie-core-pdq2). Both sites now stamp `dev`: a source build says so, and the
version a release carries is written into the artifact CI publishes. That
removes the non-file input rather than caching it, and traceability for a local
build is the checkout it came from.

`deploy.yml` published `:latest` from any push to main, so a release's own
stamping survived only until the next ordinary commit -- minutes, in an active
repo -- and a host deploying :latest reported dev, leaving the update path
unable to tell current from stale (archie-core-i5nb). `:latest` now moves only
from a commit that IS a release; every other build publishes `:edge` (mutable,
newest main) and `sha-<commit>` (immutable). A host that runs releases follows
`:latest`; a host that tracks main follows `:edge`, which RELEASING.md now says.

Verified: `task build` reports `archied dev`; the new workflow test fails
against the old workflow on all five guards and passes after; deploy.yml
parses; install.sh passes `bash -n`.

Gated with `task check` (exit 0) at 6046aaf4 -- the v1.39.0 base, CI-green --
with these five files applied. It is NOT gated at HEAD: HEAD carries an
unrelated red, `TestTaskProtoCarriesEveryDomainField`, because fe2a8e6d added
ParkClass and RemediationRounds without the proto mapping. This change neither
causes nor fixes that.
This commit is contained in:
sam committed 2026-09-23 13:09:01 +10:00
1 parent b9db00b5b6
commit 36118f4ee4
5 files changed
+87 -23

No files matched your search

+28 -11
View File
@@ -12,7 +12,7 @@ on:
# preceding fix, say) previously queued a full gate+build for each one,
# several minutes apiece, with no caching -- so a burst of pushes left
# three deploys racing in parallel for the same runner capacity. Only the
# last one's images matter (both jobs push :latest), so cancel a
# last one's images matter (each job pushes its image tags), so cancel a
# superseded run rather than let it fight for resources it can't use.
#
# Keyed on the commit, not github.ref: `git push --follow-tags` pushes
@@ -140,7 +140,7 @@ jobs:
exit 1
fi
if ! git tag --points-at HEAD | grep -qE '^(v[0-9]|archied/v|archie/v)' ; then
echo "::warning ::No release tags at HEAD, so this build publishes :latest for a commit that is not a release. It is stamped dev+<sha> and tagged sha-<sha> so it stays traceable and restorable, but it supersedes the previous release's :latest stamping until the next release is cut."
echo "This commit is not a release, so the images are published as :edge and sha-<sha>; :latest keeps naming the newest release."
fi
build-and-push:
@@ -184,12 +184,13 @@ jobs:
RUNTIME_VERSION="${RUNTIME_TAG#archie/}"
[ -n "$GATEWAY_VERSION" ] || GATEWAY_VERSION="${RELEASE_TAG#v}"
[ -n "$RUNTIME_VERSION" ] || RUNTIME_VERSION="${RELEASE_TAG#v}"
# Not a bare `dev`. A literal dev tells an operator nothing about which
# build they are running, and it is what made a release's stamping look
# transient: the next commit to main republished :latest as `dev`. The
# commit is appended so a main build is identifiable and is visibly not
# a release. (On an untagged push GITHUB_SHA IS the commit — the tag
# case is handled by the dereference above.)
# `dev+<sha>` is what a build that is NOT a release reports, and it is
# honest now that such a build no longer publishes :latest: the version
# names the commit it came from, and the sha-<sha> tag makes it
# restorable. Before, this same string reached :latest minutes after a
# release and made the release's own stamping read as dev
# (archie-core-i5nb). (On an untagged push GITHUB_SHA IS the commit —
# the tag case is handled by the dereference above.)
[ -n "$GATEWAY_VERSION" ] || GATEWAY_VERSION="dev+${GITHUB_SHA:0:12}"
[ -n "$RUNTIME_VERSION" ] || RUNTIME_VERSION="dev+${GITHUB_SHA:0:12}"
# Keep the raw tag presence separate from the version fallback: a
@@ -202,6 +203,24 @@ jobs:
echo "release_tag=$RELEASE_TAG" >> "$GITHUB_OUTPUT"
echo "gateway=$GATEWAY_VERSION" >> "$GITHUB_OUTPUT"
echo "runtime=$RUNTIME_VERSION" >> "$GITHUB_OUTPUT"
# :latest names the newest RELEASE, so only a commit that IS one
# publishes it. A push to main publishes :edge (mutable, newest main)
# and sha-<sha> (immutable) instead. That is what stops a release's
# :latest stamping from being overwritten minutes later by the next
# ordinary commit, which left a host reporting dev and made the update
# path unable to tell current from stale (archie-core-i5nb).
#
# `if` blocks rather than `[ -n "$X" ] && echo`: the latter returns
# non-zero when the test fails, and these steps run under `set -e`.
{
echo "archied_tags<<EOF"
if [ -n "$RELEASE_TAG" ]; then
echo "ghcr.io/samcharles93/archied:latest"
fi
echo "ghcr.io/samcharles93/archied:edge"
echo "ghcr.io/samcharles93/archied:sha-${{ steps.commit.outputs.sha }}"
echo "EOF"
} >> "$GITHUB_OUTPUT"
- name: Build and push archied
uses: docker/build-push-action@v6
@@ -209,9 +228,7 @@ jobs:
context: .
file: Dockerfile.archied
push: true
tags: |
ghcr.io/samcharles93/archied:latest
ghcr.io/samcharles93/archied:sha-${{ steps.commit.outputs.sha }}
tags: ${{ steps.versions.outputs.archied_tags }}
build-args: |
GATEWAY_VERSION=${{ steps.versions.outputs.gateway }}
RUNTIME_VERSION=${{ steps.versions.outputs.runtime }}