The release, release-cgo and autoupdate workflows must run on a runner
the caller chooses. Each workflow now takes a runner input. It accepts
a label or a JSON object that selects a runner group.
The workflow uses inputs.runner first. It then uses the
SHARED_ACTIONS_RUNNER variable and falls back to ubuntu-latest. Callers
that set nothing keep the old behaviour.
Every job now has a timeout. Release and autoupdate runs use a
concurrency group that never cancels a running job.
Lets a caller run the release jobs on a self-hosted runner via with.runner,
e.g. runner: self-hosted. Defaults to ubuntu-latest so existing consumers are
unaffected.
The release job now passes the calculated semver (bare VERSION and v-prefixed VERSION_TAG) into the repo-local workflow-prepare.sh hook's environment, so a consumer repo can stamp its version into source at build time. Needed for runtimes that cannot resolve their version from build info at runtime (e.g. Yaegi-interpreted Traefik plugins). Additive and backward-compatible: the test/version jobs run the hook without a version, and repos with no hook (or one that ignores VERSION) are unaffected.
lukaszraczylo