Compare commits

..

3 commits

Author SHA1 Message Date
Philip Peterson
accb20a488 vendor-images.sh: skopeo never read CONTAINERS_REGISTRIES_CONF
Not a 1.22 change, as the last commit said: 1.16 ignored it too.

[skip ci]

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-02 20:43:43 -09:00
Philip Peterson
8ab8a2162e vendor-images.sh: skopeo --registries-conf, not the env variable
skopeo 1.22 ignores CONTAINERS_REGISTRIES_CONF, so the mirror was copied
onto itself.

[skip ci]

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-02 19:50:47 -09:00
Philip Peterson
2c79c38515 iso.yml: prune the build's images afterwards
Two builds in a day filled the runner's disk (v0.1.0-alpha.1's first run).

[skip ci]

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-02 15:41:42 -09:00
2 changed files with 10 additions and 1 deletions

View file

@ -15,7 +15,12 @@ set -uo pipefail
mirror=${MIRROR:-forge.quinefoundation.com/quine-global-labs/mirror}
# Read from upstream itself, not from the mirror registries.conf points at.
export CONTAINERS_REGISTRIES_CONF=/dev/null
# The flag, not CONTAINERS_REGISTRIES_CONF: podman and buildah read that
# variable, but skopeo never has. Until 2026-11-03 this exported it, skopeo
# read the mirror, and every run copied the mirror onto itself, so tags
# stayed at their first copy. (runner-x86-linux's skopeo now honors the
# variable too; petersweb-infra, hosts/runner-x86-linux.)
skopeo() { command skopeo --registries-conf /dev/null "$@"; }
for ref in "$@"; do
src="docker://$ref"

View file

@ -183,3 +183,7 @@ jobs:
run: |
rm -rf output release
podman logout "$REGISTRY" || true
# Each build leaves ~100 GB of images and stages; the runner's daily
# prune keeps a day's worth, and a second build that day filled the
# disk. The next build pulls its bases from the forge again.
podman image prune -af >/dev/null || true