feat(cargo-cache): hardlink-clone a per-ref Cargo cache from a published snapshot

Replaces the phase-0 resolution probe with the real actions, merging the two
independent per-branch Cargo cache implementations on this forge into the
design neither of them had.

## The merge

- zemyna seeds a PR branch by `cp -al` hardlink clone (near-free: cost scales
  with inode count, not bytes) from the base branch's LIVE target dir — a
  torn read waiting for a second job slot (its own #911).
- emowheel seeds from a PUBLISHED IMMUTABLE SNAPSHOT (no race by
  construction) but with `cp -a`, duplicating ~35 GB per branch.

This ships hardlink-clone FROM a published snapshot: zemyna's cost profile,
emowheel's soundness, and #911 closed structurally rather than by the runner
happening to have one execution slot.

## The bug both implementations have

A build inside a `cp -al` clone DOES mutate the directory it was cloned from.
Cargo replaces real artifacts, but writes its metadata — and build scripts
write their OUT_DIR — with a plain truncating write, straight through the
shared inode. Measured set: `.fingerprint/<unit>/dep-<target>` (under
CARGO_UNSTABLE_CHECKSUM_FRESHNESS), `build/<pkg>/{output,root-output,out/**}`,
`deps/*.d` and `<profile>/*.d`.

The checksum-freshness case is a wrong answer, not a slow build: a PR clone
rewrites the base's dep-info to describe the PR's sources while the base's
cache still holds the artifact built from the base's; once the PR merges, the
base's next run finds the checksums match, reports `Fresh`, and links a binary
built from the pre-merge code. Reproduced end to end.

Fix: hardlink the artifacts (the GB), real-copy the metadata (the MB) — about
3.7% of a 6.9 GB Bevy target dir, against 100% for a full copy.

## Contents

- `cargo-cache/action.yml` — consume: resolve keys, seed from the base's
  snapshot via staging + one atomic rename, strip Cargo lock files, unshare
  the mutable paths, restore mtimes from git history, lock, prune.
- `cargo-cache-publish/action.yml` — publish: record the build watermark,
  atomically republish the snapshot on a protected branch, release the lock
  (`mode: release-lock` for the `if: always()` step).
- `scripts/` — all logic, so it is testable standalone; the YAML is wiring.
- `scripts/*selftest.sh` + `selftest.sh` — five suites, 63 assertions, every
  fix paired with a control that reproduces the bug. All green locally.

Eviction merges emowheel's liveness pass (dead branches pruned
unconditionally, not gated on disk pressure) with LRU-under-pressure, but
inverts the order within the pressure pass: `target-*` before `snapshot-*`,
because a snapshot is hardlinked to everything cloned from it, so evicting one
frees almost no real bytes while costing every future PR its warm start.

restore-mtimes.sh is ported from emowheel (the watermark variant, which closes
the merge hazard zemyna's copy still has) with its provenance de-projectised.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sqh2vscfzisk83VuPVQX9L
This commit is contained in:
2026-08-23 14:18:01 -05:00
co-authored by Claude Opus 5
parent 8503883138
commit 248af3061e
17 changed files with 2693 additions and 14 deletions
+174
View File
@@ -0,0 +1,174 @@
#!/usr/bin/env bash
# Regression test for prune-cache.sh. Builds a real scratch git repo standing
# in for `origin` and a real scratch directory standing in for the cache root,
# then runs the ACTUAL script against both — not a simulation of its logic.
#
# What each scenario demonstrates, and why the controls matter as much as the
# fixes (a scenario that always passes proves nothing):
#
# 1. DEAD BRANCH PRUNED, not gated on disk pressure — a cache whose branch
# no longer exists on origin is removed even with plenty of free space.
# Waiting for pressure to notice means paying for dead caches until then.
# 2. LIVE BRANCH SURVIVES despite being OLDER than the dead one — liveness,
# not age, is what decides pass 1.
# 3. PROTECTED REFS NEVER EVICTED under forced disk pressure, even when
# their caches are the oldest on disk and would rank first for LRU.
# 4. LOCKED CACHE PROTECTED even when dead, old, and under pressure.
# 5. STALE LOCK NOT HONOURED FOREVER — the same cache with a lock older than
# STALE_LOCK_SECONDS is evicted, so a crashed job cannot pin a directory
# permanently.
# 6. LIVENESS UNAVAILABLE FAILS SAFE — origin unreachable: a genuinely dead
# cache is NOT pruned, the log says so plainly, and the pressure fallback
# still works independently. "Unavailable" degrades to pressure-only, not
# to no eviction at all.
# 7. TARGET DIRS EVICTED BEFORE SNAPSHOTS — the ordering that differs from
# the obvious one. A snapshot is hardlinked to the caches cloned from it,
# so evicting it frees almost nothing while costing every future PR its
# warm start.
# 8. SELF-CLEAR REPORTS LOUDLY to the job summary, not just a log warning.
# 9. OWN CACHE NEVER EVICTED by a sibling pass.
# 10. SCOPED TO THE CACHE ROOT — a decoy outside it (standing in for another
# project's volume) is never touched.
set -euo pipefail
script_dir=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
. "$script_dir/cache-lib.sh"
prune="$script_dir/prune-cache.sh"
scratch=$(mktemp -d)
trap 'rm -rf "$scratch"' EXIT
pass_count=0
fail() { echo "ASSERTION FAILED: $*" >&2; [ -n "${1:-}" ] && [ -f "$scratch/log" ] && { echo "--- log ---" >&2; cat "$scratch/log" >&2; }; exit 1; }
ok() { pass_count=$((pass_count + 1)); echo "PASS: $*"; }
assert_gone() { [ -e "$1" ] && fail "expected gone: $1 ($2)"; ok "$2"; }
assert_kept() { [ -e "$1" ] || fail "expected kept: $1 ($2)"; ok "$2"; }
assert_log() { grep -q -- "$1" "$scratch/log" || fail "expected in log: $1 ($2)"; ok "$2"; }
echo "=== building a scratch origin with real branches ==="
origin="$scratch/origin.git"; git init -q --bare "$origin"
work="$scratch/work"; git init -q "$work"
(
cd "$work"
git -c user.email=t@t -c user.name=t commit -q --allow-empty -m init
git branch -M main
git checkout -q -b dev; git -c user.email=t@t -c user.name=t commit -q --allow-empty -m dev
git checkout -q -b feat/live; git -c user.email=t@t -c user.name=t commit -q --allow-empty -m live
git remote add origin "$origin"
git push -q origin main dev feat/live
)
cd "$work"
MAIN=$(cache_key main); DEV=$(cache_key dev); LIVE=$(cache_key feat/live)
DEAD=$(cache_key feat/dead); OWN=$(cache_key feat/own)
root="$scratch/cache"
mk() { mkdir -p "$root/$1"; head -c 4096 /dev/zero > "$root/$1/blob"; touch -d "$2" "$root/$1/.cache-last-used"; }
reset_cache() {
rm -rf "$root"; mkdir -p "$root"
mk "target-$MAIN" '2020-01-01'
mk "snapshot-$MAIN" '2020-01-01'
mk "target-$DEV" '2020-01-01'
mk "snapshot-$DEV" '2020-01-01'
mk "target-$LIVE" '2020-01-02' # older than the dead one, deliberately
mk "target-$DEAD" '2030-01-01' # newest on disk, but its branch is gone
mk "snapshot-$DEAD" '2030-01-01'
mk "target-$OWN" '2025-01-01'
}
run_prune() {
local free="${1:-}"
CACHE_DF_OVERRIDE="$free" GITHUB_STEP_SUMMARY="$scratch/summary" \
bash "$prune" "$root" "$root/target-$OWN" "dev main" 10 > "$scratch/log" 2>&1 \
|| { cat "$scratch/log"; fail "prune-cache.sh exited non-zero"; }
}
echo
echo "=== 1/2: dead pruned unconditionally; older-but-live survives ==="
reset_cache
run_prune "1000000 900000" # 90% free: no pressure at all
assert_gone "$root/target-$DEAD" "dead branch's target dir pruned with no disk pressure"
assert_gone "$root/snapshot-$DEAD" "dead branch's snapshot pruned too"
assert_kept "$root/target-$LIVE" "live branch survives despite an older marker than the dead one"
assert_log "no matching branch on origin" "eviction reason reported"
echo
echo "=== 3: protected refs never evicted under forced pressure ==="
reset_cache
run_prune "1000000 1000" # 0.1% free
assert_kept "$root/target-$DEV" "dev's target dir survives disk pressure"
assert_kept "$root/snapshot-$DEV" "dev's snapshot survives disk pressure"
assert_kept "$root/target-$MAIN" "main's target dir survives disk pressure"
assert_kept "$root/snapshot-$MAIN" "main's snapshot survives disk pressure"
echo
echo "=== 9: own cache never evicted by a sibling pass ==="
assert_kept "$root/target-$OWN" "this run's own cache survives"
echo
echo "=== 7: target dirs evicted before snapshots ==="
reset_cache
# Only the live branch is evictable; give it both a target dir and a snapshot
# with identical markers so ordering, not age, decides.
mk "snapshot-$LIVE" '2020-01-02'
# The df override is a fixed reading, so the pressure loop drains everything
# evictable — which is what makes the ORDER the observable property here, not
# what survives. Assert the eviction order directly from the log.
run_prune "1000000 1000"
order=$(grep -o "evicting \(target\|snapshot\)-$LIVE" "$scratch/log" | sed "s/evicting //")
[ "$(printf '%s\n' "$order" | head -1)" = "target-$LIVE" ] \
|| fail "expected target-$LIVE to be evicted before snapshot-$LIVE, got: $order"
ok "target dirs are evicted before snapshots"
echo
echo "=== 4: a fresh lock protects a dead, old, under-pressure cache ==="
reset_cache
date +%s > "$root/target-$DEAD/.ci-lock-ci-1"
run_prune "1000000 1000"
assert_kept "$root/target-$DEAD" "locked cache survives both passes"
assert_log "held open by" "lock reported in the log"
echo
echo "=== 5: a stale lock is not honoured forever ==="
reset_cache
echo 0 > "$root/target-$DEAD/.ci-lock-ci-1"
touch -d '2000-01-01' "$root/target-$DEAD/.ci-lock-ci-1"
run_prune "1000000 900000"
assert_gone "$root/target-$DEAD" "cache with an abandoned lock is evicted"
assert_log "treating as abandoned" "abandoned lock reported in the log"
echo
echo "=== 6: liveness unavailable fails safe, pressure fallback still works ==="
reset_cache
(
cd "$work" && git remote set-url origin "$scratch/nonexistent.git"
)
run_prune "1000000 900000" # no pressure
assert_kept "$root/target-$DEAD" "dead cache NOT pruned when liveness is unavailable"
assert_log "treating as UNAVAILABLE" "unavailability reported plainly, not folded into 'no branches'"
run_prune "1000000 1000" # now with pressure
if [ -e "$root/target-$DEAD" ] && [ -e "$root/target-$LIVE" ]; then
fail "pressure fallback did nothing when liveness was unavailable"
fi
ok "pressure fallback still evicts when liveness is unavailable"
(cd "$work" && git remote set-url origin "$origin")
echo
echo "=== 8: self-clear reports to the job summary ==="
reset_cache
rm -rf "$root/target-$DEAD" "$root/snapshot-$DEAD" "$root/target-$LIVE"
: > "$scratch/summary"
run_prune "1000000 1000" # nothing evictable left but the run's own cache
assert_log "clearing own" "self-clear reported in the log"
grep -q 'self-clear' "$scratch/summary" || fail "self-clear missing from the job summary"
ok "self-clear reported to the job summary, not only the log"
[ -d "$root/target-$OWN" ] || fail "self-clear left the own directory missing"
[ -z "$(ls -A "$root/target-$OWN")" ] || fail "self-clear did not actually empty the directory"
ok "own cache wiped and recreated empty"
echo
echo "=== 10: scoped to the cache root ==="
reset_cache
decoy="$scratch/other-project"; mkdir -p "$decoy/target-$DEAD"; touch "$decoy/target-$DEAD/blob"
run_prune "1000000 1000"
assert_kept "$decoy/target-$DEAD" "a cache outside the cache root is never touched"
echo
echo "prune-cache-selftest: ${pass_count} assertions passed"