fix(pruner): evict whole bun packages instead of single files

.github/workflows/build.yml

@@ -10,6 +10,20 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
+  scripts:
+    name: Shell scripts
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+
+      - name: shellcheck
+        run: shellcheck -x pruner.sh tests/*.sh
+
+      - name: tests
+        run: |
+          set -euo pipefail
+          for t in tests/*.sh; do bash "$t"; done
+
   matrix:
     name: Generate build matrix
     runs-on: ubuntu-latest
@@ -83,14 +97,14 @@ jobs:
 
   gate:
     name: All builds passed
-    needs: [matrix, build]
+    needs: [scripts, matrix, build]
     if: always()
     runs-on: ubuntu-latest
     steps:
       - name: Check that all required jobs succeeded
         run: |
-          if [ "${{ needs.matrix.result }}" != "success" ] || [ "${{ needs.build.result }}" != "success" ]; then
-            echo "::error::A required job failed or was cancelled (matrix: ${{ needs.matrix.result }}, build: ${{ needs.build.result }})"
+          if [ "${{ needs.scripts.result }}" != "success" ] || [ "${{ needs.matrix.result }}" != "success" ] || [ "${{ needs.build.result }}" != "success" ]; then
+            echo "::error::A required job failed or was cancelled (scripts: ${{ needs.scripts.result }}, matrix: ${{ needs.matrix.result }}, build: ${{ needs.build.result }})"
             exit 1
           fi
           echo "All builds succeeded"

README.md

@@ -125,7 +125,12 @@ The `pruner` service ([`pruner.sh`](pruner.sh)) keeps slow work off the job
 path:
 
 - **Nightly:** trims the Gradle, bun, mise and hk caches to `CACHE_CAP_BYTES`
-  each (default 20 GiB), dropping the least recently used files first. It then
+  each (default 20 GiB), least recently used first. The bun cache drops whole
+  package folders (`install/cache/<pkg>@<ver>@@@1`, by the atime of their
+  `package.json`) and never `bin/`: bun reuses a package folder as complete
+  whenever it exists, so a single evicted file would go missing from every
+  later install. mise tool versions are dropped whole too; the Gradle, mise
+  download and hk caches drop single files. It then
   runs `docker system prune --filter until=24h` and
   `docker builder prune --filter until=72h`. Named volumes are never pruned.
 - **Hourly:** fills the action archive cache, capped at 2 GiB.

pruner.sh

@@ -6,7 +6,9 @@
 # 21-39 s of runner time each time and freeing nothing (issue #16). Here they
 # run once a day, off the job path:
 #   1. LRU size caps on the per-runner cache volumes (mounted under
-#      $CACHES_ROOT/<runner>/<name>, see docker-compose.yml).
+#      $CACHES_ROOT/<runner>/<name>, see docker-compose.yml). Caches whose
+#      entries are only valid whole (bun packages, mise tool versions) evict
+#      whole entries; the rest evict single files.
 #   2. docker system prune / builder prune on the host daemon (docker.sock).
 #   3. Hourly, in the background: fill the shared action archive cache.
 #
@@ -79,7 +81,7 @@ lru_cap() {
     rec="${rec#*	}"        # drop atime
     fsize="${rec%%	*}"
     path="${rec#*	}"
-    [ -n "$path" ] && [ -n "$fsize" ] || continue
+    if [ -z "$path" ] || [ -z "$fsize" ]; then continue; fi
     rm -f "$path" 2>/dev/null || true
     freed=$((freed + fsize))
   done < <(find "$dir" -type f -printf '%A@\t%s\t%p\0' 2>/dev/null | sort -z -n)
@@ -108,16 +110,59 @@ lru_cap_mise_installs() {
   log "$installs freed ~${freed}B"
 }
 
+# bun treats install/cache/<pkg>@<ver>@@@<n>/ as complete whenever the folder
+# exists, so a file evicted from it is silently missing from every later
+# node_modules (next@16.3.8 lost dist/compiled/server-only/empty.js, breaking
+# every Next build). 0-byte files went first: copying them never reads, so
+# their atime never moved. Evict whole package folders instead, least recently
+# used first by the atime of their package.json (read on every install and
+# import), until the whole volume ($1, mounted from /root/.bun) fits under the
+# cap. Never touches $1/bin (the bun binary itself) or the @t@ manifest piles.
+lru_cap_bun() {
+  local dir="$1" cap="${2:-$CACHE_CAP_BYTES}"
+  local cache="$dir/install/cache"
+  local excess freed=0 rec path dsize base t
+  [ -d "$cache" ] || return 0
+  # Leftovers of interrupted extractions; the age guard spares live installs.
+  find "$cache" -mindepth 1 -maxdepth 2 ! -name .tmp \( -name '*.tmp' -o -name '*.part' \) \
+    -mmin +60 -exec rm -rf {} + 2>/dev/null
+  find "$cache/.tmp" -mindepth 1 -maxdepth 1 -mmin +60 -exec rm -rf {} + 2>/dev/null
+  excess=$(excess_bytes "$dir" "$cap") || return 0
+  log "evicting least recently used packages from $cache"
+  while IFS= read -r -d '' rec; do
+    [ "$freed" -ge "$excess" ] && break
+    path="${rec#*	}"
+    [ -d "$path" ] || continue
+    dsize=$(dir_bytes "$path")
+    rm -rf "$path" 2>/dev/null || true
+    freed=$((freed + ${dsize:-0}))
+    # Drop the <pkg>/<ver>@@@<n> symlink pointing at it, and the dir if empty.
+    base="${path##*/}"
+    rm -f "${path%/*}/${base%%@*}/${base#*@}" 2>/dev/null
+    rmdir "${path%/*}/${base%%@*}" 2>/dev/null
+  done < <(
+    # <pkg>@<ver>... at the top level, @scope/<pkg>@<ver>... one level down.
+    find "$cache" -mindepth 1 -maxdepth 2 -type d -name '*@*' ! -name '@*' \
+      \( -path "$cache/@*/*" -o ! -path "$cache/*/*" \) -print0 2>/dev/null |
+      while IFS= read -r -d '' path; do
+        t=$(stat -c %X "$path/package.json" 2>/dev/null || stat -c %Y "$path" 2>/dev/null) || continue
+        printf '%s\t%s\0' "$t" "$path"
+      done | sort -z -n
+  )
+  find "$cache" -mindepth 1 -maxdepth 1 -name '@*' ! -name '@t@' -type d -empty -delete 2>/dev/null || true
+  log "$cache freed ~${freed}B"
+}
+
 prune_caches() {
   local runner
   for runner in "$CACHES_ROOT"/*/; do
     [ -d "$runner" ] || continue
     runner="${runner%/}"
     log "capping caches under $runner"
     lru_cap "$runner/gradle"
-    lru_cap "$runner/bun"
+    lru_cap_bun "$runner/bun"
     lru_cap "$runner/mise-cache"
-    lru_cap "$runner/hk"
+    lru_cap "$runner/hk" # configs/<hash>.json + pklr/packages/<hash>.zip: whole files
     lru_cap_mise_installs "$runner/mise-data"
   done
   lru_cap "$ACTION_CACHE_DIR" "$ACTION_CACHE_CAP_BYTES"
@@ -146,8 +191,11 @@ prefetch_one() {
   local name="$1" ref="$2" sha="$2" dir file tmp
   dir="$ACTION_CACHE_DIR/${name//\//_}"
   if ! [[ "$sha" =~ ^[0-9a-f]{40}$ ]]; then
-    sha=$(gh_curl -H 'Accept: application/vnd.github.sha' "$GITHUB_API/repos/$name/commits/$ref") &&
-      [[ "$sha" =~ ^[0-9a-f]{40}$ ]] || { log "prefetch: cannot resolve $name@$ref" >&2; return 1; }
+    sha=$(gh_curl -H 'Accept: application/vnd.github.sha' "$GITHUB_API/repos/$name/commits/$ref")
+    if ! [[ "$sha" =~ ^[0-9a-f]{40}$ ]]; then
+      log "prefetch: cannot resolve $name@$ref" >&2
+      return 1
+    fi
   fi
   file="$dir/$sha.tar.gz"
   [ -s "$file" ] && return 0
@@ -191,7 +239,7 @@ prefetch_actions() {
   refs=$(sort -u "$work/refs")
   rm -rf "$work"
   while read -r name ref; do
-    [ -n "$name" ] && [ -n "$ref" ] || continue
+    if [ -z "$name" ] || [ -z "$ref" ]; then continue; fi
     n=$((n + 1))
     if out=$(prefetch_one "$name" "$ref"); then
       [ "$out" = 1 ] && fetched=$((fetched + 1))
@@ -225,6 +273,9 @@ run_once() {
   log "done. Disk free: ${before} -> ${after}"
 }
 
+# Sourced (tests/): define the functions only.
+[ "${BASH_SOURCE[0]}" = "$0" ] || return 0
+
 if [ "${1:-}" = prefetch ]; then
   prefetch_actions
   exit 0

tests/pruner-bun.sh

@@ -0,0 +1,103 @@
+#!/bin/bash
+# pruner.sh must evict whole bun packages, never single files: bun reuses a
+# cache folder as complete whenever it exists. Builds a fake /root/.bun volume
+# (packages holding 0-byte files with old atimes, a scoped package, the
+# <pkg>/<ver> symlinks, @t@ piles and bin/bun), runs prune_caches under a small
+# cap, then checks every surviving package is complete and every evicted one
+# is fully gone. Needs GNU find/stat/du (Linux, or `docker run ubuntu`):
+#   bash tests/pruner-bun.sh
+set -euo pipefail
+
+here="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+work="$(mktemp -d)"
+trap 'rm -rf "$work"' EXIT
+
+fails=0
+fail() { echo "FAIL: $*"; fails=$((fails + 1)); }
+
+bun="$work/caches/1/bun"
+cache="$bun/install/cache"
+mkdir -p "$bun/bin" "$cache/@t@"
+head -c 300000 /dev/zero >"$bun/bin/bun"
+ln -s /root/.bun/bin/bun "$bun/bin/bunx"
+head -c 2000 /dev/zero >"$cache/@t@/0061322e0ad01aa4.pile"
+
+# make_pkg <entry relative to cache> <age in days of package.json atime>
+make_pkg() {
+  local entry="$1" days="$2" d="$cache/$1" base
+  mkdir -p "$d/dist/compiled/server-only"
+  head -c 100000 /dev/urandom >"$d/index.js"
+  echo '{"name":"x"}' >"$d/package.json"
+  : >"$d/dist/compiled/server-only/empty.js" # 0-byte, atime never refreshes
+  : >"$d/LICENSE.empty"
+  # 0-byte files are the oldest thing in the volume, as in production.
+  find "$d" -type f -empty -exec touch -a -d '400 days ago' {} +
+  touch -a -d "$days days ago" "$d/package.json"
+  base="${entry##*/}"
+  mkdir -p "$(dirname "$d")/${base%%@*}"
+  ln -s "/root/.bun/install/cache/$entry" "$(dirname "$d")/${base%%@*}/${base#*@}"
+  find "$d" -type f | sed "s|^$d/||" | sort >"$work/${entry//\//_}.files"
+}
+
+make_pkg 'old-a@1.0.0@@@1' 90
+make_pkg 'old-b@2.0.0@@@1' 80
+make_pkg '@scope/old-c@3.0.0@@@1' 70
+make_pkg 'next@16.3.8@@@1' 1
+make_pkg 'next@16.4.2@@@1' 2
+make_pkg '@scope/fresh@1.0.0@@@1' 3
+make_pkg 'nopkgjson@1.0.0@@@1' 0
+rm "$cache/nopkgjson@1.0.0@@@1/package.json" # falls back to dir mtime
+sed -i '/^package.json$/d' "$work/nopkgjson@1.0.0@@@1.files"
+touch -d '1 hour ago' "$cache/nopkgjson@1.0.0@@@1"
+
+# Stale and live partial downloads.
+mkdir -p "$cache/.tmp/stale" "$cache/.tmp/live"
+touch -d '2 days ago' "$cache/.tmp/stale"
+head -c 10 /dev/zero >"$cache/half@1.0.0@@@1.tgz.part"
+touch -d '2 days ago' "$cache/half@1.0.0@@@1.tgz.part"
+
+# Cap = volume minus the three oldest packages (old-a, old-b, @scope/old-c),
+# plus half a package of slack: exactly those three must go, nothing else.
+sz() { du -sb "$1" | awk '{print $1}'; }
+cap=$(($(sz "$bun") - $(sz "$cache/old-a@1.0.0@@@1") - $(sz "$cache/old-b@2.0.0@@@1") \
+  - $(sz "$cache/@scope/old-c@3.0.0@@@1") - 2 * 4096 + 50000))
+# shellcheck source=pruner.sh
+source "$here/../pruner.sh"
+CACHES_ROOT="$work/caches" CACHE_CAP_BYTES="$cap" prune_caches
+
+for entry in 'old-a@1.0.0@@@1' 'old-b@2.0.0@@@1' '@scope/old-c@3.0.0@@@1'; do
+  base="${entry##*/}"
+  [ ! -e "$cache/$entry" ] || fail "$entry should be evicted"
+  [ ! -L "$(dirname "$cache/$entry")/${base%%@*}/${base#*@}" ] || fail "$entry symlink left behind"
+  [ ! -e "$(dirname "$cache/$entry")/${base%%@*}" ] || fail "$entry name dir left behind"
+done
+
+for entry in 'next@16.3.8@@@1' 'next@16.4.2@@@1' '@scope/fresh@1.0.0@@@1' 'nopkgjson@1.0.0@@@1'; do
+  d="$cache/$entry"
+  if [ ! -d "$d" ]; then
+    fail "$entry should survive"
+    continue
+  fi
+  diff <(find "$d" -type f | sed "s|^$d/||" | sort) "$work/${entry//\//_}.files" >/dev/null ||
+    fail "$entry is incomplete"
+  base="${entry##*/}"
+  [ -L "$(dirname "$d")/${base%%@*}/${base#*@}" ] || fail "$entry symlink removed"
+done
+
+[ -f "$cache/next@16.3.8@@@1/dist/compiled/server-only/empty.js" ] || fail "next lost server-only/empty.js"
+[ -s "$bun/bin/bun" ] || fail "bin/bun removed"
+[ -L "$bun/bin/bunx" ] || fail "bin/bunx removed"
+[ -f "$cache/@t@/0061322e0ad01aa4.pile" ] || fail "@t@ pile removed"
+[ -d "$cache/@scope" ] || fail "@scope dir removed while it still holds a package"
+[ ! -e "$cache/.tmp/stale" ] || fail "stale .tmp entry kept"
+[ -e "$cache/.tmp/live" ] || fail "live .tmp entry removed"
+[ ! -e "$cache/half@1.0.0@@@1.tgz.part" ] || fail "stale .part kept"
+
+size=$(sz "$bun")
+[ "$size" -le "$cap" ] || fail "volume ${size}B still over the ${cap}B cap"
+
+if [ "$fails" -gt 0 ]; then
+  echo "$fails check(s) failed"
+  exit 1
+fi
+echo "PASS: bun cache evicted by whole package (volume ${size}B)"