vouchington-tooling 0.9.0 → 0.10.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vouchington-tooling",
-  "version": "0.9.0",
+  "version": "0.10.0",
   "description": "Vouchington CLI and extractable tooling libraries.",
   "homepage": "https://github.com/vouchington/vouchington-tooling/tree/main/packages/vouchington-tooling#readme",
   "bugs": {

package/scripts/gha/assert-vitest-test-passed.sh

@@ -0,0 +1,27 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Guards against a `it.runIf(...)` (or equivalent conditional test) silently
+# degrading to a skip: `--reporter=json` records every test's status, and a
+# skipped or absent test must fail the job, not report green. A CI job whose
+# only assertion is "did the suite exit 0" cannot tell "ran and passed" apart
+# from "never ran at all".
+
+report_path="${1:?usage: assert-vitest-test-passed.sh <report-json-path> <full-test-name>}"
+full_name="${2:?usage: assert-vitest-test-passed.sh <report-json-path> <full-test-name>}"
+
+if [ ! -r "$report_path" ]; then
+  echo "::error::vitest report not found at $report_path"
+  exit 1
+fi
+
+status=$(jq -r --arg name "$full_name" '
+  [.testResults[].assertionResults[] | select(.fullName == $name) | .status] | first // "absent"
+' "$report_path")
+
+echo "test \"$full_name\": $status"
+
+if [ "$status" != passed ]; then
+  echo "::error::required test did not pass (status: $status) -- a skipped or absent test must not report green"
+  exit 1
+fi

package/scripts/gha/host-pressure-diagnostics.sh

@@ -10,6 +10,26 @@ section() {
   printf '\n== %s ==\n' "$1"
 }
 
+is_number() {
+  [[ "$1" =~ ^[0-9]+(\.[0-9]+)?$ ]]
+}
+
+show_load_per_cpu() {
+  load1=$1
+  load5=$2
+  load15=$3
+  cpu_count=$4
+  printf 'load average: %s %s %s\n' "${load1:-unavailable}" "${load5:-unavailable}" "${load15:-unavailable}"
+  ratio=''
+  if is_number "$load1" && is_number "$cpu_count" && [ "$cpu_count" != '0' ]; then
+    # Capture rather than let awk print directly: a forked subprocess can
+    # fail or be killed under host pressure -- the exact condition this
+    # script exists to diagnose -- and produce no output on either branch.
+    ratio=$(awk -v load="$load1" -v cpus="$cpu_count" 'BEGIN { printf "%.2f", load / cpus }' 2>/dev/null || true)
+  fi
+  printf 'load1 per cpu: %s\n' "${ratio:-unavailable}"
+}
+
 show_meminfo_field() {
   field=$1
   if [ -r /proc/meminfo ]; then
@@ -54,14 +74,24 @@ show_oom_lines() {
 
 show_top_rss_processes() {
   if have ps && have sort && have sed; then
-    ps ax -o pid= -o rss= -o comm= 2>/dev/null | sort -k2,2nr | sed -n '1,15p'
+    output=$(ps ax -o pid= -o rss= -o comm= 2>/dev/null | sort -k2,2nr | sed -n '1,15p')
+  else
+    output=''
+  fi
+  if [ -n "$output" ]; then
+    printf '%s\n' "$output"
   else
-    printf 'unavailable\n'
+    printf 'unavailable: top rss processes\n'
   fi
 }
 
 show_runner_counts() {
-  ps ax -o comm= 2>/dev/null | awk '
+  process_list=$(ps ax -o comm= 2>/dev/null)
+  if [ -z "$process_list" ]; then
+    printf 'unavailable: process list\n'
+    return
+  fi
+  printf '%s\n' "$process_list" | awk '
     {
       command = $0
       sub(/^[[:space:]]+/, "", command)
@@ -137,18 +167,21 @@ case "$platform" in
     done
 
     section 'load and cpu'
+    loadavg_line=''
     if [ -r /proc/loadavg ]; then
-      printf 'load average: '
-      awk '{ print $1, $2, $3 }' /proc/loadavg 2>/dev/null || printf 'unavailable\n'
-    else
-      printf 'load average unavailable\n'
+      loadavg_line=$(awk '{ print $1, $2, $3 }' /proc/loadavg 2>/dev/null || true)
     fi
+    load1=$(printf '%s' "$loadavg_line" | awk '{ print $1 }')
+    load5=$(printf '%s' "$loadavg_line" | awk '{ print $2 }')
+    load15=$(printf '%s' "$loadavg_line" | awk '{ print $3 }')
     printf 'nproc: '
     if have nproc; then
-      nproc 2>/dev/null || printf 'unavailable\n'
+      cpu_count=$(nproc 2>/dev/null || true)
     else
-      getconf _NPROCESSORS_ONLN 2>/dev/null || printf 'unavailable\n'
+      cpu_count=$(getconf _NPROCESSORS_ONLN 2>/dev/null || true)
     fi
+    printf '%s\n' "${cpu_count:-unavailable}"
+    show_load_per_cpu "$load1" "$load5" "$load15" "$cpu_count"
 
     section 'kernel OOM evidence'
     show_oom_lines
@@ -160,13 +193,69 @@ case "$platform" in
   Darwin)
     section 'system basics'
     uptime 2>/dev/null || printf 'uptime unavailable\n'
-    printf 'hw.ncpu: '
-    sysctl -n hw.ncpu 2>/dev/null || printf 'unavailable\n'
     printf 'hw.memsize: '
     sysctl -n hw.memsize 2>/dev/null || printf 'unavailable\n'
     section 'swap and vm'
     vm_stat 2>/dev/null || printf 'vm_stat unavailable\n'
     sysctl vm.swapusage 2>/dev/null || printf 'vm.swapusage unavailable\n'
+
+    section 'load and cpu'
+    loadavg_raw=$(sysctl -n vm.loadavg 2>/dev/null || true)
+    load1=$(printf '%s' "$loadavg_raw" | awk '{ print $2 }')
+    load5=$(printf '%s' "$loadavg_raw" | awk '{ print $3 }')
+    load15=$(printf '%s' "$loadavg_raw" | awk '{ print $4 }')
+    printf 'hw.logicalcpu: '
+    logicalcpu=$(sysctl -n hw.logicalcpu 2>/dev/null || true)
+    printf '%s\n' "${logicalcpu:-unavailable}"
+    printf 'hw.physicalcpu: '
+    sysctl -n hw.physicalcpu 2>/dev/null || printf 'unavailable\n'
+    printf 'hw.ncpu: '
+    sysctl -n hw.ncpu 2>/dev/null || printf 'unavailable\n'
+    show_load_per_cpu "$load1" "$load5" "$load15" "$logicalcpu"
+    # Apple-Silicon performance/efficiency core split. These sysctls do not
+    # exist on Intel Macs and must degrade to unavailable there. On some
+    # runners the OID resolves with exit 0 but zero bytes of output rather
+    # than failing outright -- capture and fall back on emptiness, not just
+    # on a non-zero exit, or that case silently swallows the trailing
+    # newline and glues the next section onto this line.
+    perflevel0=$(sysctl -n hw.perflevel0.logicalcpu 2>/dev/null || true)
+    printf 'hw.perflevel0.logicalcpu: %s\n' "${perflevel0:-unavailable}"
+    perflevel1=$(sysctl -n hw.perflevel1.logicalcpu 2>/dev/null || true)
+    printf 'hw.perflevel1.logicalcpu: %s\n' "${perflevel1:-unavailable}"
+    # iostat -c 2 is the closest PSI analogue on Darwin: a since-boot row plus
+    # a 1-second live row with us/sy/id and 1m/5m/15m load. Do not gate this on
+    # `have timeout` -- base macOS ships no timeout binary (only gtimeout via
+    # coreutils/Homebrew), and `-c 2` is self-bounding at ~1s regardless.
+    if have iostat; then
+      if have timeout; then
+        timeout 5 iostat -c 2 2>/dev/null || printf 'iostat unavailable\n'
+      else
+        iostat -c 2 2>/dev/null || printf 'iostat unavailable\n'
+      fi
+    else
+      printf 'iostat unavailable\n'
+    fi
+
+    section 'memory pressure'
+    if have memory_pressure; then
+      memory_pressure -Q 2>/dev/null || printf 'memory_pressure unavailable\n'
+    else
+      printf 'memory_pressure unavailable\n'
+    fi
+    printf 'kern.memorystatus_vm_pressure_level: '
+    pressure_level=$(sysctl -n kern.memorystatus_vm_pressure_level 2>/dev/null || true)
+    case "$pressure_level" in
+      1) printf 'normal (1)\n' ;;
+      2) printf 'warning (2)\n' ;;
+      4) printf 'critical (4)\n' ;;
+      '') printf 'unavailable\n' ;;
+      *) printf 'unknown (%s)\n' "$pressure_level" ;;
+    esac
+    printf 'vm.compressor_bytes_used: '
+    sysctl -n vm.compressor_bytes_used 2>/dev/null || printf 'unavailable\n'
+    # No jetsam/OOM-kill log scrape here (unlike Linux show_oom_lines): `log
+    # show` is unbounded and slow, with no cheap bounded equivalent on macOS.
+
     section 'top rss processes'
     show_top_rss_processes
     section 'runner worker/listener counts'