mirror of
https://github.com/hanzoai/benchmarks.git
synced 2026-10-30 14:59:11 +00:00
zeekayandHanzo Dev
377c4f4416
Twelve lanes in the order that needs the least, so the first failure names the smallest missing piece: node, then go, then the cloud source, then fetched data. A lane whose data nobody has fetched is SKIPPED and says which command fetches it, counted separately from the passes. Twelve ok over four silent skips is the same green a broken suite prints, and this suite has spent a week finding exactly that shape. It names the tree the binary lanes build from, above their results. Two different repositories check out as `cloud`; against the wrong one all five run and fail, which reads as five broken lanes rather than one wrong path. With the line printed, `hanzo-inc/cloud · c17b10d8` sits directly above four failures and the diagnosis is immediate. Verified from a fresh clone: 12 ok, 0 failed, 2 skipped with the source named. Co-authored-by: Hanzo Dev <[email protected]>
57 lines
2.4 KiB
Bash
Executable File
57 lines
2.4 KiB
Bash
Executable File
#!/usr/bin/env bash
|
|
# Does this checkout work?
|
|
#
|
|
# Not a benchmark: a benchmark tells you what something costs, and this tells you
|
|
# whether the thing that measures it runs at all. Every lane, in the order that
|
|
# needs the least, so the first failure names the smallest missing piece.
|
|
#
|
|
# bash check.sh # everything it can reach
|
|
# CLOUD_SRC=/path bash check.sh # including the four that build the binary
|
|
#
|
|
# A lane that needs data nobody has fetched is SKIPPED and says so. A lane that
|
|
# runs and fails is a failure. The difference matters: this file existing and
|
|
# reporting "0 failed" over four skipped lanes would be the same green a broken
|
|
# suite prints.
|
|
set -uo pipefail
|
|
cd "$(dirname "$0")"
|
|
ok=0 bad=0 skip=0
|
|
run() {
|
|
local name=$1; shift
|
|
printf ' %-32s ' "$name"
|
|
if "$@" >/tmp/bench-check.log 2>&1; then printf 'ok\n'; ok=$((ok + 1))
|
|
else printf 'FAIL — %s\n' "$(grep -vE '_nvm_lazy|^ld: warning' /tmp/bench-check.log | tail -1 | cut -c1-58)"; bad=$((bad + 1)); fi
|
|
}
|
|
skipped() { printf ' %-32s skipped — %s\n' "$1" "$2"; skip=$((skip + 1)); }
|
|
|
|
echo 'needs only node'
|
|
run 'fleet' node fleet/fleet.mjs /tmp/bench-check-fleet
|
|
run 'pricing' node pricing/pricing.mjs
|
|
run 'sandbox' node sandbox/sandbox.mjs
|
|
|
|
echo 'needs go'
|
|
if command -v go >/dev/null; then
|
|
run 'goroutine builds' env GOWORK=off go build -o /tmp/bench-check-g ./goroutine
|
|
run 'transport harness builds' env GOWORK=off go build -o /tmp/bench-check-t ./transport/harness
|
|
else
|
|
skipped 'goroutine' 'no go toolchain'; skipped 'transport harness' 'no go toolchain'
|
|
fi
|
|
|
|
echo 'needs the cloud source (CLOUD_SRC, or ../cloud)'
|
|
# NAME THE SUBJECT BEFORE MEASURING IT. These five build a binary, and two
|
|
# different repositories check out as `cloud`. Against the wrong one they run
|
|
# and fail, which reads as five broken lanes instead of one wrong path — so the
|
|
# tree is printed here, above its own results.
|
|
if . ./host.sh && cloud_src >/dev/null 2>&1; then
|
|
cloud_src >/dev/null 2>/tmp/bench-check-src; sed 's/^/ /' /tmp/bench-check-src | grep . | head -1
|
|
run 'self' bash self/run.sh
|
|
run 'self/modules' bash self/modules.sh
|
|
run 'cipher' bash cipher/run.sh
|
|
run 'egress' bash egress/run.sh
|
|
run 'transport' bash transport/run.sh
|
|
else
|
|
for l in self self/modules cipher egress transport; do skipped "$l" 'no cloud source — see README'; done
|
|
fi
|
|
|
|
printf '\n%d ok, %d failed, %d skipped\n' "$ok" "$bad" "$skip"
|
|
[ "$bad" -eq 0 ]
|