mirror of
https://github.com/hanzoai/ci.git
synced 2026-10-27 09:19:34 +00:00
zeekayandHanzo Dev
a0de959480
Hanzo CI/CD / cicd (push) Waiting to run
CI/CD / gate (push) Waiting to run
The delegate lane assembled its own image ref -- repo:sha-<short7>-amd64 -- and the door takes one, so that was the only tag those repos ever published. The fleet pins semver and pin.sh refuses anything else, so every repo on mode: delegate built green and could not be pinned or deployed at all; ghcr.io/hanzoai/console has semvers up to 8.5.199 and sha- tags for every build since it switched. The tag shape is now bin/imgtags, one rule read by both lanes, and the delegate lane derives its number with bin/imgver like the buildx lane always has. The door still gets one ref -- the immutable sha- one -- and the version is added to the manifest it pushed with crane tag, so both names carry one digest. Posting a second build for the second tag would build the same commit twice and produce two. The completion signal is the sha- tag resolving at the registry rather than a status read: the door answers 202 and reports nothing further, and the manifest existing is what the alias needs anyway. A version that does not land fails the run. bin/imgtags_test.sh pins the rule and asserts both lanes take their tags from it. The four gates that read the pipeline still named .hanzo/workflows, where it has not lived since the move, so they read nothing and passed; they read .github/workflows now, as do bin/kind_test.sh and bin/tally_test.sh. Co-authored-by: Hanzo Dev <[email protected]>
93 lines
6.1 KiB
Bash
93 lines
6.1 KiB
Bash
#!/usr/bin/env bash
|
|
# Tests for the `tally` function that decides whether a test gate reported any
|
|
# result. There is no bin/tally: the function lives inside the one definition of
|
|
# the pipeline, .github/workflows/build.yml, and this suite LIFTS IT OUT and runs
|
|
# it. So there is nothing to keep in step — what is tested here is the same text
|
|
# the runner executes, and a rule edited in one place cannot drift from a rule
|
|
# tested in another.
|
|
#
|
|
# It has to be lifted rather than called because a run cannot reach a new file in
|
|
# this repo: the tools checkout derives its ref from GITHUB_WORKFLOW_REF, which
|
|
# git.hanzo.ai does not set, so $CI_HOME resolves to whatever `v1` names rather
|
|
# than to the tag the caller pinned.
|
|
#
|
|
# Every fixture is output captured from a real gate in the fleet, colour bytes
|
|
# and all, because colour is what the first version of this got wrong.
|
|
# Offline and deterministic: strings in, a number out, no network.
|
|
# Run: bash bin/tally_test.sh
|
|
set -uo pipefail
|
|
cd "$(dirname "$0")/.."
|
|
DEF=.github/workflows/build.yml
|
|
tmp=$(mktemp -d); trap 'rm -rf "$tmp"' EXIT
|
|
fail=0
|
|
|
|
sed -n '/^ # tally BEGIN$/,/^ # tally END$/p' "$DEF" | sed 's/^ //' > "$tmp/tally.sh"
|
|
grep -q '^tally() {' "$tmp/tally.sh" || {
|
|
echo "FAIL could not lift tally() out of $DEF — the BEGIN/END markers moved, fix this suite"; exit 1; }
|
|
# shellcheck disable=SC1090
|
|
. "$tmp/tally.sh"
|
|
|
|
E=$'\033' # the escapes every JS and Python runner writes on CI
|
|
|
|
# t <name> <want> <log...>
|
|
t() {
|
|
local name=$1 want=$2; shift 2
|
|
printf '%s\n' "$@" > "$tmp/log"
|
|
local got; got=$(tally "$tmp/log")
|
|
if [ "$got" = "$want" ]; then printf 'ok %-58s %s\n' "$name" "$got"
|
|
else printf 'FAIL %-58s got=%s want=%s\n' "$name" "$got" "$want"; fail=1; fi
|
|
}
|
|
|
|
echo "COUNTED — a runner reporting results"
|
|
t "go, per package" 1 "ok github.com/hanzoai/ci 0.005s"
|
|
t "go, three packages" 3 "ok x/a 0.1s" "ok x/b (cached)" "ok x/c 0.2s"
|
|
t "go -v, per test beats package" 3 "--- PASS: TestA (0.00s)" "--- PASS: TestB (0.00s)" "--- FAIL: TestC (0.00s)" "FAIL x/a 0.1s"
|
|
t "go, a failing package" 1 "FAIL github.com/hanzoai/x 0.104s"
|
|
t "gotestsum, summed over runs" 80 "DONE 40 tests in 3.399s" "DONE 40 tests in 1.100s"
|
|
t "cargo" 5 "test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured"
|
|
t "cargo, with failures" 7 "test result: FAILED. 5 passed; 2 failed; 0 ignored"
|
|
t "rust -v" 3 "test tally::a ... ok" "test tally::b ... FAILED" "test tally::c ... ignored"
|
|
t "mocha" 415 " 412 passing (2s)" " 3 failing"
|
|
# ava attaches the count to `tests` rather than to the verb, so the rules that
|
|
# read the word before passed/failed find `tests` there. Both captured from the
|
|
# fleet: hanzoai/openid-client and hanzoai/insights-node.
|
|
t "ava" 94 " 94 tests passed"
|
|
t "ava, with a failure" 94 " 1 test failed" " 93 tests passed"
|
|
t "bun test, per case" 4 "(pass) api > get [0.52ms]" "(pass) api > put [0.3ms]" "(pass) api > del [0.2ms]" "(fail) api > patch [0.1ms]"
|
|
t "TAP" 2 "ok 1 - the first" "not ok 2 - the second"
|
|
t "a bin/*_test.sh suite" 17 " ok clean repo is green" " ok the mark is refused" "" "vendormark_test: 17 passed, 0 failed"
|
|
t "pytest, plain summary" 90 "======================== 90 passed, 15 skipped in 0.34s ========================"
|
|
t "pytest, node ids" 3 "tests/t.py::TestA::test_x PASSED" "tests/t.py::TestA::test_y PASSED" "tests/t.py::test_z FAILED"
|
|
|
|
echo
|
|
echo "COLOUR — the same runners as CI actually writes them"
|
|
t "pytest in colour" 90 "tests/t.py::TestA::test_x ${E}[32mPASSED${E}[0m" \
|
|
"${E}[32m======================== ${E}[32m${E}[1m90 passed${E}[0m, ${E}[33m15 skipped${E}[0m${E}[32m in 0.34s${E}[0m${E}[32m ========================${E}[0m"
|
|
t "vitest in colour" 9 "${E}[2m Test Files ${E}[22m ${E}[1m${E}[32m1 passed${E}[39m${E}[22m${E}[90m (1)${E}[39m" \
|
|
"${E}[2m Tests ${E}[22m ${E}[1m${E}[32m9 passed${E}[39m${E}[22m${E}[90m (9)${E}[39m"
|
|
t "jest in colour" 412 "${E}[1mTests:${E}[22m ${E}[1m${E}[32m411 passed${E}[39m${E}[22m, ${E}[1m1 failed${E}[22m, 412 total"
|
|
t "a progress bar's carriage returns" 1 $'#=#=# 50.0%\r######## 100.0%\rok x/a 0.1s'
|
|
|
|
echo
|
|
echo "NOT COUNTED — a gate that is not a test, and a runner that ran nothing"
|
|
t "empty output" 0 ""
|
|
t "go, no test files" 0 "? github.com/hanzoai/x [no test files]"
|
|
t "go, build tag excluded the suite" 0 "testing: warning: no tests to run" "PASS" "ok x/a 0.002s [no tests to run]"
|
|
t "an assertion gate saying OK:" 0 "OK: one definition (2058 lines), no forwarder, no second address"
|
|
t "a linter" 0 "shell: no authored utility classes in 36 files ✓" "☔️ success workspaces valid!"
|
|
t "a typescript build" 0 "> [email protected] build" "> tsc && tsc -p tsconfig.esm.json"
|
|
t "npm install noise" 0 "added 30 packages in 1s" "14 high severity vulnerabilities" "60 packages are looking for funding"
|
|
t "a turbo build" 0 "• Packages in scope: @hanzo/gui, @hanzogui/core" "• Running build in 197 packages" "cache bypass, force executing 5701289d"
|
|
t "a word ending in ok" 0 "notebook something" "OK" "hook fired"
|
|
t "vitest reporting nothing" 0 "${E}[2m Tests ${E}[22m ${E}[1m${E}[32m0 passed${E}[39m${E}[22m${E}[90m (0)${E}[39m"
|
|
|
|
echo
|
|
echo "NO DOUBLE COUNT — one run reported twice is still one run"
|
|
t "cargo prints its own total once" 5 "test tally::a ... ok" "test tally::b ... ok" "test tally::c ... ok" "test tally::d ... ok" "test tally::e ... ok" "test result: ok. 5 passed; 0 failed"
|
|
t "pytest ids and summary agree" 3 "tests/t.py::a PASSED" "tests/t.py::b PASSED" "tests/t.py::c PASSED" "===== 3 passed in 0.1s ====="
|
|
t "vitest files line is not a total" 9 "${E}[2m Test Files ${E}[22m 1 passed (1)" "${E}[2m Tests ${E}[22m 9 passed (9)"
|
|
|
|
echo
|
|
[ $fail = 0 ] && echo "all tally tests passed" || echo "tally tests FAILED"
|
|
exit $fail
|