Read the ecosystem's own definition of the two test-outcome rules #44

Workflow file for this run

	name: Go

	on:
	push:
	branches: [master]
	pull_request:
	branches: [master]

	jobs:
	# Everything that needs no database and no library: the formatting, and that
	# every build tag combination compiles and vets. No test runs here, because a
	# test collected without the database it reads can only skip, and a skip is a
	# test nobody has run reported inside a green job.
	build:
	name: Build and vet, every tag
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@v4

	- uses: actions/setup-go@v5
	with:
	go-version-file: go.mod
	cache: true

	- name: Formatting
	run: \|
	unformatted=$(gofmt -l .)
	if [ -n "$unformatted" ]; then
	echo "gofmt reports:"; echo "$unformatted"; exit 1
	fi

	# A file behind a build tag no job ever passes is invisible to every check
	# the repository has: it is not compiled, so it is not vetted, and it
	# carries no test a count could miss. Compiling each tag set here is what
	# keeps backend_duck.go and backend_spark.go from rotting; the meos tag
	# needs the library and is compiled in the job below, which builds it.
	- name: Vet and build, default tags
	run: go vet ./... && go build ./...

	- name: Vet and build, duckdb backend
	run: go vet -tags duckdb ./... && go build -tags duckdb ./...

	- name: Vet and build, spark backend
	run: go vet -tags spark ./... && go build -tags spark ./...

	# EVERY TEST THIS REPOSITORY CARRIES RUNS HERE, AND NONE OF THEM SKIPS.
	# A test skips when a precondition is missing from the job that collects it,
	# so this job supplies all three: the fixture database the Annex A data tests
	# read, the libmeos the streaming engine links, and the network the published
	# OGC schemas are fetched over. It then refuses a skip outright, so a
	# precondition that stops being supplied fails the job instead of decaying
	# into a green run that exercised nothing.
	#
	# THE DATABASE IS MOBILITYDB MASTER, BUILT HERE FROM SOURCE. A job pinned to a
	# release measures a MobilityDB the project no longer develops, so a defect fixed
	# upstream reads here as a live one and blocks work that is not blocked: deleting
	# a temporal value that spans a whole composing sequence segfaults the backend on
	# 1.3.0 and answers correctly on master. Naming a version somebody can install is
	# a reason to ALSO run a release, never a reason to develop against one.
	#
	# AND A PUBLISHED IMAGE IS NOT THE SOURCE. The mobilitydb/mobilitydb tags are
	# built by MobilityDB/MobilityDB-docker, a separate repository on its own schedule,
	# so what they carry is that repository's packaging rather than the source this
	# tier is written against: `18-3.6-master` fails to load the library on a missing
	# libgdal.so.39, and `17-3.5-master` exits its own entrypoint because the
	# pointcloud extension it asks for is not available. Pinning whichever tag happens
	# to work makes the packaging the source of truth and dates the measurement to
	# whenever that tag was pushed. The recipe below is MobilityDB's own
	# .github/workflows/pgversion.yml, read from the checkout under test.
	conformance:
	name: Every test, against MobilityDB master
	runs-on: ubuntu-24.04
	env:
	PGVERSION: "18"
	POSTGISVERSION: "3"
	MFAPI_DSN: postgres://postgres:conformance@127.0.0.1:5432/mfapi?sslmode=disable
	# The vendored OGC schema bundle is compared against what OGC publishes.
	MFAPI_SCHEMA_FRESHNESS: "1"
	# The floor is every result line `go test -v` writes, subtests included:
	# 77 top-level tests and 64 subtests. Subtests belong in it because a
	# t.Run whose body skips is as unrun as its parent would be, and deleting
	# a case from a table-driven test moves nothing else. Raise it when the
	# suite grows. Lowering it belongs in the same commit as the removal it
	# accounts for, because deleting a test is what a skip refusal cannot see.
	MFAPI_TEST_FLOOR: "141"
	steps:
	- uses: actions/checkout@v4

	- name: Check out MobilityDB master
	uses: actions/checkout@v4
	with:
	repository: MobilityDB/MobilityDB
	ref: master
	path: mobilitydb

	- name: Name the commit under test
	run: git -C mobilitydb log -1 --format='MobilityDB master %H %s'

	- uses: actions/setup-go@v5
	with:
	go-version-file: go.mod
	cache: true

	- name: Remove the runner's PostgreSQL, so apt.postgresql.org answers
	run: \|
	sudo service postgresql stop \|\| true
	sudo apt-get --purge remove postgresql* -y \|\| true
	sudo rm -rf /var/lib/postgresql/ /etc/postgresql/ /var/log/postgresql/ \|\| true

	# MobilityDB's own apt hardening, taken from the checkout above rather than
	# copied: the runner image ships two package sources that answer 403 from the
	# Azure runners, which fails an apt-get update asking only for Ubuntu packages.
	- uses: ./mobilitydb/.github/actions/apt-resilient

	- name: Add the PostgreSQL APT repository
	run: \|
	sudo apt-get install -y curl ca-certificates gnupg
	curl https://www.postgresql.org/media/keys/ACCC4CF8.asc \| sudo apt-key add -
	codename=$(lsb_release -cs)
	echo "deb http://apt.postgresql.org/pub/repos/apt/ ${codename}-pgdg main ${PGVERSION}" \
	\| sudo tee /etc/apt/sources.list.d/pgdg.list

	- name: Install the build and server dependencies
	run: \|
	sudo apt-get update
	sudo apt-get install -y \
	libgeos++-dev \
	libproj-dev \
	libgdal-dev \
	libjson-c-dev \
	postgresql-${PGVERSION} \
	postgresql-${PGVERSION}-postgis-${POSTGISVERSION} \
	postgresql-server-dev-${PGVERSION}

	# The default family set is what this tier reads: temporal geometry points
	# and the scalar temporal types. The optional families carry dependencies of
	# their own and no route here reaches one, so the build stays at the defaults
	# and the job measures the surface it uses.
	- name: Build and install MobilityDB
	run: \|
	export PATH=/usr/lib/postgresql/${PGVERSION}/bin:$PATH
	mkdir mobilitydb/build
	cd mobilitydb/build
	cmake -DCMAKE_BUILD_TYPE=Release ..
	make -j "$(nproc)"
	sudo make install

	# The same source, configured as the standalone library the streaming
	# engine links. Building it from the checkout above is what makes the
	# engine tests measure the same MobilityDB the SQL routes measure; a
	# libmeos from anywhere else would be a second, unnamed commit.
	- name: Build and install libmeos from the same checkout
	run: \|
	mkdir mobilitydb/build-meos
	cd mobilitydb/build-meos
	cmake -DMEOS=ON -DCMAKE_BUILD_TYPE=Release ..
	make -j "$(nproc)"
	sudo make install
	sudo ldconfig

	# LOADING THE EXTENSION IS THE ASSERTION THE PUBLISHED IMAGES FAILED. Doing
	# it in a step of its own, ahead of the fixture, is what makes a database that
	# cannot carry MobilityDB say so here rather than inside a fixture load.
	- name: Create the database and load the extension
	run: \|
	sudo service postgresql start
	sudo -u postgres psql -v ON_ERROR_STOP=1 \
	-c "ALTER USER postgres PASSWORD 'conformance'"
	sudo -u postgres createdb mfapi
	psql "$MFAPI_DSN" -v ON_ERROR_STOP=1 -q \
	-c "CREATE EXTENSION mobilitydb CASCADE"
	psql "$MFAPI_DSN" -X -A -t -c "SELECT mobilitydbFullVersion()"

	- name: Load the conformance fixture
	run: psql "$MFAPI_DSN" -v ON_ERROR_STOP=1 -q -f tutorial/setup/load_conformance.sql

	# THE EXIT STATUS OF A PIPELINE IS ITS LAST STAGE'S. Without pipefail the
	# status read here is tee's, which is always 0, and a failing test reaches the
	# guard below as though it had passed.
	- name: The whole suite, with every precondition supplied
	run: \|
	set -o pipefail
	go test -tags meos -v ./... 2>&1 \| tee suite.log

	# A SKIP IS NOT A PASS, AND A DELETION IS NOT VISIBLE IN THE SKIP COUNT.
	# Two rules, because each is blind to what the other catches. The first
	# refuses a test that reported itself unrun; once that stands, deleting a
	# test is the remaining way to stop running it, and it leaves no skip
	# behind, so only the total moves.
	#
	# The rules live once, in MEOS-API beside the catalog, and every consumer
	# in the ecosystem reads that one definition rather than a copy of it: a
	# copy is what drifts, and a rule about coverage that drifts is worse than
	# none. The same script answers over a local log, so a by-hand answer and
	# the CI answer cannot differ.
	- name: Refuse a skip, and a suite that shrank
	uses: MobilityDB/MEOS-API/.github/actions/check-test-outcome@master
	with:
	log: suite.log
	min-tests: ${{ env.MFAPI_TEST_FLOOR }}

	# A SAMPLE THAT HAS DRIFTED IS WORSE THAN NO SAMPLE: it states what the
	# service used to answer, and a reader has no way to tell. Re-emitting here
	# and diffing against what is committed is what keeps samples/ the current
	# answer rather than a snapshot somebody forgot to refresh.
	#
	# The TemporalProperties document carries the time it was written, which the
	# standard defines it to carry, so that one line is normalised on both sides.
	- name: The committed samples are what the service writes
	run: \|
	set -o pipefail
	go run . -emit samples.emitted
	norm() { sed -E 's/"timeStamp": "[^"]*"/"timeStamp": "<written>"/' "$1"; }
	rc=0
	for f in samples/*; do
	b=$(basename "$f")
	if [ ! -f "samples.emitted/$b" ]; then
	echo "::error::samples/$b is no longer a resource the service serves"; rc=1; continue
	fi
	if ! diff -q <(norm "$f") <(norm "samples.emitted/$b") >/dev/null; then
	echo "::error::samples/$b is not what the service writes; re-run: go run . -emit samples"
	diff -u <(norm "$f") <(norm "samples.emitted/$b") \| head -40
	rc=1
	fi
	done
	for f in samples.emitted/*; do
	b=$(basename "$f")
	if [ ! -f "samples/$b" ]; then
	echo "::error::the service serves a resource samples/ does not carry: $b"; rc=1
	fi
	done
	if [ "$rc" -eq 0 ]; then
	echo "$(ls samples/*.json \| wc -l) sample documents are what the service writes"
	fi
	exit $rc

Navigation Menu

Read the ecosystem's own definition of the two test-outcome rules #44

Workflow file

Read the ecosystem's own definition of the two test-outcome rules #44

Uh oh!

Workflow file for this run