test: the matrix must reconcile registered suites against accounted ones (#916)

.github/workflows/ci.yml

@@ -472,7 +472,7 @@ jobs:
       - name: Run the suite matrix on PG ${{ matrix.pg }}
         run: |
           set -euo pipefail
-          # PGC_SKIP_TIMING drops the three wall-clock suites: a shared runner
+          # PGC_SKIP_TIMING drops the four wall-clock suites: a shared runner
           # cannot hold a ratio still, and a gate that reds for reasons unrelated
           # to the change is worse than one that does not run. They stay in the
           # local matrix, which is where those numbers mean anything.

bench/build_citus.sh

@@ -28,7 +28,9 @@ echo "== building Citus $TAG against $("$PG_CONFIG" --version) ($PG_CONFIG)"
 
 # Benchmark arms must run on a non-assert PostgreSQL. An assert build's numbers
 # are invalid, so refuse one the way the rest of the bench tooling does.
-if "$PG_CONFIG" --configure | grep -q -- '--enable-cassert'; then
+# grep -c, not grep -q; see test/lib.sh's pgc_is_columnar_scan.
+_cfg="$("$PG_CONFIG" --configure 2>/dev/null || true)"
+if [ "$(grep -c -- '--enable-cassert' <<<"$_cfg" || true)" != 0 ]; then
 	echo "REFUSING: $PG_CONFIG is an assert build; benchmark numbers from it are invalid" >&2
 	exit 1
 fi

bench/build_timescaledb.sh

@@ -28,7 +28,9 @@ echo "== building TimescaleDB $VERSION against $("$PG_CONFIG" --version) ($PG_CO
 
 # Benchmark arms must run on a non-assert PostgreSQL. An assert build's numbers
 # are invalid, so refuse one the way the rest of the bench tooling does.
-if "$PG_CONFIG" --configure | grep -q -- '--enable-cassert'; then
+# grep -c, not grep -q; see test/lib.sh's pgc_is_columnar_scan.
+_cfg="$("$PG_CONFIG" --configure 2>/dev/null || true)"
+if [ "$(grep -c -- '--enable-cassert' <<<"$_cfg" || true)" != 0 ]; then
 	echo "REFUSING: $PG_CONFIG is an assert build; benchmark numbers from it are invalid" >&2
 	exit 1
 fi

bench/provision.sh

@@ -76,7 +76,9 @@ pg_present() { [ -x "$PREFIX_ROOT/$1/bin/pg_config" ]; }
 # built with cassert would produce benchmark numbers that are quietly wrong, and
 # nothing else on the box would notice.
 pg_is_assert() {
-	"$PREFIX_ROOT/$1/bin/pg_config" --configure 2>/dev/null | grep -q -- '--enable-cassert'
+	# grep -c, not grep -q; see test/lib.sh's pgc_is_columnar_scan.
+	_cfg="$("$PREFIX_ROOT/$1/bin/pg_config" --configure 2>/dev/null || true)"
+	[ "$(grep -c -- '--enable-cassert' <<<"$_cfg" || true)" != 0 ]
 }
 
 check_pg() {

test/arrow_import.sh

@@ -579,9 +579,12 @@ idx_count() {  # force an index scan
 	   SELECT count(*) FROM ix_tgt WHERE id BETWEEN 100 AND 199;" | tail -1
 }
 idx_plan_is_index_scan() {
-	q "$IDX_SETUP
-	   EXPLAIN (COSTS OFF) SELECT count(*) FROM ix_tgt WHERE id BETWEEN 100 AND 199;" \
-		| grep -qi 'Index.*Scan' && echo yes || echo no
+	# grep -c on a captured value, not a pipe into grep -q; see lib.sh's
+	# pgc_is_columnar_scan for the mechanism and the measurement.
+	local _plan
+	_plan="$(q "$IDX_SETUP
+	   EXPLAIN (COSTS OFF) SELECT count(*) FROM ix_tgt WHERE id BETWEEN 100 AND 199;")"
+	[ "$(grep -ci 'Index.*Scan' <<<"$_plan" || true)" != 0 ] && echo yes || echo no
 }
 seq_count() {  # force a sequential scan
 	q "SET enable_indexscan = off; SET enable_bitmapscan = off;

test/concurrency.sh

@@ -67,7 +67,11 @@ LOGFILE="$WORKDIR/server.log"
 # postmaster.
 port_is_free() {  # port -> 0 if nothing is listening on it
 	if command -v ss >/dev/null 2>&1; then
-		! ss -Htln "sport = :$1" 2>/dev/null | grep -q ":$1"
+		# grep -c on a captured value, not a pipe into grep -q. A spurious
+		# EPIPE here answers "nothing is listening" for a port that IS taken,
+		# and the suite then starts a cluster on an occupied port.
+		_pif="$(ss -Htln "sport = :$1" 2>/dev/null || true)"
+		[ "$(grep -c ":$1" <<<"$_pif" || true)" = 0 ]
 	else
 		# fall back to a connect probe: a refused connection means free
 		! (exec 3<>"/dev/tcp/127.0.0.1/$1") 2>/dev/null

test/entry_point_privilege.sh

@@ -307,9 +307,12 @@ as_ep() {  # as_ep <sql> -> the SQLSTATE, or the literal noerror
 	local out
 	out="$(env PATH="$PGC_BINDIR:$PATH" psql -h 127.0.0.1 -p "$PGC_PORT" -U t_ep -d "$PGC_DB" \
 		-At -v VERBOSITY=sqlstate -v ON_ERROR_STOP=0 -c "$1" 2>&1)"
-	printf '%s\n' "$out" | sed -n 's/^.*ERROR:[[:space:]]*\([0-9A-Z]\{5\}\).*$/\1/p' | head -1 \
-		| grep -q . && printf '%s\n' "$out" | sed -n 's/^.*ERROR:[[:space:]]*\([0-9A-Z]\{5\}\).*$/\1/p' | head -1 \
-		|| echo noerror
+	# Extract once into a variable rather than twice through a pipeline whose
+	# STATUS is the answer: `... | grep -q .` reports "no sqlstate" whenever the
+	# writer takes EPIPE, and this function's answer is a SQLSTATE.
+	local _sqlstate
+	_sqlstate="$(sed -n 's/^.*ERROR:[[:space:]]*\([0-9A-Z]\{5\}\).*$/\1/p' <<<"$out" | head -1)"
+	if [ -n "$_sqlstate" ]; then printf '%s\n' "$_sqlstate"; else echo noerror; fi
 }
 
 # The premise that makes every SQLSTATE arm below mean anything: this role can

test/lib.sh

@@ -1249,9 +1249,17 @@ check_ratio_needs_quiet_machine() {  # <name> <a> <b> <bound>
 # EXPLAIN without ANALYZE is enough: the line comes from the plan rather than the
 # run, so the assertion costs a plan and does not execute the query.
 pgc_is_columnar_scan() {	# query -> yes|no
-	env PATH="$PGC_BINDIR:$PATH" psql -h 127.0.0.1 -p "$PGC_PORT" -U postgres \
-		-d "$PGC_DB" -At -c "EXPLAIN (COSTS OFF) $1" 2>/dev/null \
-		| grep -q 'Columnar Projected Columns' && echo yes || echo no
+	# grep -c, NOT grep -q. grep -q exits the moment it matches, psql is still
+	# writing, and under a suite's `set -o pipefail` the pipeline reports failure
+	# though the pattern WAS present -- so this helper answers "no" for a plan that
+	# contains the line. Latent rather than live at EXPLAIN size (0 wrong in 200
+	# trials at 1.9KB) but with no floor in the mechanism: measured 1/200 wrong at
+	# 8.9KB and 40/40 at 289KB, on a loaded machine. grep -c reads to EOF.
+	local _plan
+	_plan="$(env PATH="$PGC_BINDIR:$PATH" psql -h 127.0.0.1 -p "$PGC_PORT" -U postgres \
+		-d "$PGC_DB" -At -c "EXPLAIN (COSTS OFF) $1" 2>/dev/null)"
+	[ "$(grep -c 'Columnar Projected Columns' <<<"$_plan" || true)" != 0 ] \
+		&& echo yes || echo no
 }
 
 # Does this query's plan drive the per-row fetch path?
@@ -1278,9 +1286,12 @@ pgc_is_columnar_scan() {	# query -> yes|no
 # report "no" for a query that does exercise the cache. Measured: an UPDATE over
 # 2,000 rows made 20,366 fetch_row calls under a Custom Scan plan (#797).
 pgc_uses_row_fetch() {	# setup query -> yes|no
-	env PATH="$PGC_BINDIR:$PATH" psql -h 127.0.0.1 -p "$PGC_PORT" -U postgres \
-		-d "$PGC_DB" -At -c "$1" -c "EXPLAIN (COSTS OFF) $2" 2>/dev/null \
-		| grep -q 'Index Scan using' && echo yes || echo no
+	# grep -c, not grep -q; see pgc_is_columnar_scan above for the measurement.
+	local _plan
+	_plan="$(env PATH="$PGC_BINDIR:$PATH" psql -h 127.0.0.1 -p "$PGC_PORT" -U postgres \
+		-d "$PGC_DB" -At -c "$1" -c "EXPLAIN (COSTS OFF) $2" 2>/dev/null)"
+	[ "$(grep -c 'Index Scan using' <<<"$_plan" || true)" != 0 ] \
+		&& echo yes || echo no
 }
 
 # Order-independent set hash of an arbitrary query's result. The row is cast to

test/native_groupagg.sh

@@ -45,9 +45,13 @@ groupvec_off() { psql_run "ALTER DATABASE $PGC_DB SET pgcolumnar.enable_group_ve
 # "Columnar Vectorized Group Keys"; no other node emits it, so a positive grep is
 # proof of the node rather than an absence test that a fallback would also pass.
 pgc_is_groupvec() {	# query -> yes|no
-	env PATH="$PGC_BINDIR:$PATH" psql -h 127.0.0.1 -p "$PGC_PORT" -U postgres \
-		-d "$PGC_DB" -At -c "EXPLAIN (COSTS OFF) $1" 2>/dev/null \
-		| grep -q 'Columnar Vectorized Group Keys' && echo yes || echo no
+	# grep -c on a captured plan, not a pipe into grep -q; see lib.sh's
+	# pgc_is_columnar_scan for the mechanism and the measurement.
+	local _plan
+	_plan="$(env PATH="$PGC_BINDIR:$PATH" psql -h 127.0.0.1 -p "$PGC_PORT" -U postgres \
+		-d "$PGC_DB" -At -c "EXPLAIN (COSTS OFF) $1" 2>/dev/null)"
+	[ "$(grep -c 'Columnar Vectorized Group Keys' <<<"$_plan" || true)" != 0 ] \
+		&& echo yes || echo no
 }
 
 # toggle_diff LABEL "QUERY on t_col": same query, path off vs on, byte-exact.

test/native_groupagg_batch.sh

@@ -59,7 +59,11 @@ q1() { q "$1" | tail -1; }
 # Keys" is emitted by no other node, so a positive grep proves the node rather
 # than an absence test a fallback would also satisfy.
 is_groupvec() {	# query -> yes|no
-	q "EXPLAIN (COSTS OFF) $1" | grep -q 'Columnar Vectorized Group Keys' \
+	# grep -c on a captured value, not a pipe into grep -q; see lib.sh's
+	# pgc_is_columnar_scan for the mechanism and the measurement.
+	local _plan
+	_plan="$(q "EXPLAIN (COSTS OFF) $1")"
+	[ "$(grep -c 'Columnar Vectorized Group Keys' <<<"$_plan" || true)" != 0 ] \
 		&& echo yes || echo no
 }
 

test/native_vecskip.sh

@@ -42,7 +42,12 @@ explain_of() {
 		-d "$PGC_DB" -At -c "EXPLAIN (ANALYZE, COSTS OFF, TIMING OFF, SUMMARY OFF) $1" 2>/dev/null
 }
 is_scalar_scan() {
-	explain_of "$1" | grep -q 'Columnar Projected Columns' && echo yes || echo no
+	# grep -c, not grep -q: see lib.sh's pgc_is_columnar_scan for why and for the
+	# measurement. This file holds the worst instance of the shape -- the arm at
+	# "premise: and it is not the scalar scan" WANTS "no", so a spurious EPIPE
+	# answer makes that premise pass for the wrong reason. Vacuity, not a red.
+	[ "$(explain_of "$1" | grep -c 'Columnar Projected Columns' || true)" != 0 ] \
+		&& echo yes || echo no
 }
 
 # The node, before any counter is read out of it.
@@ -99,13 +104,16 @@ explain_agg() {
 # this query falls back to the scalar scan -- which DOES print the line, so the
 # check below would pass while testing nothing at all.
 check "premise: the aggregate arm really is a vectorized aggregate" \
-	"$(explain_agg "$AGGQ" | grep -q 'Columnar Vectorized Aggregates' && echo yes || echo no)" \
+	"$([ "$(explain_agg "$AGGQ" | grep -c 'Columnar Vectorized Aggregates' || true)" != 0 ] \
+		&& echo yes || echo no)" \
 	"yes"
 check "premise: and it is not the scalar scan" \
-	"$(explain_agg "$AGGQ" | grep -q 'Columnar Projected Columns' && echo yes || echo no)" "no"
+	"$([ "$(explain_agg "$AGGQ" | grep -c 'Columnar Projected Columns' || true)" != 0 ] \
+		&& echo yes || echo no)" "no"
 
 check "the vectorized aggregate reports Columnar Vectors Skipped" \
-	"$(explain_agg "$AGGQ" | grep -q 'Columnar Vectors Skipped' && echo yes || echo no)" \
+	"$([ "$(explain_agg "$AGGQ" | grep -c 'Columnar Vectors Skipped' || true)" != 0 ] \
+		&& echo yes || echo no)" \
 	"yes"
 
 # Boundary and cross-vector ranges still return exactly the heap rows.

test/objstore_s3_read.sh

@@ -166,7 +166,9 @@ pg_restart_env "AWS_ENDPOINT_URL='https://127.0.0.1:$S3_PORT'" \
 	"AWS_ACCESS_KEY_ID='$AKID'" "AWS_SECRET_ACCESS_KEY='$SECRET'" \
 	"AWS_REGION='$REGION'"
 MOD_SO="$(pgc_pg "$PGC_BINDIR/pg_config --pkglibdir" | tail -1)/pgcolumnar_objstore.so"
-if pgc_pg "ldd '$MOD_SO'" 2>/dev/null | grep -q libssl; then
+# grep -c, not grep -q; see lib.sh's pgc_is_columnar_scan.
+_ldd_out="$(pgc_pg "ldd '$MOD_SO'" 2>/dev/null || true)"
+if [ "$(grep -c libssl <<<"$_ldd_out" || true)" != 0 ]; then
 	HTTPS_WANT="08006"
 else
 	HTTPS_WANT="0A000"

test/parquet_import.sh

@@ -167,9 +167,12 @@ idx_count() {  # force an index scan
 	   SELECT count(*) FROM ix_tgt WHERE id BETWEEN 100 AND 199;" | tail -1
 }
 idx_plan_is_index_scan() {
-	q "$IDX_SETUP
-	   EXPLAIN (COSTS OFF) SELECT count(*) FROM ix_tgt WHERE id BETWEEN 100 AND 199;" \
-		| grep -qi 'Index.*Scan' && echo yes || echo no
+	# grep -c on a captured value, not a pipe into grep -q; see lib.sh's
+	# pgc_is_columnar_scan for the mechanism and the measurement.
+	local _plan
+	_plan="$(q "$IDX_SETUP
+	   EXPLAIN (COSTS OFF) SELECT count(*) FROM ix_tgt WHERE id BETWEEN 100 AND 199;")"
+	[ "$(grep -ci 'Index.*Scan' <<<"$_plan" || true)" != 0 ] && echo yes || echo no
 }
 seq_count() {  # force a sequential scan
 	q "SET enable_indexscan = off; SET enable_bitmapscan = off;