Skip to content

test: the row-identity lock arms wait for the holder's UPDATE to finish; main is red on them (#1368) #1390

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jdatcmd merged 1 commit into from
Oct 14, 2026
Merged

test: the row-identity lock arms wait for the holder's UPDATE to finish; main is red on them (#1368) #1390

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 14 additions & 9 deletions test/advisory_lock_class.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -212,20 +212,25 @@ check "a duplicate key is still rejected" \
# so it is fixed at server start. Neither was set or read by any test.
psql_run "CREATE TABLE rl (id int, v int) USING pgcolumnar;"
psql_run "INSERT INTO rl SELECT g, g FROM generate_series(1, 100) g;"
# row_locks_while_updating SETTING -> "<class-104 locks>|<holder's UPDATE running>", read
# in one query once the holder's UPDATE is running. WAITED FOR, NOT SLEPT ON (#1368
# review): read after a fixed sleep, a holder that had not reached its UPDATE yet made
# the off arm read 0 on a build that ignores the setting, because nothing was locked.
# The second field proves the UPDATE was running when the first was read.
# row_locks_while_updating SETTING -> "<class-104 locks>|<holder past its UPDATE>", read
# in one query once the holder's UPDATE has FINISHED. WAITED FOR, NOT SLEPT ON (#1368
# review), and waited for on the right event: the holder names itself in
# application_name after its UPDATE returns, and the poll waits for that name. Waiting
# for its RowExclusiveLock on rl was not enough, because the UPDATE takes that at
# executor start, before it reaches row 7 and takes the class-104 lock; a poll landing
# between the two read "none|1" on CI (main f601281, PG 17 and PG 18). The CTE sleep
# ahead of the row keeps that window open on every run, so waiting on the early event
# fails every time rather than now and then.
row_locks_while_updating() {
local tag="rl_updated_$1"
env PATH="$PGC_BINDIR:$PATH" psql -h 127.0.0.1 -p "$PGC_PORT" -U postgres -d "$PGC_DB" -Atq \
-c "BEGIN; SET LOCAL pgcolumnar.enable_row_update_lock = $1; UPDATE rl SET v = v + 1 WHERE id = 7; SELECT pg_sleep(3); COMMIT;" \
-c "BEGIN; SET LOCAL pgcolumnar.enable_row_update_lock = $1; WITH d AS MATERIALIZED (SELECT pg_sleep(1)) UPDATE rl SET v = v + 1 FROM d WHERE id = 7; SET application_name = '$tag'; SELECT pg_sleep(3); COMMIT;" \
>/dev/null 2>&1 &
local holder=$! got="" i
for i in $(seq 1 100); do
got="$(q "SELECT count(*) FILTER (WHERE locktype = 'advisory' AND objsubid = 104) || '|' ||
count(*) FILTER (WHERE locktype = 'relation' AND relation = 'rl'::regclass AND mode = 'RowExclusiveLock')
FROM pg_locks WHERE pid <> pg_backend_pid()")"
got="$(q "SELECT (SELECT count(*) FROM pg_locks WHERE locktype = 'advisory' AND objsubid = 104
AND pid <> pg_backend_pid()) || '|' ||
(SELECT count(*) FROM pg_stat_activity WHERE application_name = '$tag')")"
[ "${got#*|}" = 1 ] && break
sleep 0.1
done
Expand Down
Loading