Re: Improving tracking/processing of buildfarm test failures

Hello hackers,

Please take a look at the August report on buildfarm failures:
# SELECT br, COUNT(*) FROM failures WHERE dt >= '2026-08-01' AND
 dt < '2026-09-01' GROUP BY br;
REL_14_STABLE: 14
REL_15_STABLE: 7
REL_16_STABLE: 16
REL_17_STABLE: 13
REL_18_STABLE: 27
REL_19_STABLE: 62
master: 68
-- Total: 207
(Counting test failures only, excluding indent-check, Configure, Build
errors, also excluding failures from fruitcrow.)

# SELECT COUNT(*) FROM (SELECT DISTINCT issue_link FROM failures WHERE
 dt >= '2026-08-01' AND dt < '2026-09-01');
21

# SELECT issue_link, COUNT(*) FROM failures WHERE dt >= '2026-08-01' AND
 dt < '2026-09-01' GROUP BY issue_link ORDER BY 2 DESC LIMIT 5;
https://www.postgresql.org/message-id/f071e691-5930-4738-9dbd-43ed2da367fe%40gmail.com : 41
-- plpython tests fail against python 3.14 under Valgrind
-- Fixed

https://www.postgresql.org/message-id/1383452.1787067554%40sss.pgh.pa.us : 25
-- test_shm_mq.sql fails due to error on registering background process
-- Fixed

https://www.postgresql.org/message-id/d92b18e8-00aa-4c32-95ff-7d807a427481%40vondra.me : 20
-- occasional ECPG failures on dikkop (FreeBSD) due to non-thread-safe snprintf

https://www.postgresql.org/message-id/7a68bee2-91bd-481d-be44-160e6beebc83%40gmail.com : 18
-- xid_wraparound/002_limits.pl might fail due to transient warning

https://www.postgresql.org/message-id/361531e2-52b5-499c-a126-815f277bbef2%40gmail.com : 15
-- test_checksums/001_basic.pl fails on dropping database on slow storage (after 51f55b13a)

# SELECT COUNT(*) FROM failures WHERE dt >= '2026-08-01' AND
 dt < '2026-09-01' AND issue_link IS NULL; -- Unsorted/unhelpful failures
45

Short-lived failures: 36

Best regards,
Alexander