Move code from ExecutorCheckPerms into ExecutorStart

[jeltz]

The data we get in the ExecutorCheckPerms hook is also available in the QueryDesc object we get in the other executor hooks so doing the relation name lookup in a separate hook only makes the code more complicated for no gain.

It might make even more sense to move this lookup to the ExecutorEnd hook but that risks changing the behavior as we cannot do catalog lookups when the transaction has been rolled back. It is unclear if we need to support that, but that kind of change should be its own commit.

PG-0

[codecov]

Codecov Report

❌ Patch coverage is 95.83333% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 88.04%. Comparing base (16107e3) to head (53b3df0).

Files with missing lines	Patch %	Lines
src/pg_stat_monitor.c	95.83%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #789      +/-   ##
==========================================
+ Coverage   87.93%   88.04%   +0.10%     
==========================================
  Files           3        3              
  Lines        1318     1313       -5     
  Branches      186      185       -1     
==========================================
- Hits         1159     1156       -3     
+ Misses         77       76       -1     
+ Partials       82       81       -1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[dutow]

This introduced a bug (or two, depending on how count it): the utility path doesn't update relations and num_relations

-- Observed results for the trailing COPY:
--   base (main)     : {public.logtbl}          (correct — COPY's own table)
--   patched (pr789) : {public.sensitive_audit} (WRONG — from the prior errored query)
CREATE EXTENSION IF NOT EXISTS pg_stat_monitor;
CREATE TABLE sensitive_audit(id int);
INSERT INTO sensitive_audit VALUES (1);
CREATE TABLE logtbl(id int);
SELECT pg_stat_monitor_reset();

-- Tracked query touching sensitive_audit, errors during ExecutorRun:
SELECT id, 1/(id-1) FROM sensitive_audit;   -- division by zero
-- Unrelated utility statement stored via the ProcessUtility path:
COPY logtbl TO '/tmp/pgsm_stale.csv' CSV;

SELECT substring(query for 45) AS query, relations
FROM pg_stat_monitor
WHERE query ILIKE '%COPY logtbl%' OR query ILIKE '%sensitive%'
ORDER BY query;

and

-- Observed results:
--                                      base (main)                 patched (pr789)
--   SELECT ... FROM parent             {public.parent}             {public.parent}   (unchanged)
--   INSERT INTO child (has FK)         {}  (clobbered/empty)       {public.child}    (FIXED)
--   UPDATE child (has FK)              {}  (clobbered/empty)       {public.child}    (FIXED)
--   COPY parent TO ...                 {public.parent}             {}                (REGRESSION)

CREATE EXTENSION IF NOT EXISTS pg_stat_monitor;
CREATE TABLE parent(id int primary key);
CREATE TABLE child(id int primary key, pid int references parent(id));
INSERT INTO parent VALUES (1),(2),(3);
SELECT pg_stat_monitor_reset();

SELECT * FROM parent WHERE id = 1;
INSERT INTO child VALUES (10, 1);
COPY parent TO '/tmp/pgsm_copy_test.csv' CSV;
UPDATE child SET pid = 2 WHERE id = 10;

SELECT substring(query for 42) AS query, relations
FROM pg_stat_monitor
WHERE query NOT ILIKE '%pg_stat_monitor%' AND query NOT ILIKE '%CREATE%'
ORDER BY query;

[dutow]

typo