Fix results ordering, control file whitespace, ENABLE_* overrides

.claude/agents/test.md

@@ -35,7 +35,7 @@ make clean-envs && test/bats/bin/bats tests/04-pgtle.bats
 ✅ **DO THIS**:
 ```bash
 test/bats/bin/bats tests/04-pgtle.bats  # Auto-rebuilds if needed
-DEBUG=5 test/bats/bin/bats tests/04-pgtle.bats  # For investigation
+TESTDEBUG=5 test/bats/bin/bats tests/04-pgtle.bats  # For investigation
 ```
 
 **ONLY clean when debugging the cleanup mechanism itself** and you MUST document what cleanup failure you're investigating.
@@ -228,11 +228,11 @@ test/bats/bin/bats tests/test-pgtle-install.bats
 
 ### Debugging
 ```bash
-DEBUG=2 test/bats/bin/bats tests/01-meta.bats  # Debug output
+TESTDEBUG=2 test/bats/bin/bats tests/01-meta.bats  # Debug output
 test/bats/bin/bats --verbose tests/01-meta.bats  # BATS verbose mode
 ```
 
-**Debug levels**: 10 (critical), 20 (significant), 30 (general), 40 (verbose), 50+ (maximum)
+**Debug levels** (set via `TESTDEBUG`): 1 (pollution/critical), 2 (major workflow), 3 (detailed state), 5 (verbose internals)
 
 ---
 
@@ -247,7 +247,8 @@ Tests set these automatically (from `tests/helpers.bash`):
 - `PGXNBRANCH` - Branch to use (defaults to `master`)
 - `TEST_TEMPLATE` - Template directory (defaults to `${TOPDIR}/template`)
 - `PG_LOCATION` - PostgreSQL installation path
-- `DEBUG` - Debug level (0-5)
+- `TESTDEBUG` - Test infrastructure debug level (0-5); controls helpers.bash debug() output
+- `DEBUG` - pgxntool script debug level; controls debug() in pgxntool/lib.sh (independent of TESTDEBUG)
 
 ---
 
@@ -292,7 +293,7 @@ test/bats/bin/bats tests/04-pgtle.bats  # Auto-rebuilds foundation via ensure_fo
 ### Debugging Test Failures
 
 1. Read test output (which assertion failed?)
-2. Use DEBUG mode: `DEBUG=5 test/bats/bin/bats tests/test-name.bats`
+2. Use TESTDEBUG mode: `TESTDEBUG=5 test/bats/bin/bats tests/test-name.bats`
 3. Inspect environment: `cd .envs/sequential/repo && ls -la`
 4. Check state markers: `ls .envs/sequential/.bats-state/`
 5. **Work top-down**: Fix earliest failure first (downstream failures often cascade)
@@ -361,7 +362,7 @@ make test
 test/bats/bin/bats tests/04-pgtle.bats
 
 # Debug
-DEBUG=5 test/bats/bin/bats tests/04-pgtle.bats
+TESTDEBUG=5 test/bats/bin/bats tests/04-pgtle.bats
 
 # Test infrastructure
 make test-recursion

.claude/settings.json

@@ -2,17 +2,17 @@
   "permissions": {
     "allow": [
       "Bash(make:*)",
-      "Bash(DEBUG=1 make:*)",
-      "Bash(DEBUG=2 make:*)",
-      "Bash(DEBUG=3 make:*)",
-      "Bash(DEBUG=4 make:*)",
-      "Bash(DEBUG=5 make:*)",
+      "Bash(TESTDEBUG=1 make:*)",
+      "Bash(TESTDEBUG=2 make:*)",
+      "Bash(TESTDEBUG=3 make:*)",
+      "Bash(TESTDEBUG=4 make:*)",
+      "Bash(TESTDEBUG=5 make:*)",
       "Bash(test/bats/bin/bats:*)",
-      "Bash(DEBUG=1 test/bats/bin/bats:*)",
-      "Bash(DEBUG=2 test/bats/bin/bats:*)",
-      "Bash(DEBUG=3 test/bats/bin/bats:*)",
-      "Bash(DEBUG=4 test/bats/bin/bats:*)",
-      "Bash(DEBUG=5 test/bats/bin/bats:*)",
+      "Bash(TESTDEBUG=1 test/bats/bin/bats:*)",
+      "Bash(TESTDEBUG=2 test/bats/bin/bats:*)",
+      "Bash(TESTDEBUG=3 test/bats/bin/bats:*)",
+      "Bash(TESTDEBUG=4 test/bats/bin/bats:*)",
+      "Bash(TESTDEBUG=5 test/bats/bin/bats:*)",
       "Edit",
       "WebFetch",
 

.claude/skills/commit/SKILL.md

@@ -41,6 +41,18 @@ bash .claude/skills/commit/scripts/gather-repo-info.sh ../pgxntool .
 
 Review output. Identify which repos have changes.
 
+### 2b. Check Debug Level Distribution (pgxntool .sh changes only)
+
+If `../pgxntool/*.sh` files are in the diff, scan each changed file:
+
+1. Collect all `debug N` call levels
+2. Note whether the file contains loops (`for`, `while`)
+3. Flag if: file has loops AND every debug call uses a single-digit level (1-9)
+4. Flag if: file has 5+ debug calls AND they all use the exact same level
+
+This is **advisory only** — present any findings in Step 5 and let the user decide.
+If nothing looks off, say nothing (don't add noise to the common case).
+
 ### 3. Check HISTORY.asc (pgxntool changes only)
 
 Read `../pgxntool/HISTORY.asc`. Determine if changes are significant and user-visible:

.claude/skills/test/scripts/run-tests.sh

@@ -8,6 +8,16 @@
 
 set -euo pipefail
 
+# Set PGPORT from PGCLUSTER if not already set, so pg_regress and other raw
+# PostgreSQL binaries use the correct cluster (Debian pg_wrapper handles
+# PGCLUSTER but raw binaries only respect PGPORT).
+if [ -z "${PGPORT:-}" ] && [ -n "${PGCLUSTER:-}" ]; then
+    _port=$(pg_lsclusters -h 2>/dev/null | awk -v c="$PGCLUSTER" \
+        'BEGIN{split(c,a,"/"); v=a[1]; n=a[2]} $1==v && $2==n {print $3; exit}')
+    [ -n "$_port" ] && export PGPORT=$_port
+    unset _port
+fi
+
 # Determine log directory (per-project, based on working directory)
 LOG_DIR="/tmp/pgxntool-test-logs$(pwd | sed 's/\//_/g')"
 mkdir -p "$LOG_DIR"

template-multi-extension/ext_beta.control

@@ -1,3 +1,3 @@
 comment = 'Beta extension for multi-extension testing'
-default_version = '2.5.0' # DO NOT REMOVE: trailing comment exercises parse_control_file comment handling (issue #25)
+default_version = '2.5.0'   # DO NOT REMOVE: multiple spaces before comment exercises parse_control_file whitespace trimming (issues #25, #26)
 requires = 'plpgsql'

template/test/expected/base.out

@@ -1,2 +1,3 @@
+\set ECHO none
 
 single_ext_ok

template/test/expected/pgxntool-test.out

@@ -1,2 +1,4 @@
+\set ECHO none
 1..1
 ok 1
+# TRANSACTION INTENTIONALLY LEFT OPEN!

template/test/sql/base.sql

@@ -1,3 +1,6 @@
+\set ECHO none
+-- DO NOT REMOVE: pg_regress runs psql with echo-all, so without this the SQL
+-- input is echoed into the output and never matches expected/base.out.
 \a
 \t
 SELECT pg_sleep(.1);

template/test/sql/pgxntool-test.sql

@@ -1,3 +1,8 @@
+\set ECHO none
+-- DO NOT REMOVE: pg_regress runs psql with echo-all, so without this every
+-- input line (including all of setup.sql) is echoed into the output and never
+-- matches expected/pgxntool-test.out. This mirrors the standard pgxntool pgtap
+-- convention (every test file starts with \set ECHO none).
 \i test/pgxntool/setup.sql
 
 SELECT plan(1);

test/lib/helpers.bash

@@ -70,13 +70,15 @@ error() {
 
 # Debug output function
 # Usage: debug LEVEL "message"
-# Outputs message if DEBUG >= LEVEL
+# Outputs message if TESTDEBUG >= LEVEL
+# Uses TESTDEBUG (not DEBUG) to avoid conflating test infrastructure verbosity
+# with pgxntool script verbosity — set them independently as needed.
 debug() {
   local level=$1
   shift
   local message="$*"
 
-  if [ "${DEBUG:-0}" -ge "$level" ]; then
+  if [ "${TESTDEBUG:-0}" -ge "$level" ]; then
     out -f "DEBUG[$level]: $message"
   fi
 }

test/sequential/04-pgtle.bats

@@ -228,6 +228,33 @@ teardown_file() {
   rm -f empty.control
 }
 
+@test "pgtle: parse_control_file handles whitespace before trailing comment" {
+  # Multiple spaces and a tab before a comment would leave stray whitespace after
+  # comment removal if only a single space was stripped (issue #26 follow-up to #25).
+  # Verify parse_control_file correctly strips all trailing whitespace so pgtle.sh
+  # finds the versioned SQL file and succeeds.
+
+  # Multiple spaces before comment
+  printf "default_version = '7.7.7'   # multiple spaces\n" > wspace_test.control
+  echo "SELECT 1;" > sql/wspace_test--7.7.7.sql
+
+  run "$TEST_REPO/pgxntool/pgtle.sh" --extension wspace_test --pgtle-version 1.5.0+
+  assert_success
+  assert_file_exists "pg_tle/1.5.0+/wspace_test.sql"
+
+  # Tab before comment
+  printf "default_version = '6.6.6'\t# tab before comment\n" > tab_test.control
+  echo "SELECT 1;" > sql/tab_test--6.6.6.sql
+
+  run "$TEST_REPO/pgxntool/pgtle.sh" --extension tab_test --pgtle-version 1.5.0+
+  assert_success
+  assert_file_exists "pg_tle/1.5.0+/tab_test.sql"
+
+  # Cleanup
+  rm -f wspace_test.control sql/wspace_test--7.7.7.sql pg_tle/1.5.0+/wspace_test.sql
+  rm -f tab_test.control sql/tab_test--6.6.6.sql pg_tle/1.5.0+/tab_test.sql
+}
+
 @test "pgtle: warning on module_pathname in control" {
   # Create a C extension control file
   echo "comment = 'C extension'" > cext.control

test/standard/make-results.bats

@@ -11,8 +11,6 @@
 # Not tested here:
 # - make results when already up-to-date (idempotent case). Safe to re-run,
 #   but the extra make test invocation adds non-trivial time for little value.
-# - verify-results behavior (blocking make results when tests fail) is tested
-#   separately in test-verify-results.bats.
 
 load ../lib/helpers
 
@@ -73,11 +71,21 @@ setup() {
   [ -d "test/output" ] || echo "$output" | grep -q "diff"
 }
 
+@test "make results is blocked by verify-results when tests fail" {
+  # The previous test ran make test which failed (mismatch), creating regression.diffs.
+  # With the fixed ordering (results: test verify-results), make results re-runs
+  # make test first (still fails → regression.diffs), then verify-results blocks.
+  # This test verifies that verify-results correctly intercepts failing results.
+  run make results
+  assert_failure
+  assert_contains "$output" "Cannot run 'make results'"
+}
+
 @test "make results updates expected output" {
-  # Run make results to fix the expected output.
-  # Must disable verify-results: the previous test created a mismatch, which caused
-  # make test to generate regression.diffs. verify-results blocks make results when
-  # regression.diffs exists, so we bypass it here to test the fix-mismatch workflow.
+  # Run make results with verify-results disabled to actually fix the mismatch.
+  # regression.diffs still exists from the previous test's make test run.
+  # Disabling verify-results lets make results run make test and copy the fresh
+  # results to expected/, fixing the mismatch.
   run make PGXNTOOL_ENABLE_VERIFY_RESULTS=no results
   assert_success
 }

test/standard/multi-extension.bats

@@ -16,9 +16,10 @@
 # - sql/ext_beta.sql
 # - META.in.json (with provides for both extensions)
 #
-# Note: ext_beta.control intentionally has a trailing comment on default_version.
-# This exercises pgtle.sh's parse_control_file to catch the quote-before-comment bug
-# (issue #25). Do not remove the comment from ext_beta.control.
+# Note: ext_beta.control intentionally has MULTIPLE spaces before the trailing comment
+# on default_version. This exercises parse_control_file's whitespace trimming to catch
+# both the single-space-only bug (issue #25) and the multiple-spaces bug (issue #26).
+# Do not reduce the spaces or remove the comment from ext_beta.control.
 
 load ../lib/helpers
 
@@ -112,8 +113,9 @@ setup() {
   run make all
   assert_success
 
-  # Also verify pgtle generation works — ext_beta.control has a trailing comment on
-  # default_version to exercise parse_control_file's comment handling (issue #25).
+  # Also verify pgtle generation works — ext_beta.control has multiple spaces before
+  # a trailing comment on default_version to exercise parse_control_file whitespace
+  # trimming (issues #25, #26).
   run make pgtle
   assert_success
 }

test/standard/test-verify-results.bats

@@ -5,10 +5,10 @@
 # Validates that verify-results correctly blocks make results when tests are
 # failing, detects pgtap failures, and can be disabled.
 #
-# Separate from make-results.bats because verify-results is pure file-checking
-# logic that doesn't require PostgreSQL. Keeping it in its own suite means these
-# tests always run, even when PostgreSQL is unavailable (make-results.bats skips
-# entirely in that case).
+# NOTE: pgxntool makes verify-results depend on test (verify-results: test), so
+# `make verify-results` re-runs the suite and therefore requires PostgreSQL. The
+# pgtap-detection cases below plant files in test/results/ (which survive the
+# fresh test run) rather than a regression.diffs (which the fresh run overwrites).
 
 load ../lib/helpers
 
@@ -29,38 +29,37 @@ setup() {
   assert_success
 }
 
-@test "verify-results has no dependencies on test execution" {
+@test "verify-results depends on test execution" {
+  # pgxntool makes verify-results depend on test (verify-results: test) so that
+  # `make results` re-runs the tests and checks the FRESH regression.diffs, even
+  # under make -j. Confirm the dependency is wired: a dry-run of verify-results
+  # must include the installcheck recipe pulled in via the test target.
   run make -n verify-results 2>&1
   assert_success
-  assert_not_contains "$output" "installcheck"
+  assert_contains "$output" "installcheck"
 }
 
-@test "regression.diffs blocks both verify-results and make results" {
-  cat > test/regression.diffs <<'EOF'
-*** test/expected/test.out
---- test/results/test.out
-***************
-*** 1 ****
-! expected
---- 1 ----
-! actual
-EOF
+@test "verify-results blocks when a test actually fails" {
+  # verify-results depends on test, so `make verify-results` re-runs the suite.
+  # A planted regression.diffs would just be overwritten by that fresh run, so
+  # force a REAL failure by corrupting an expected file, then confirm
+  # verify-results refuses to let make results proceed.
+  echo >> test/expected/base.out
 
   run make verify-results
   assert_failure
   assert_contains "$output" "Tests are failing"
   assert_contains "$output" "Cannot run 'make results'"
 
-  # make results itself should also be blocked
-  run make results
-  assert_failure
-  assert_contains "$output" "Cannot run 'make results'"
-
-  # State left for next test (regression.diffs still present)
+  # Restore clean state for the tests that follow.
+  run git checkout -- test/expected/base.out
+  assert_success
+  rm -f test/regression.diffs
 }
 
 @test "verify-results can be disabled" {
-  # regression.diffs still present from previous test
+  # With verify-results disabled, results depends only on test, so its dry-run
+  # never mentions the verify-results block message.
   run make -n results PGXNTOOL_ENABLE_VERIFY_RESULTS=no 2>&1
   assert_success
   # Check that verify-results recipe commands are not in the dry-run output.