Update to lib_pb_query 6.2.2

.gitignore

@@ -27,3 +27,12 @@ pg_query-*.tar
 
 # compiled nif
 priv/*.so
+
+# Fuzz build outputs (compiled binaries)
+fuzz/_build/
+
+# Generated deparse corpus - rebuild with: make fuzz-corpus
+fuzz/corpus/deparse/
+
+# libFuzzer-discovered corpus entries (hash-named files added during fuzzing runs)
+fuzz/corpus/parse/[0-9a-f][0-9a-f][0-9a-f][0-9a-f]*

Makefile

@@ -11,19 +11,75 @@ ifeq ($(shell uname -s),Darwin)
     LDFLAGS += -undefined dynamic_lookup
 endif
 
-.PHONY: all libpg_query_ex clean update-libpg_query
+FUZZ_CC      = clang
+FUZZ_CFLAGS  = -g -O1 -fsanitize=fuzzer,address -I$(LIBPG_QUERY_PATH)
+FUZZ_LDFLAGS = -fsanitize=fuzzer,address -lpthread
+FUZZ_SRCS    = fuzz/fuzz_parse_protobuf.c \
+               fuzz/fuzz_scan.c \
+               fuzz/fuzz_deparse.c \
+               fuzz/fuzz_roundtrip.c
+FUZZ_BINS    = $(patsubst fuzz/%.c,fuzz/_build/%,$(FUZZ_SRCS))
+
+.PHONY: all libpg_query_ex clean fuzz fuzz-corpus fuzz-clean patch update-libpg_query
 
 all: priv/libpg_query_ex.so
 
-priv:
-	mkdir -p priv
+
+fuzz/_build:
+	mkdir -p fuzz/_build
 
 $(LIBPG_QUERY_PATH)/libpg_query.a:
 	$(MAKE) -B -C $(LIBPG_QUERY_PATH) libpg_query.a
 
-priv/libpg_query_ex.so: priv $(LIBPG_QUERY_PATH)/libpg_query.a c_src/libpg_query_ex.c
+priv/libpg_query_ex.so: $(LIBPG_QUERY_PATH)/libpg_query.a c_src/libpg_query_ex.c
+	mkdir -p priv
 	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ c_src/libpg_query_ex.c $(LIBPG_QUERY_PATH)/libpg_query.a
 
+# Build all fuzz targets (requires clang with libFuzzer)
+fuzz: fuzz/_build $(LIBPG_QUERY_PATH)/libpg_query.a $(FUZZ_BINS)
+	@echo ""
+	@echo "Fuzz targets built in fuzz/_build/. Next steps:"
+	@echo "  1. Populate the deparse corpus:  make fuzz-corpus"
+	@echo "  2. Run a fuzzer, e.g.:"
+	@echo "       fuzz/_build/fuzz_parse_protobuf -max_len=65536 fuzz/corpus/parse/"
+	@echo "       fuzz/_build/fuzz_deparse        -max_len=65536 fuzz/corpus/deparse/"
+
+fuzz/_build/%: fuzz/%.c $(LIBPG_QUERY_PATH)/libpg_query.a
+	$(FUZZ_CC) $(FUZZ_CFLAGS) -o $@ $< $(LIBPG_QUERY_PATH)/libpg_query.a $(FUZZ_LDFLAGS)
+
+# Also build the corpus generator (not a fuzz target, no -fsanitize=fuzzer)
+fuzz/_build/gen_deparse_corpus: fuzz/gen_deparse_corpus.c $(LIBPG_QUERY_PATH)/libpg_query.a
+	$(FUZZ_CC) -g -O1 -fsanitize=address -I$(LIBPG_QUERY_PATH) -o $@ $< \
+		$(LIBPG_QUERY_PATH)/libpg_query.a -fsanitize=address -lpthread
+
+# Generate binary protobuf seeds for fuzz_deparse from the SQL corpus
+fuzz-corpus: fuzz/_build/gen_deparse_corpus
+	@mkdir -p fuzz/corpus/deparse
+	fuzz/_build/gen_deparse_corpus fuzz/corpus/deparse \
+		"SELECT 1" \
+		"SELECT id, name FROM users WHERE active = true" \
+		"SELECT u.id, o.total FROM users u JOIN orders o ON u.id = o.user_id" \
+		"INSERT INTO users (name, email) VALUES ('Alice', 'alice@example.com')" \
+		"UPDATE users SET active = false WHERE last_login < NOW() - INTERVAL '90 days'" \
+		"DELETE FROM sessions WHERE expires_at < NOW()" \
+		"WITH recent AS (SELECT * FROM events WHERE created_at > NOW() - INTERVAL '1 hour') SELECT count(*) FROM recent" \
+		"SELECT * FROM users WHERE id IN (SELECT user_id FROM orders WHERE total > 100)" \
+		"SELECT department, count(*), avg(salary) FROM employees GROUP BY department HAVING count(*) > 5" \
+		"SELECT id, salary, rank() OVER (PARTITION BY department ORDER BY salary DESC) FROM employees" \
+		"SELECT id::text, created_at::date, price::numeric(10,2) FROM products" \
+		"SELECT CASE WHEN status = 1 THEN 'active' WHEN status = 2 THEN 'pending' ELSE 'unknown' END FROM orders" \
+		"SELECT ARRAY[1,2,3], unnest(tags) FROM posts" \
+		"SELECT data->>'name', data->'address'->>'city' FROM profiles WHERE data @> '{\"active\":true}'" \
+		"SELECT id FROM users WHERE EXISTS (SELECT 1 FROM orders WHERE orders.user_id = users.id)" \
+		"SELECT * FROM events WHERE created_at BETWEEN '2024-01-01' AND '2024-12-31'" \
+		"SELECT * FROM users WHERE email ILIKE '%@example.com' AND name NOT LIKE 'test%'" \
+		"SELECT * FROM users WHERE deleted_at IS NULL AND parent_id IS NOT NULL" \
+		"SELECT * FROM users WHERE id = \$$1 AND status = \$$2" \
+		"SELECT id FROM admins UNION ALL SELECT id FROM moderators EXCEPT SELECT id FROM banned_users"
+
+fuzz-clean:
+	$(RM) -r fuzz/_build fuzz/corpus/deparse fuzz/crashes
+
 protobuf:
 	MIX_ENV=prod mix protox.generate --output-path=lib/pg_query/proto --multiple-files c_src/libpg_query/protobuf/pg_query.proto
 
@@ -32,6 +88,15 @@ clean:
 	$(MAKE) -C $(LIBPG_QUERY_PATH) clean
 	$(RM) priv/libpg_query_ex.so
 
+# Apply all local patches to the vendored libpg_query source
+patch:
+	@for p in patches/*.patch; do \
+		echo "Applying $$p ..."; \
+		git apply "$$p" || { echo "FAILED: $$p"; exit 1; }; \
+	done
+	@echo "All patches applied."
+
 update-libpg_query:
-	git subtree pull -P "c_src/libpg_query" --squash https://github.com/pganalyze/libpg_query.git 15-latest
+	git subtree pull -P "c_src/libpg_query" --squash https://github.com/pganalyze/libpg_query.git 17-latest
+	$(MAKE) patch
 

README.md

@@ -35,6 +35,46 @@ def deps do
 end
 ```
 
+## Fuzzing
+
+The `fuzz/` directory contains [libFuzzer](https://llvm.org/docs/LibFuzzer.html) harnesses that exercise the NIF boundary. You need clang with libFuzzer support (standard in most LLVM distributions).
+
+Build all four targets:
+
+```bash
+make fuzz
+```
+
+Optionally seed the deparse corpus from known-good SQL strings:
+
+```bash
+make fuzz-corpus
+```
+
+Then run a target, for example:
+
+```bash
+# Fuzz the parse → deparse round-trip (highest-value target)
+fuzz/_build/fuzz_roundtrip -max_len=4096 -artifact_prefix=fuzz/crashes/ fuzz/corpus/parse/
+
+# Fuzz raw protobuf bytes fed directly into the deparser
+fuzz/_build/fuzz_deparse -max_len=65536 -artifact_prefix=fuzz/crashes/ fuzz/corpus/deparse/
+
+# Fuzz the SQL parser
+fuzz/_build/fuzz_parse_protobuf -max_len=4096 -artifact_prefix=fuzz/crashes/ fuzz/corpus/parse/
+
+# Fuzz the SQL scanner
+fuzz/_build/fuzz_scan -max_len=4096 -artifact_prefix=fuzz/crashes/ fuzz/corpus/parse/
+```
+
+Crash artifacts are written to `fuzz/crashes/` (the `-artifact_prefix` flag controls this; without it libFuzzer writes to the current directory). Replay a crash by passing the file as a positional argument:
+
+```bash
+fuzz/_build/fuzz_roundtrip fuzz/crashes/<file>
+```
+
+Clean up build artefacts and generated corpus with `make fuzz-clean`.
+
 ## License
 
 This Elixir interface is distributed under the terms of the [Apache 2.0 license](./LICENSE).

c_src/libpg_query/.gitattributes

@@ -1 +1,2 @@
-*.sql binary
+*.sql binary
+*.psql binary

c_src/libpg_query/.github/workflows/ci.yml

@@ -8,10 +8,15 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-20.04]
+        os: [ubuntu-24.04]
         compiler: [clang, gcc]
         protobuf_lib: [protobuf-c, protobuf-cpp]
         valgrind: [valgrind,no-valgrind]
+        exclude:
+          # this combination hits linking errors: see https://github.com/pganalyze/libpg_query/pull/289
+          - compiler: clang
+            protobuf_lib: protobuf-cpp
+            valgrind: valgrind
     steps:
     - name: Check out code
       uses: actions/checkout@v4
@@ -21,7 +26,7 @@ jobs:
       uses: actions/cache@v4
       with:
         path: protobuf-25.1
-        key: ${{ runner.os }}-protobuf-25.1
+        key: ${{ runner.os }}-${{ matrix.compiler }}-protobuf-25.1
     - name: Build protobuf library
       if: matrix.protobuf_lib == 'protobuf-cpp' && steps.cache-protobuf.outputs.cache-hit != 'true'
       run: |

c_src/libpg_query/CHANGELOG.md

@@ -2,6 +2,50 @@
 
 All versions are tagged by the major Postgres version, plus an individual semver for this library itself.
 
+## 17-6.2.2   2026-01-26
+
+* pg_query_normalize: Fix handling of special strings in DefElem [#325](https://github.com/pganalyze/libpg_query/pull/325)
+  - This avoids a crash when running the normalize function on certain utility statements
+* pg_query_deparse_comments_for_query: Add missing initialization [#324](https://github.com/pganalyze/libpg_query/pull/324)
+  - This avoids a crash for callers that read the error field of the result when there is no error
+
+## 17-6.2.1   2026-01-14
+
+* Add pg_query_is_utility_stmt function to determine if query text contains utility statements [#313](https://github.com/pganalyze/libpg_query/pull/313)
+  * This is a fast check for callers that don't actually need the parse tree itself
+* Add missing top-level postgres_deparse.h in Makefile install step
+  - This was an oversight from the previous 6.2.0 release
+* Improve pg_query_summary function:
+  - Speed up summary truncate replacement logic
+  - Correctly handle `GRANT .. ON ALL TABLES IN SCHEMA` statements
+  - Correctly handle schema qualified filter columns
+
+## 17-6.2.0   2025-12-10
+
+* Add fast summary information function (pg_query_summary)
+  - This allows gathering certain information, for example which tables are referenced in a
+    statement, without requiring a Protobuf serialization step in a higher level library
+  - Additionally this can also be used to perform "smart truncation" of a query by
+    omitting deeply nested information (e.g. a CTE definition, or a target list) whilst
+    preserving more essential parts like the FROM claus
+* Deparser:
+  - Introduce pretty printing / formatting
+    - Introduces a new optional pretty print mode that emits a human readable
+      output. A detailed explanation of the mechanism can be found at the start
+      of the deparser file.
+  - Rework handling of expressions inside typecasts
+    - Prefer (..)::type syntax, unless we are already in a function call.
+  - Use lowercase keywords in xmlroot functions
+    - This matches other XML functions as well as the Postgres documentation,
+      since these are closer to function argument names than regular keywords.
+  - Fix deparse of ALTER TABLE a ALTER b SET STATISTICS DEFAULT
+  - Fix the null pointer dereference when handling identity columns
+* Allow alternate definitions of NAMEDATALEN identifier limit
+  - This allows building libpg_query with an override of the built-time limit of
+    Postgres identifiers (typically 63 characters)
+* Normalization: Add support for CALL statements
+* Bump Postgres to 17.7 and switch back to release tarballs
+
 ## 17-6.1.0   2025-04-02
 
 * Update to Postgres 17.4, and add recent patches scheduled for Postgres 17.5 (not yet released)

c_src/libpg_query/Makefile

@@ -7,13 +7,12 @@ PGDIR = $(root_dir)/tmp/postgres
 PGDIRBZ2 = $(root_dir)/tmp/postgres.tar.bz2
 PGDIRZIP = $(root_dir)/tmp/postgres.zip
 
-PG_VERSION = 17.4
+PG_VERSION = 17.7
 PG_VERSION_MAJOR = $(call word-dot,$(PG_VERSION),1)
-PG_VERSION_NUM = 170004
-PG_BRANCH = REL_17_STABLE
+PG_VERSION_NUM = 170007
 PROTOC_VERSION = 25.1
 
-VERSION = 6.1.0
+VERSION = 6.2.2
 VERSION_MAJOR = $(call word-dot,$(VERSION),1)
 VERSION_MINOR = $(call word-dot,$(VERSION),2)
 VERSION_PATCH = $(call word-dot,$(VERSION),3)
@@ -39,6 +38,7 @@ override CFLAGS += -g -I. -I./vendor -I./src/include -I./src/postgres/include -W
 
 ifeq ($(OS),Windows_NT)
 override CFLAGS += -I./src/postgres/include/port/win32
+override TEST_CFLAGS += -I./src/postgres/include/port/win32
 endif
 
 override PG_CONFIGURE_FLAGS += -q --without-readline --without-zlib --without-icu
@@ -116,14 +116,9 @@ clean:
 .PHONY: all clean build build_shared extract_source examples test install
 
 $(PGDIR):
-#	We temporarily build off REL_17_STABLE to pull in https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=6da2ba1d8a031984eb016fed6741bb2ac945f19d
-#   TODO: Go back to upstream tarball once 17.5 is released
-#	tar -xjf $(PGDIRBZ2)
-#	curl -o $(PGDIRBZ2) https://ftp.postgresql.org/pub/source/v$(PG_VERSION)/postgresql-$(PG_VERSION).tar.bz2
-#	mv $(root_dir)/postgresql-$(PG_VERSION) $(PGDIR)
-	curl -L -o $(PGDIRZIP) https://github.com/postgres/postgres/archive/refs/heads/$(PG_BRANCH).zip
-	unzip $(PGDIRZIP)
-	mv $(root_dir)/postgres-$(PG_BRANCH) $(PGDIR)
+	curl -o $(PGDIRBZ2) https://ftp.postgresql.org/pub/source/v$(PG_VERSION)/postgresql-$(PG_VERSION).tar.bz2
+	tar -xjf $(PGDIRBZ2)
+	mv $(root_dir)/postgresql-$(PG_VERSION) $(PGDIR)
 	cd $(PGDIR); patch -p1 < $(root_dir)/patches/01_parser_additional_param_ref_support.patch
 	cd $(PGDIR); patch -p1 < $(root_dir)/patches/03_lexer_track_yyllocend.patch
 	cd $(PGDIR); patch -p1 < $(root_dir)/patches/04_lexer_comments_as_tokens.patch
@@ -133,6 +128,7 @@ $(PGDIR):
 	cd $(PGDIR); patch -p1 < $(root_dir)/patches/08_avoid_zero_length_delimiter_in_regression_tests.patch
 	cd $(PGDIR); patch -p1 < $(root_dir)/patches/09_allow_param_junk.patch
 	cd $(PGDIR); patch -p1 < $(root_dir)/patches/10_avoid_namespace_hashtab_impl_gen.patch
+	cd $(PGDIR); patch -p1 < $(root_dir)/patches/11_ifndef_namedatalen.patch
 	cd $(PGDIR); ./configure $(PG_CONFIGURE_FLAGS)
 	cd $(PGDIR); make -C src/pl/plpgsql/src pl_gram.h plerrcodes.h pl_reserved_kwlist_d.h pl_unreserved_kwlist_d.h
 	cd $(PGDIR); make -C src/port pg_config_paths.h
@@ -241,14 +237,15 @@ examples/normalize_error: examples/normalize_error.c $(ARLIB)
 examples/simple_plpgsql: examples/simple_plpgsql.c $(ARLIB)
 	$(CC) $(TEST_CFLAGS) -o $@ -g examples/simple_plpgsql.c $(ARLIB) $(TEST_LDFLAGS)
 
-TESTS = test/complex test/concurrency test/deparse test/fingerprint test/fingerprint_opts test/normalize test/normalize_utility test/parse test/parse_opts test/parse_protobuf test/parse_protobuf_opts test/parse_plpgsql test/scan test/split
+TESTS = test/complex test/concurrency test/deparse test/fingerprint test/fingerprint_opts test/is_utility_stmt test/normalize test/normalize_utility test/parse test/parse_opts test/parse_protobuf test/parse_protobuf_opts test/parse_plpgsql test/scan test/split test/summary test/summary_truncate
 test: $(TESTS)
 ifeq ($(VALGRIND),1)
 	$(VALGRIND_MEMCHECK) test/complex || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/concurrency || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/deparse || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/fingerprint || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/fingerprint_opts || (cat test/valgrind.log && false)
+	$(VALGRIND_MEMCHECK) test/is_utility_stmt || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/normalize || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/normalize_utility || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/parse || (cat test/valgrind.log && false)
@@ -257,6 +254,8 @@ ifeq ($(VALGRIND),1)
 	$(VALGRIND_MEMCHECK) test/parse_protobuf_opts || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/scan || (cat test/valgrind.log && false)
 	$(VALGRIND_MEMCHECK) test/split || (cat test/valgrind.log && false)
+	$(VALGRIND_MEMCHECK) test/summary || (cat test/valgrind.log && false)
+	$(VALGRIND_MEMCHECK) test/summary_truncate || (cat test/valgrind.log && false)
 	# Output-based tests
 	$(VALGRIND_MEMCHECK) test/parse_plpgsql || (cat test/valgrind.log && false)
 	diff -Naur test/plpgsql_samples.expected.json test/plpgsql_samples.actual.json
@@ -266,6 +265,7 @@ else
 	test/deparse
 	test/fingerprint
 	test/fingerprint_opts
+	test/is_utility_stmt
 	test/normalize
 	test/normalize_utility
 	test/parse
@@ -274,6 +274,8 @@ else
 	test/parse_protobuf_opts
 	test/scan
 	test/split
+	test/summary
+	test/summary_truncate
 	# Output-based tests
 	test/parse_plpgsql
 	diff -Naur test/plpgsql_samples.expected.json test/plpgsql_samples.actual.json
@@ -297,6 +299,10 @@ test/fingerprint_opts: test/fingerprint_opts.c test/fingerprint_opts_tests.c $(A
 	# We have "-Isrc/" because this test uses pg_query_fingerprint_with_opts
 	$(CC) $(TEST_CFLAGS) -o $@ -Isrc/ test/fingerprint_opts.c $(ARLIB) $(TEST_LDFLAGS)
 
+test/is_utility_stmt: test/framework/main.c test/is_utility_stmt.c $(ARLIB)
+	# We have "-Isrc/postgres/include" because this test uses pg_query_summary_direct
+	$(CC) $(TEST_CFLAGS) -o $@ -Isrc/postgres/include test/framework/main.c test/is_utility_stmt.c $(ARLIB) $(TEST_LDFLAGS)
+
 test/normalize: test/normalize.c test/normalize_tests.c $(ARLIB)
 	$(CC) $(TEST_CFLAGS) -o $@ test/normalize.c $(ARLIB) $(TEST_LDFLAGS)
 
@@ -306,6 +312,14 @@ test/normalize_utility: test/normalize_utility.c test/normalize_utility_tests.c
 test/parse: test/parse.c test/parse_tests.c $(ARLIB)
 	$(CC) $(TEST_CFLAGS) -o $@ test/parse.c $(ARLIB) $(TEST_LDFLAGS)
 
+test/summary: test/framework/main.c test/summary.c test/summary_tests.c test/summary_tests_list.c $(ARLIB)
+	# We have "-Isrc/postgres/include" because this test uses pg_query_summary_direct
+	$(CC) $(TEST_CFLAGS) -o $@ -Isrc/postgres/include test/framework/main.c test/summary.c $(ARLIB) $(TEST_LDFLAGS)
+
+test/summary_truncate: test/framework/main.c test/summary_truncate.c $(ARLIB)
+	# We have "-Isrc/postgres/include" because this test uses pg_query_summary_direct
+	$(CC) $(TEST_CFLAGS) -o $@ -Isrc/postgres/include test/framework/main.c test/summary_truncate.c $(ARLIB) $(TEST_LDFLAGS)
+
 test/parse_opts: test/parse_opts.c test/parse_opts_tests.c $(ARLIB)
 	$(CC) $(TEST_CFLAGS) -o $@ test/parse_opts.c $(ARLIB) $(TEST_LDFLAGS)
 
@@ -336,4 +350,5 @@ install: $(ARLIB) $(SOLIB)
 	$(LN_S) $(SOLIBVER) "$(DESTDIR)"$(libdir)/$(SOLIB)
 	$(INSTALL) -d "$(DESTDIR)"$(includedir)/$(TARGET)
 	$(INSTALL) -m 644 pg_query.h "$(DESTDIR)"$(includedir)/pg_query.h
+	$(INSTALL) -m 644 postgres_deparse.h "$(DESTDIR)"$(includedir)/postgres_deparse.h
 	$(INSTALL) -m 644 protobuf/pg_query.proto "$(DESTDIR)"$(includedir)/$(TARGET)/pg_query.proto