asciiman: support attribute passing

Also add a VIM syntax highlighting file.

This also fixes some previously untriggered bugs.

.clang-format

@@ -0,0 +1,32 @@
+# clang-format is fairly limited, and these rules are approximate:
+#  - array initializers can get terribly mangled with clang-format 12.0,
+#  - sometimes it still aligns with space characters,
+#  - struct name NL { NL ... NL } NL name; is unachievable.
+BasedOnStyle: GNU
+ColumnLimit: 80
+IndentWidth: 4
+TabWidth: 4
+UseTab: ForContinuationAndIndentation
+BreakBeforeBraces: Allman
+SpaceAfterCStyleCast: true
+AlignAfterOpenBracket: DontAlign
+AlignOperands: DontAlign
+AlignConsecutiveMacros: Consecutive
+AllowAllArgumentsOnNextLine: false
+AllowAllParametersOfDeclarationOnNextLine: false
+IndentGotoLabels: false
+
+# IncludeCategories has some potential, but it may also break the build.
+# Note that the documentation says the value should be "Never".
+SortIncludes: false
+
+# This is a compromise, it generally works out aesthetically better.
+BinPackArguments: false
+
+# Unfortunately, this can't be told to align to column 40 or so.
+SpacesBeforeTrailingComments: 2
+
+# liberty-specific macro body wrappers.
+MacroBlockBegin: "BLOCK_START"
+MacroBlockEnd: "BLOCK_END"
+ForEachMacros: ["LIST_FOR_EACH"]

.gitignore

@@ -7,3 +7,5 @@
 /liberty.files
 /liberty.creator*
 /liberty.includes
+/liberty.cflags
+/liberty.cxxflags

.travis.yml

@@ -1,18 +0,0 @@
-language: c
-notifications:
- irc:
-  channels: "irc.janouch.name#dev"
-  use_notice: true
-  skip_join: true
-compiler:
- - clang
- - gcc
-before_install:
- - sudo apt-get update -qq
-before_script:
- - mkdir build
- - cd build
-script:
- - cmake .. -DCMAKE_INSTALL_PREFIX=/usr
- - make
- - ctest -V

CMakeLists.txt

@@ -1,11 +1,12 @@
 project (liberty C)
-cmake_minimum_required (VERSION 2.8.5)
+cmake_minimum_required (VERSION 2.8.12)
 
 # Moar warnings
-if ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUC)
+if ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUCC)
 	# -Wunused-function is pretty annoying here, as everything is static
-	set (CMAKE_C_FLAGS "-std=c99 -Wall -Wextra -Wno-unused-function")
-endif ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUC)
+	set (wdisabled "-Wno-unused-function")
+	set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -std=c99 -Wall -Wextra ${wdisabled}")
+endif ()
 
 # Dependencies
 set (CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake)
@@ -15,11 +16,9 @@ find_package (PkgConfig REQUIRED)
 pkg_check_modules (libssl REQUIRED libssl libcrypto)
 
 if ("${CMAKE_SYSTEM_NAME}" MATCHES "BSD")
-	include_directories (/usr/local/include)
-	link_directories (/usr/local/lib)
 	# Our POSIX version macros make these undefined
 	add_definitions (-D__BSD_VISIBLE=1 -D_BSD_SOURCE=1)
-endif ("${CMAKE_SYSTEM_NAME}" MATCHES "BSD")
+endif ()
 
 set (common_libraries ${libssl_LIBRARIES})
 include_directories (${libssl_INCLUDE_DIRS})
@@ -31,22 +30,81 @@ foreach (extra iconv rt)
 	find_library (extra_lib_${extra} ${extra})
 	if (extra_lib_${extra})
 		list (APPEND common_libraries ${extra})
-	endif (extra_lib_${extra})
-endforeach (extra)
-
-# Generate a configuration file
-# TODO: actualy use the configuration file for something; so far we allow
-#   for direct inclusion without running this CMakeLists.txt
-configure_file (${PROJECT_SOURCE_DIR}/liberty-config.h.in
-	${PROJECT_BINARY_DIR}/liberty-config.h)
-include_directories (${PROJECT_SOURCE_DIR} ${PROJECT_BINARY_DIR})
-set (common_sources ${PROJECT_BINARY_DIR}/liberty-config.h)
+	endif ()
+endforeach ()
 
 # Build some unit tests
+include_directories (${PROJECT_SOURCE_DIR})
 enable_testing ()
-foreach (name liberty proto)
+set (tests liberty proto)
+
+pkg_check_modules (libpulse libpulse)
+if (libpulse_FOUND)
+	list (APPEND tests pulse)
+	list (APPEND common_libraries ${libpulse_LIBRARIES})
+	include_directories (${libpulse_INCLUDE_DIRS})
+	link_directories (${libpulse_LIBRARY_DIRS})
+endif ()
+
+foreach (name ${tests})
 	add_executable (test-${name} tests/${name}.c ${common_sources})
 	add_threads (test-${name})
 	target_link_libraries (test-${name} ${common_libraries})
 	add_test (NAME test-${name} COMMAND test-${name})
-endforeach (name)
+endforeach ()
+
+# --- Tools --------------------------------------------------------------------
+
+# Test the AsciiDoc manual page generator for a successful parse
+set (ASCIIMAN ${PROJECT_SOURCE_DIR}/tools/asciiman.awk)
+add_custom_command (OUTPUT libertyxdr.7
+	COMMAND env LC_ALL=C awk -f ${ASCIIMAN}
+		"${PROJECT_SOURCE_DIR}/libertyxdr.adoc" > libertyxdr.7
+	DEPENDS libertyxdr.adoc ${ASCIIMAN}
+	COMMENT "Generating man page for libertyxdr" VERBATIM)
+add_custom_target (docs ALL DEPENDS libertyxdr.7)
+
+# Test CMake script parsing
+add_test (test-cmake-parser
+	env LC_ALL=C awk -f ${PROJECT_SOURCE_DIR}/tools/cmake-parser.awk
+	-f ${PROJECT_SOURCE_DIR}/tools/cmake-dump.awk ${CMAKE_CURRENT_LIST_FILE})
+
+# Test protocol code generation
+set (lxdrgen_outputs)
+set (lxdrgen_base "${PROJECT_BINARY_DIR}/lxdrgen.lxdr")
+foreach (backend c go mjs)
+	list (APPEND lxdrgen_outputs ${lxdrgen_base}.${backend})
+	add_custom_command (OUTPUT ${lxdrgen_base}.${backend}
+		COMMAND env LC_ALL=C awk
+			-f ${PROJECT_SOURCE_DIR}/tools/lxdrgen.awk
+			-f ${PROJECT_SOURCE_DIR}/tools/lxdrgen-${backend}.awk
+			-v PrefixCamel=ProtoGen
+			${PROJECT_SOURCE_DIR}/tests/lxdrgen.lxdr
+			> ${lxdrgen_base}.${backend}
+		DEPENDS
+			${PROJECT_SOURCE_DIR}/tools/lxdrgen.awk
+			${PROJECT_SOURCE_DIR}/tools/lxdrgen-${backend}.awk
+			${PROJECT_SOURCE_DIR}/tests/lxdrgen.lxdr
+		COMMENT "Generating test protocol code (${backend})" VERBATIM)
+endforeach ()
+add_custom_target (test-lxdrgen-outputs ALL DEPENDS ${lxdrgen_outputs})
+
+set_source_files_properties (${lxdrgen_base}.c
+	PROPERTIES HEADER_FILE_ONLY TRUE)
+add_executable (test-lxdrgen tests/lxdrgen.c ${lxdrgen_base}.c)
+target_include_directories (test-lxdrgen PUBLIC ${PROJECT_BINARY_DIR})
+add_test (NAME test-lxdrgen-c COMMAND test-lxdrgen)
+
+find_program (GO_EXECUTABLE go)
+if (GO_EXECUTABLE)
+	add_test (test-lxdrgen-go ${GO_EXECUTABLE} vet ${lxdrgen_base}.go)
+else ()
+	message (WARNING "Cannot test generated protocol code for Go")
+endif ()
+
+find_program (NODE_EXECUTABLE node)
+if (NODE_EXECUTABLE)
+	add_test (test-lxdrgen-mjs ${NODE_EXECUTABLE} -c ${lxdrgen_base}.mjs)
+else ()
+	message (WARNING "Cannot test generated protocol code for Javascript")
+endif ()

LICENSE

@@ -1,9 +1,7 @@
- Copyright (c) 2014 - 2016, Přemysl Janouch <p.janouch@gmail.com>
- All rights reserved.
+Copyright (c) 2014 - 2022, Přemysl Eric Janouch <p@janouch.name>
 
 Permission to use, copy, modify, and/or distribute this software for any
- purpose with or without fee is hereby granted, provided that the above
- copyright notice and this permission notice appear in all copies.
+purpose with or without fee is hereby granted.
 
 THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF

README.adoc

@@ -17,18 +17,51 @@ All development is done on Linux, but other POSIX-compatible operating systems
 should be supported as well.  They have an extremely low priority, however, and
 I'm not testing them at all, with the exception of OpenBSD.
 
+Tools
+-----
+This project also hosts a number of supporting scripts written in portable AWK:
+
+asciiman.awk::
+	A fallback manual page generator for AsciiDoc documents,
+	motivated by the hugeness of AsciiDoc's and Asciidoctor's dependency trees.
+	It uses the _man_ macro package.
+
+cmake-parser.awk::
+	Parses the CMake language to the extent that is necessary to reliably
+	extract project versions.  Its greatest limitation is its inability
+	to expand variables, which would require a full interpreter.
+
+cmake-dump.awk::
+	This can be used in conjunction with the previous script to dump CMake
+	scripts in a normalized format for further processing.
+
+lxdrgen.awk::
+	Protocol code generator for a variant of XDR,
+	which is link:libertyxdr.adoc[documented separately].
+	Successfully employed in https://git.janouch.name/p/xK[xK].
+
+lxdrgen-c.awk::
+	LibertyXDR backend that builds on top of the C pseudolibrary.
+
+lxdrgen-go.awk::
+	LibertyXDR backend for Go, supporting _encoding/json_ interfaces.  It also
+	produces optimized JSON marshallers (however, note that the _json.Marshaler_
+	interface is bound to be underperforming, due to the amount of otherwise
+	avoidable memory allocations it necessitates).
+
+lxdrgen-mjs.awk::
+	LibertyXDR backend for Javascript, currently for decoding only.
+	It cuts a corner by not using BigInts, on par with `JSON.parse()`.
+
 Contributing and Support
 ------------------------
-Use this project's GitHub to report any bugs, request features, or submit pull
-requests.  If you want to discuss this project, or maybe just hang out with
-the developer, feel free to join me at irc://irc.janouch.name, channel #dev.
+Use https://git.janouch.name/p/liberty to report any bugs, request features,
+or submit pull requests.  `git send-email` is tolerated.  If you want to discuss
+the project, feel free to join me at ircs://irc.janouch.name, channel #dev.
+
+Bitcoin donations are accepted at: 12r5uEWEgcHC46xd64tt3hHt9EUvYYDHe9
 
 License
 -------
-'liberty' is written by Přemysl Janouch <p.janouch@gmail.com>.
-
-You may use the software under the terms of the ISC license, the text of which
-is included within the package, or, at your option, you may relicense the work
-under the MIT or the Modified BSD License, as listed at the following site:
-
-http://www.gnu.org/licenses/license-list.html
+This software is released under the terms of the 0BSD license, the text of which
+is included within the package along with the list of authors.

cmake/FindNcursesw.cmake

@@ -0,0 +1,17 @@
+# Public Domain
+
+find_package (PkgConfig REQUIRED)
+pkg_check_modules (Ncursesw QUIET ncursesw)
+
+# OpenBSD doesn't provide a pkg-config file
+set (required_vars Ncursesw_LIBRARIES)
+if (NOT Ncursesw_FOUND)
+	find_library (Ncursesw_LIBRARIES NAMES ncursesw)
+	find_path (Ncursesw_INCLUDE_DIRS ncurses.h)
+	list (APPEND required_vars Ncursesw_INCLUDE_DIRS)
+endif (NOT Ncursesw_FOUND)
+
+include (FindPackageHandleStandardArgs)
+FIND_PACKAGE_HANDLE_STANDARD_ARGS (Ncursesw DEFAULT_MSG ${required_vars})
+
+mark_as_advanced (Ncursesw_LIBRARIES Ncursesw_INCLUDE_DIRS)

cmake/FindUnistring.cmake

@@ -0,0 +1,10 @@
+# Public Domain
+
+find_path (Unistring_INCLUDE_DIRS unistr.h)
+find_library (Unistring_LIBRARIES NAMES unistring libunistring)
+
+include (FindPackageHandleStandardArgs)
+FIND_PACKAGE_HANDLE_STANDARD_ARGS (Unistring DEFAULT_MSG
+	Unistring_INCLUDE_DIRS Unistring_LIBRARIES)
+
+mark_as_advanced (Unistring_LIBRARIES Unistring_INCLUDE_DIRS)

fuzz

@@ -0,0 +1,18 @@
+#!/bin/sh
+# I'm not sure how to make maximum use of this invention
+# Make sure to have llvm-symbolizer installed
+clang -g -fsanitize=address,undefined,fuzzer -fno-sanitize-recover=all \
+	tests/fuzz.c -o fuzz-executor
+
+fuzz () {
+	echo "`tput bold`-- Fuzzing $1`tput sgr0`"
+	mkdir -p /tmp/corpus-$1
+	./fuzz-executor -test=$1 -artifact_prefix=$1- \
+		-max_total_time=600 -timeout=1 /tmp/corpus-$1
+}
+
+if [ $# -gt 0 ]; then
+	for test in "$@"; do fuzz $test; done
+else
+	for test in $(./fuzz-executor); do fuzz $test; done
+fi

liberty-proto.c

@@ -1,12 +1,10 @@
 /*
  * liberty-proto.c: the ultimate C unlibrary: protocols
  *
- * Copyright (c) 2014 - 2016, Přemysl Janouch <p.janouch@gmail.com>
- * All rights reserved.
+ * Copyright (c) 2014 - 2016, Přemysl Eric Janouch <p@janouch.name>
  *
  * Permission to use, copy, modify, and/or distribute this software for any
- * purpose with or without fee is hereby granted, provided that the above
- * copyright notice and this permission notice appear in all copies.
+ * purpose with or without fee is hereby granted.
  *
  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
@@ -29,15 +27,13 @@ struct irc_message
 	struct str_map tags;                ///< IRC 3.2 message tags
 	char *prefix;                       ///< Message prefix
 	char *command;                      ///< IRC command
-	struct str_vector params;           ///< Command parameters
+	struct strv params;                 ///< Command parameters
 };
 
 static char *
 irc_unescape_message_tag (const char *value)
 {
-	struct str s;
-	str_init (&s);
-
+	struct str s = str_make ();
 	bool escape = false;
 	for (const char *p = value; *p; p++)
 	{
@@ -64,8 +60,7 @@ irc_unescape_message_tag (const char *value)
 static void
 irc_parse_message_tags (const char *tags, struct str_map *out)
 {
-	struct str_vector v;
-	str_vector_init (&v);
+	struct strv v = strv_make ();
 	cstr_split (tags, ";", true, &v);
 
 	for (size_t i = 0; i < v.len; i++)
@@ -79,19 +74,16 @@ irc_parse_message_tags (const char *tags, struct str_map *out)
 		else
 			str_map_set (out, key, xstrdup (""));
 	}
-
-	str_vector_free (&v);
+	strv_free (&v);
 }
 
 static void
 irc_parse_message (struct irc_message *msg, const char *line)
 {
-	str_map_init (&msg->tags);
-	msg->tags.free = free;
-
+	msg->tags = str_map_make (free);
 	msg->prefix = NULL;
 	msg->command = NULL;
-	str_vector_init (&msg->params);
+	msg->params = strv_make ();
 
 	// IRC 3.2 message tags
 	if (*line == '@')
@@ -132,7 +124,7 @@ irc_parse_message (struct irc_message *msg, const char *line)
 
 		if (*line == ':')
 		{
-			str_vector_add (&msg->params, ++line);
+			strv_append (&msg->params, ++line);
 			break;
 		}
 
@@ -140,7 +132,7 @@ irc_parse_message (struct irc_message *msg, const char *line)
 		if (!param_len)
 			break;
 
-		str_vector_add_owned (&msg->params, xstrndup (line, param_len));
+		strv_append_owned (&msg->params, xstrndup (line, param_len));
 		line += param_len;
 	}
 }
@@ -151,7 +143,7 @@ irc_free_message (struct irc_message *msg)
 	str_map_free (&msg->tags);
 	free (msg->prefix);
 	free (msg->command);
-	str_vector_free (&msg->params);
+	strv_free (&msg->params);
 }
 
 static void
@@ -221,6 +213,7 @@ irc_fnmatch (const char *pattern, const char *string)
 	char x_pattern[pattern_size], x_string[string_size];
 	irc_strxfrm (x_pattern, pattern, pattern_size);
 	irc_strxfrm (x_string,  string,  string_size);
+	// FIXME: this supports [], which is not mentioned in RFC 2812
 	return fnmatch (x_pattern, x_string, 0);
 }
 
@@ -289,14 +282,15 @@ struct http_tokenizer
 	struct str string;                  ///< "token" / "quoted-string" content
 };
 
-static void
-http_tokenizer_init (struct http_tokenizer *self, const char *input, size_t len)
+static struct http_tokenizer
+http_tokenizer_make (const char *input, size_t len)
 {
-	memset (self, 0, sizeof *self);
-	self->input = (const unsigned char *) input;
-	self->input_len = len;
-
-	str_init (&self->string);
+	return (struct http_tokenizer)
+	{
+		.input = (const unsigned char *) input,
+		.input_len = len,
+		.string = str_make (),
+	};
 }
 
 static void
@@ -429,8 +423,8 @@ http_parse_media_type (const char *media_type,
 	char **type, char **subtype, struct str_map *parameters)
 {
 	bool result = false;
-	struct http_tokenizer t;
-	http_tokenizer_init (&t, media_type, strlen (media_type));
+	struct http_tokenizer t =
+		http_tokenizer_make (media_type, strlen (media_type));
 
 	if (http_tokenizer_next (&t, true) != HTTP_T_TOKEN)
 		goto end;
@@ -491,8 +485,7 @@ http_parse_upgrade (const char *upgrade, struct http_protocol **out)
 	struct http_protocol *list = NULL;
 	struct http_protocol *tail = NULL;
 
-	struct http_tokenizer t;
-	http_tokenizer_init (&t, upgrade, strlen (upgrade));
+	struct http_tokenizer t = http_tokenizer_make (upgrade, strlen (upgrade));
 
 	enum {
 		STATE_PROTOCOL_NAME,
@@ -619,16 +612,16 @@ struct scgi_parser
 	void *user_data;                    ///< User data passed to callbacks
 };
 
-static void
-scgi_parser_init (struct scgi_parser *self)
+static struct scgi_parser
+scgi_parser_make (void)
 {
-	memset (self, 0, sizeof *self);
-
-	str_init (&self->input);
-	str_map_init (&self->headers);
-	self->headers.free = free;
-	str_init (&self->name);
-	str_init (&self->value);
+	return (struct scgi_parser)
+	{
+		.input = str_make (),
+		.headers = str_map_make (free),
+		.name = str_make (),
+		.value = str_make (),
+	};
 }
 
 static void
@@ -671,10 +664,11 @@ scgi_parser_push (struct scgi_parser *self,
 		if (digit == ':')
 		{
 			self->state = SCGI_READING_NAME;
+			str_remove_slice (&self->input, 0, 1);
 			break;
 		}
 
-		if (digit < '0' || digit >= '9')
+		if (digit < '0' || digit > '9')
 			return error_set (e, "invalid header netstring");
 
 		size_t new_len = self->headers_len * 10 + (digit - '0');
@@ -707,6 +701,7 @@ scgi_parser_push (struct scgi_parser *self,
 			self->state = SCGI_READING_VALUE;
 
 		str_remove_slice (&self->input, 0, 1);
+		self->headers_len--;
 		break;
 	}
 	case SCGI_READING_VALUE:
@@ -729,12 +724,13 @@ scgi_parser_push (struct scgi_parser *self,
 				self->name.str, str_steal (&self->value));
 
 			str_reset (&self->name);
-			str_init (&self->value);
+			self->value = str_make ();
 
 			self->state = SCGI_READING_NAME;
 		}
 
 		str_remove_slice (&self->input, 0, 1);
+		self->headers_len--;
 		break;
 	}
 	case SCGI_READING_CONTENT:
@@ -799,7 +795,8 @@ enum fcgi_protocol_status
 
 struct fcgi_parser;
 
-typedef void (*fcgi_message_fn)
+/// Message handler, returns false if further processing should be stopped
+typedef bool (*fcgi_message_fn)
 	(const struct fcgi_parser *parser, void *user_data);
 
 enum fcgi_parser_state
@@ -828,12 +825,11 @@ struct fcgi_parser
 	void *user_data;                    ///< User data
 };
 
-static void
-fcgi_parser_init (struct fcgi_parser *self)
+static struct fcgi_parser
+fcgi_parser_make (void)
 {
-	memset (self, 0, sizeof *self);
-	str_init (&self->input);
-	str_init (&self->content);
+	return (struct fcgi_parser)
+		{ .input = str_make (), .content = str_make () };
 }
 
 static void
@@ -846,8 +842,8 @@ fcgi_parser_free (struct fcgi_parser *self)
 static void
 fcgi_parser_unpack_header (struct fcgi_parser *self)
 {
-	struct msg_unpacker unpacker;
-	msg_unpacker_init (&unpacker, self->input.str, self->input.len);
+	struct msg_unpacker unpacker =
+		msg_unpacker_make (self->input.str, self->input.len);
 
 	bool success = true;
 	uint8_t reserved;
@@ -862,7 +858,7 @@ fcgi_parser_unpack_header (struct fcgi_parser *self)
 	str_remove_slice (&self->input, 0, unpacker.offset);
 }
 
-static void
+static bool
 fcgi_parser_push (struct fcgi_parser *self, const void *data, size_t len)
 {
 	// This could be made considerably faster for high-throughput applications
@@ -874,14 +870,14 @@ fcgi_parser_push (struct fcgi_parser *self, const void *data, size_t len)
 	{
 	case FCGI_READING_HEADER:
 		if (self->input.len < FCGI_HEADER_LEN)
-			return;
+			return true;
 
 		fcgi_parser_unpack_header (self);
 		self->state = FCGI_READING_CONTENT;
 		break;
 	case FCGI_READING_CONTENT:
 		if (self->input.len < self->content_length)
-			return;
+			return true;
 
 		// Move an appropriate part of the input buffer to the content buffer
 		str_reset (&self->content);
@@ -891,10 +887,11 @@ fcgi_parser_push (struct fcgi_parser *self, const void *data, size_t len)
 		break;
 	case FCGI_READING_PADDING:
 		if (self->input.len < self->padding_length)
-			return;
+			return true;
 
 		// Call the callback to further process the message
-		self->on_message (self, self->user_data);
+		if (!self->on_message (self, self->user_data))
+			return false;
 
 		// Remove the padding from the input buffer
 		str_remove_slice (&self->input, 0, self->padding_length);
@@ -929,11 +926,10 @@ struct fcgi_nv_parser
 	char *value;                        ///< The current value, 0-terminated
 };
 
-static void
-fcgi_nv_parser_init (struct fcgi_nv_parser *self)
+static struct fcgi_nv_parser
+fcgi_nv_parser_make (void)
 {
-	memset (self, 0, sizeof *self);
-	str_init (&self->input);
+	return (struct fcgi_nv_parser) { .input = str_make () };
 }
 
 static void
@@ -952,8 +948,8 @@ fcgi_nv_parser_push (struct fcgi_nv_parser *self, const void *data, size_t len)
 
 	while (true)
 	{
-		struct msg_unpacker unpacker;
-		msg_unpacker_init (&unpacker, self->input.str, self->input.len);
+		struct msg_unpacker unpacker =
+			msg_unpacker_make (self->input.str, self->input.len);
 
 	switch (self->state)
 	{
@@ -1050,8 +1046,7 @@ fcgi_nv_convert_len (size_t len, struct str *output)
 static void
 fcgi_nv_convert (struct str_map *map, struct str *output)
 {
-	struct str_map_iter iter;
-	str_map_iter_init (&iter, map);
+	struct str_map_iter iter = str_map_iter_make (map);
 	while (str_map_iter_next (&iter))
 	{
 		const char *name  = iter.link->key;
@@ -1090,8 +1085,7 @@ ws_encode_response_key (const char *key)
 	SHA1 ((unsigned char *) response_key, strlen (response_key), hash);
 	free (response_key);
 
-	struct str base64;
-	str_init (&base64);
+	struct str base64 = str_make ();
 	base64_encode (hash, sizeof hash, &base64);
 	return str_steal (&base64);
 }
@@ -1169,11 +1163,10 @@ struct ws_parser
 	void *user_data;                    ///< User data for callbacks
 };
 
-static void
-ws_parser_init (struct ws_parser *self)
+static struct ws_parser
+ws_parser_make (void)
 {
-	memset (self, 0, sizeof *self);
-	str_init (&self->input);
+	return (struct ws_parser) { .input = str_make () };
 }
 
 static void
@@ -1201,8 +1194,10 @@ ws_parser_unmask (char *payload, uint64_t len, uint32_t mask)
 	{
 	case 3:
 		payload[end + 2] ^= (mask >>  8) & 0xFF;
+		// Fall-through
 	case 2:
 		payload[end + 1] ^= (mask >> 16) & 0xFF;
+		// Fall-through
 	case 1:
 		payload[end    ] ^= (mask >> 24) & 0xFF;
 	}
@@ -1214,8 +1209,8 @@ ws_parser_push (struct ws_parser *self, const void *data, size_t len)
 	bool success = false;
 	str_append_data (&self->input, data, len);
 
-	struct msg_unpacker unpacker;
-	msg_unpacker_init (&unpacker, self->input.str, self->input.len);
+	struct msg_unpacker unpacker =
+		msg_unpacker_make (self->input.str, self->input.len);
 
 	while (true)
 	switch (self->state)
@@ -1276,10 +1271,7 @@ ws_parser_push (struct ws_parser *self, const void *data, size_t len)
 	case WS_PARSER_PAYLOAD:
 		// Move the buffer so that payload data is at the front
 		str_remove_slice (&self->input, 0, unpacker.offset);
-
-		// And continue unpacking frames past the payload
-		msg_unpacker_init (&unpacker, self->input.str, self->input.len);
-		unpacker.offset = self->payload_len;
+		unpacker = msg_unpacker_make (self->input.str, self->input.len);
 
 		if (self->input.len < self->payload_len)
 			goto need_data;
@@ -1288,6 +1280,8 @@ ws_parser_push (struct ws_parser *self, const void *data, size_t len)
 		if (!self->on_frame (self->user_data, self))
 			goto fail;
 
+		// And continue unpacking frames past the payload
+		unpacker.offset = self->payload_len;
 		self->state = WS_PARSER_FIXED;
 		break;
 	}
@@ -1327,7 +1321,6 @@ enum mpd_subsystem
 #define XX(a, b, c) MPD_SUBSYSTEM_ ## a = (1 << b),
 	MPD_SUBSYSTEM_TABLE (XX)
 #undef XX
-	MPD_SUBSYSTEM_MAX
 };
 
 static const char *mpd_subsystem_names[] =
@@ -1358,9 +1351,9 @@ struct mpd_response
 	char *message_text;                 ///< Error message
 };
 
-/// Task completion callback
+/// Task completion callback; on connection abortion most fields are 0
 typedef void (*mpd_client_task_cb) (const struct mpd_response *response,
-	const struct str_vector *data, void *user_data);
+	const struct strv *data, void *user_data);
 
 struct mpd_client_task
 {
@@ -1396,7 +1389,7 @@ struct mpd_client
 
 	struct mpd_client_task *tasks;      ///< Task queue
 	struct mpd_client_task *tasks_tail; ///< Tail of task queue
-	struct str_vector data;             ///< Data from last command
+	struct strv data;                   ///< Data from last command
 
 	// User configuration:
 
@@ -1420,21 +1413,19 @@ struct mpd_client
 static void mpd_client_reset (struct mpd_client *self);
 static void mpd_client_destroy_connector (struct mpd_client *self);
 
-static void
-mpd_client_init (struct mpd_client *self, struct poller *poller)
+static struct mpd_client
+mpd_client_make (struct poller *poller)
 {
-	memset (self, 0, sizeof *self);
-
-	self->poller = poller;
-	self->socket = -1;
-
-	str_init (&self->read_buffer);
-	str_init (&self->write_buffer);
-
-	str_vector_init (&self->data);
-
-	poller_fd_init (&self->socket_event, poller, -1);
-	poller_timer_init (&self->timeout_timer, poller);
+	return (struct mpd_client)
+	{
+		.poller = poller,
+		.socket = -1,
+		.read_buffer = str_make (),
+		.write_buffer = str_make (),
+		.data = strv_make (),
+		.socket_event = poller_fd_make (poller, -1),
+		.timeout_timer = poller_timer_make (poller),
+	};
 }
 
 static void
@@ -1446,15 +1437,36 @@ mpd_client_free (struct mpd_client *self)
 	str_free (&self->read_buffer);
 	str_free (&self->write_buffer);
 
-	str_vector_free (&self->data);
+	strv_free (&self->data);
 }
 
 // - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 
+static void
+mpd_client_dispatch (struct mpd_client *self, struct mpd_response *response)
+{
+	struct mpd_client_task *task;
+	if (!(task = self->tasks))
+		return;
+
+	if (task->callback)
+		task->callback (response, &self->data, task->user_data);
+	strv_reset (&self->data);
+
+	LIST_UNLINK_WITH_TAIL (self->tasks, self->tasks_tail, task);
+	free (task);
+}
+
 /// Reinitialize the interface so that you can reconnect anew
 static void
 mpd_client_reset (struct mpd_client *self)
 {
+	// Get rid of all pending tasks to release resources etc.
+	strv_reset (&self->data);
+	struct mpd_response aborted = { .message_text = "Disconnected" };
+	while (self->tasks)
+		mpd_client_dispatch (self, &aborted);
+
 	if (self->state == MPD_CONNECTING)
 		mpd_client_destroy_connector (self);
 
@@ -1462,6 +1474,7 @@ mpd_client_reset (struct mpd_client *self)
 		xclose (self->socket);
 	self->socket = -1;
 
+	// FIXME: this is not robust wrt. forking
 	self->socket_event.closed = true;
 	poller_fd_reset (&self->socket_event);
 	poller_timer_reset (&self->timeout_timer);
@@ -1469,17 +1482,11 @@ mpd_client_reset (struct mpd_client *self)
 	str_reset (&self->read_buffer);
 	str_reset (&self->write_buffer);
 
-	str_vector_reset (&self->data);
-
 	self->got_hello = false;
 	self->idling = false;
 	self->idling_subsystems = 0;
 	self->in_list = false;
 
-	LIST_FOR_EACH (struct mpd_client_task, iter, self->tasks)
-		free (iter);
-	self->tasks = self->tasks_tail = NULL;
-
 	self->state = MPD_DISCONNECTED;
 }
 
@@ -1530,21 +1537,6 @@ mpd_client_parse_response (const char *p, struct mpd_response *response)
 	return true;
 }
 
-static void
-mpd_client_dispatch (struct mpd_client *self, struct mpd_response *response)
-{
-	struct mpd_client_task *task;
-	if (!(task = self->tasks))
-		return;
-
-	if (task->callback)
-		task->callback (response, &self->data, task->user_data);
-	str_vector_reset (&self->data);
-
-	LIST_UNLINK_WITH_TAIL (self->tasks, self->tasks_tail, task);
-	free (task);
-}
-
 static bool
 mpd_client_parse_hello (struct mpd_client *self, const char *line)
 {
@@ -1572,15 +1564,14 @@ mpd_client_parse_line (struct mpd_client *self, const char *line)
 	struct mpd_response response;
 	memset (&response, 0, sizeof response);
 	if (!strcmp (line, "list_OK"))
-		str_vector_add_owned (&self->data, NULL);
+		strv_append_owned (&self->data, NULL);
 	else if (mpd_client_parse_response (line, &response))
-	{
 		mpd_client_dispatch (self, &response);
+	else
+		strv_append (&self->data, line);
+
 	free (response.current_command);
 	free (response.message_text);
-	}
-	else
-		str_vector_add (&self->data, line);
 	return true;
 }
 
@@ -1695,8 +1686,9 @@ mpd_client_add_task
 static void mpd_client_send_command
 	(struct mpd_client *self, const char *command, ...) ATTRIBUTE_SENTINEL;
 
+/// Avoid calling this method directly if you don't want things to explode
 static void
-mpd_client_send_commandv (struct mpd_client *self, char **commands)
+mpd_client_send_command_raw (struct mpd_client *self, const char *raw)
 {
 	// Automatically interrupt idle mode
 	if (self->idling)
@@ -1708,48 +1700,52 @@ mpd_client_send_commandv (struct mpd_client *self, char **commands)
 		mpd_client_send_command (self, "noidle", NULL);
 	}
 
-	struct str line;
-	str_init (&line);
-
-	for (; *commands; commands++)
-	{
-		if (line.len)
-			str_append_c (&line, ' ');
-
-		if (mpd_client_must_quote (*commands))
-			mpd_client_quote (*commands, &line);
-		else
-			str_append (&line, *commands);
-	}
-
 	if (self->on_io_hook)
-		self->on_io_hook (self->user_data, true, line.str);
+		self->on_io_hook (self->user_data, true, raw);
 
-	str_append_c (&line, '\n');
-	str_append_str (&self->write_buffer, &line);
-	str_free (&line);
+	str_append (&self->write_buffer, raw);
+	str_append_c (&self->write_buffer, '\n');
 
 	mpd_client_update_poller (self);
 }
 
+static void
+mpd_client_send_commandv (struct mpd_client *self, char **fields)
+{
+	struct str line = str_make ();
+	for (; *fields; fields++)
+	{
+		if (line.len)
+			str_append_c (&line, ' ');
+
+		if (mpd_client_must_quote (*fields))
+			mpd_client_quote (*fields, &line);
+		else
+			str_append (&line, *fields);
+	}
+	mpd_client_send_command_raw (self, line.str);
+	str_free (&line);
+}
+
 static void
 mpd_client_send_command (struct mpd_client *self, const char *command, ...)
 {
-	struct str_vector v;
-	str_vector_init (&v);
+	struct strv v = strv_make ();
 
 	va_list ap;
 	va_start (ap, command);
 	for (; command; command = va_arg (ap, const char *))
-		str_vector_add (&v, command);
+		strv_append (&v, command);
 	va_end (ap);
 
 	mpd_client_send_commandv (self, v.vector);
-	str_vector_free (&v);
+	strv_free (&v);
 }
 
 // - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 
+/// "On success for all commands, OK is returned.  If a command fails, no more
+/// commands are executed and the appropriate ACK error is returned"
 static void
 mpd_client_list_begin (struct mpd_client *self)
 {
@@ -1793,7 +1789,7 @@ mpd_resolve_subsystem (const char *name, unsigned *output)
 
 static void
 mpd_client_on_idle_return (const struct mpd_response *response,
-	const struct str_vector *data, void *user_data)
+	const struct strv *data, void *user_data)
 {
 	(void) response;
 
@@ -1836,16 +1832,14 @@ mpd_client_idle (struct mpd_client *self, unsigned subsystems)
 {
 	hard_assert (!self->in_list);
 
-	struct str_vector v;
-	str_vector_init (&v);
-
-	str_vector_add (&v, "idle");
+	struct strv v = strv_make ();
+	strv_append (&v, "idle");
 	for (size_t i = 0; i < N_ELEMENTS (mpd_subsystem_names); i++)
 		if (subsystems & (1 << i))
-			str_vector_add (&v, mpd_subsystem_names[i]);
+			strv_append (&v, mpd_subsystem_names[i]);
 
 	mpd_client_send_commandv (self, v.vector);
-	str_vector_free (&v);
+	strv_free (&v);
 
 	self->timeout_timer.dispatcher = mpd_client_on_timeout;
 	self->timeout_timer.user_data = self;
@@ -1865,7 +1859,7 @@ mpd_client_finish_connection (struct mpd_client *self, int socket)
 	self->socket = socket;
 	self->state = MPD_CONNECTED;
 
-	poller_fd_init (&self->socket_event, self->poller, self->socket);
+	self->socket_event = poller_fd_make (self->poller, self->socket);
 	self->socket_event.dispatcher = mpd_client_on_ready;
 	self->socket_event.user_data = self;
 
@@ -1925,7 +1919,11 @@ mpd_client_connect_unix (struct mpd_client *self, const char *address,
 	free (expanded);
 
 	if (connect (fd, (struct sockaddr *) &sun, sizeof sun))
-		return error_set (e, "%s: %s", "connect", strerror (errno));
+	{
+		error_set (e, "%s: %s", "connect", strerror (errno));
+		xclose (fd);
+		return false;
+	}
 
 	mpd_client_finish_connection (self, fd);
 	return true;

liberty-pulse.c

@@ -0,0 +1,348 @@
+/*
+ * liberty-pulse.c: PulseAudio mainloop abstraction
+ *
+ * Copyright (c) 2016 - 2021, Přemysl Eric Janouch <p@janouch.name>
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ */
+
+#include <pulse/mainloop.h>
+
+// --- PulseAudio mainloop abstraction -----------------------------------------
+
+struct pa_io_event
+{
+	LIST_HEADER (pa_io_event)
+
+	pa_mainloop_api *api;               ///< Parent structure
+	struct poller_fd fd;                ///< Underlying FD event
+
+	pa_io_event_cb_t dispatch;          ///< Dispatcher
+	pa_io_event_destroy_cb_t free;      ///< Destroyer
+	void *user_data;                    ///< User data
+};
+
+struct pa_time_event
+{
+	LIST_HEADER (pa_time_event)
+
+	pa_mainloop_api *api;               ///< Parent structure
+	struct poller_timer timer;          ///< Underlying timer event
+
+	pa_time_event_cb_t dispatch;        ///< Dispatcher
+	pa_time_event_destroy_cb_t free;    ///< Destroyer
+	void *user_data;                    ///< User data
+};
+
+struct pa_defer_event
+{
+	LIST_HEADER (pa_defer_event)
+
+	pa_mainloop_api *api;               ///< Parent structure
+	struct poller_idle idle;            ///< Underlying idle event
+
+	pa_defer_event_cb_t dispatch;       ///< Dispatcher
+	pa_defer_event_destroy_cb_t free;   ///< Destroyer
+	void *user_data;                    ///< User data
+};
+
+struct poller_pa
+{
+	struct poller *poller;              ///< The underlying event loop
+	pa_io_event *io_list;               ///< I/O events
+	pa_time_event *time_list;           ///< Timer events
+	pa_defer_event *defer_list;         ///< Deferred events
+};
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+static short
+poller_pa_flags_to_events (pa_io_event_flags_t flags)
+{
+	short result = 0;
+	if (flags & PA_IO_EVENT_ERROR)   result |= POLLERR;
+	if (flags & PA_IO_EVENT_HANGUP)  result |= POLLHUP;
+	if (flags & PA_IO_EVENT_INPUT)   result |= POLLIN;
+	if (flags & PA_IO_EVENT_OUTPUT)  result |= POLLOUT;
+	return result;
+}
+
+static pa_io_event_flags_t
+poller_pa_events_to_flags (short events)
+{
+	pa_io_event_flags_t result = 0;
+	if (events & POLLERR)  result |= PA_IO_EVENT_ERROR;
+	if (events & POLLHUP)  result |= PA_IO_EVENT_HANGUP;
+	if (events & POLLIN)   result |= PA_IO_EVENT_INPUT;
+	if (events & POLLOUT)  result |= PA_IO_EVENT_OUTPUT;
+	return result;
+}
+
+static struct timeval
+poller_pa_get_current_time (void)
+{
+	struct timeval tv;
+#ifdef _POSIX_TIMERS
+	struct timespec tp;
+	hard_assert (clock_gettime (CLOCK_REALTIME, &tp) != -1);
+	tv.tv_sec = tp.tv_sec;
+	tv.tv_usec = tp.tv_nsec / 1000;
+#else
+	gettimeofday (&tv, NULL);
+#endif
+	return tv;
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+static void
+poller_pa_io_dispatcher (const struct pollfd *pfd, void *user_data)
+{
+	pa_io_event *self = user_data;
+	self->dispatch (self->api, self,
+		pfd->fd, poller_pa_events_to_flags (pfd->revents), self->user_data);
+}
+
+static void
+poller_pa_io_enable (pa_io_event *self, pa_io_event_flags_t events)
+{
+	struct poller_fd *fd = &self->fd;
+	if (events)
+		poller_fd_set (fd, poller_pa_flags_to_events (events));
+	else
+		poller_fd_reset (fd);
+}
+
+static pa_io_event *
+poller_pa_io_new (pa_mainloop_api *api, int fd_, pa_io_event_flags_t events,
+	pa_io_event_cb_t cb, void *userdata)
+{
+	pa_io_event *self = xcalloc (1, sizeof *self);
+	self->api = api;
+	self->dispatch = cb;
+	self->user_data = userdata;
+
+	struct poller_pa *data = api->userdata;
+	self->fd = poller_fd_make (data->poller, fd_);
+	self->fd.user_data = self;
+	self->fd.dispatcher = poller_pa_io_dispatcher;
+
+	// FIXME: under x2go PA tries to register twice for the same FD,
+	//   which fails with our curent poller implementation;
+	//   we could maintain a list of { poller_fd, listeners } structures;
+	//   or maybe we're doing something wrong, which is yet to be determined
+	poller_pa_io_enable (self, events);
+	LIST_PREPEND (data->io_list, self);
+	return self;
+}
+
+static void
+poller_pa_io_free (pa_io_event *self)
+{
+	if (self->free)
+		self->free (self->api, self, self->user_data);
+
+	struct poller_pa *data = self->api->userdata;
+	poller_fd_reset (&self->fd);
+	LIST_UNLINK (data->io_list, self);
+	free (self);
+}
+
+static void
+poller_pa_io_set_destroy (pa_io_event *self, pa_io_event_destroy_cb_t cb)
+{
+	self->free = cb;
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+static void
+poller_pa_time_dispatcher (void *user_data)
+{
+	pa_time_event *self = user_data;
+	// XXX: the meaning of the time argument is undocumented,
+	//   so let's just put current Unix time in there
+	struct timeval now = poller_pa_get_current_time ();
+	self->dispatch (self->api, self, &now, self->user_data);
+}
+
+static void
+poller_pa_time_restart (pa_time_event *self, const struct timeval *tv)
+{
+	struct poller_timer *timer = &self->timer;
+	if (tv)
+	{
+		struct timeval now = poller_pa_get_current_time ();
+		poller_timer_set (timer,
+			(tv->tv_sec  - now.tv_sec)  * 1000 +
+			(tv->tv_usec - now.tv_usec) / 1000);
+	}
+	else
+		poller_timer_reset (timer);
+}
+
+static pa_time_event *
+poller_pa_time_new (pa_mainloop_api *api, const struct timeval *tv,
+	pa_time_event_cb_t cb, void *userdata)
+{
+	pa_time_event *self = xcalloc (1, sizeof *self);
+	self->api = api;
+	self->dispatch = cb;
+	self->user_data = userdata;
+
+	struct poller_pa *data = api->userdata;
+	self->timer = poller_timer_make (data->poller);
+	self->timer.user_data = self;
+	self->timer.dispatcher = poller_pa_time_dispatcher;
+
+	poller_pa_time_restart (self, tv);
+	LIST_PREPEND (data->time_list, self);
+	return self;
+}
+
+static void
+poller_pa_time_free (pa_time_event *self)
+{
+	if (self->free)
+		self->free (self->api, self, self->user_data);
+
+	struct poller_pa *data = self->api->userdata;
+	poller_timer_reset (&self->timer);
+	LIST_UNLINK (data->time_list, self);
+	free (self);
+}
+
+static void
+poller_pa_time_set_destroy (pa_time_event *self, pa_time_event_destroy_cb_t cb)
+{
+	self->free = cb;
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+static void
+poller_pa_defer_dispatcher (void *user_data)
+{
+	pa_defer_event *self = user_data;
+	self->dispatch (self->api, self, self->user_data);
+}
+
+static pa_defer_event *
+poller_pa_defer_new (pa_mainloop_api *api,
+	pa_defer_event_cb_t cb, void *userdata)
+{
+	pa_defer_event *self = xcalloc (1, sizeof *self);
+	self->api = api;
+	self->dispatch = cb;
+	self->user_data = userdata;
+
+	struct poller_pa *data = api->userdata;
+	self->idle = poller_idle_make (data->poller);
+	self->idle.user_data = self;
+	self->idle.dispatcher = poller_pa_defer_dispatcher;
+
+	poller_idle_set (&self->idle);
+	LIST_PREPEND (data->defer_list, self);
+	return self;
+}
+
+static void
+poller_pa_defer_enable (pa_defer_event *self, int enable)
+{
+	struct poller_idle *idle = &self->idle;
+	if (enable)
+		poller_idle_set (idle);
+	else
+		poller_idle_reset (idle);
+}
+
+static void
+poller_pa_defer_free (pa_defer_event *self)
+{
+	if (self->free)
+		self->free (self->api, self, self->user_data);
+
+	struct poller_pa *data = self->api->userdata;
+	poller_idle_reset (&self->idle);
+	LIST_UNLINK (data->defer_list, self);
+	free (self);
+}
+
+static void
+poller_pa_defer_set_destroy (pa_defer_event *self,
+	pa_defer_event_destroy_cb_t cb)
+{
+	self->free = cb;
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+static void
+poller_pa_quit (pa_mainloop_api *api, int retval)
+{
+	(void) api;
+	(void) retval;
+
+	// This is not called from within libpulse
+	hard_assert (!"quitting the libpulse event loop is unimplemented");
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+static struct pa_mainloop_api g_poller_pa_template =
+{
+	.io_new            = poller_pa_io_new,
+	.io_enable         = poller_pa_io_enable,
+	.io_free           = poller_pa_io_free,
+	.io_set_destroy    = poller_pa_io_set_destroy,
+
+	.time_new          = poller_pa_time_new,
+	.time_restart      = poller_pa_time_restart,
+	.time_free         = poller_pa_time_free,
+	.time_set_destroy  = poller_pa_time_set_destroy,
+
+	.defer_new         = poller_pa_defer_new,
+	.defer_enable      = poller_pa_defer_enable,
+	.defer_free        = poller_pa_defer_free,
+	.defer_set_destroy = poller_pa_defer_set_destroy,
+
+	.quit              = poller_pa_quit,
+};
+
+static struct pa_mainloop_api *
+poller_pa_new (struct poller *self)
+{
+	struct poller_pa *data = xcalloc (1, sizeof *data);
+	data->poller = self;
+
+	struct pa_mainloop_api *api = xmalloc (sizeof *api);
+	*api = g_poller_pa_template;
+	api->userdata = data;
+	return api;
+}
+
+static void
+poller_pa_destroy (struct pa_mainloop_api *api)
+{
+	struct poller_pa *data = api->userdata;
+
+	LIST_FOR_EACH (pa_io_event, iter, data->io_list)
+		poller_pa_io_free (iter);
+	LIST_FOR_EACH (pa_time_event, iter, data->time_list)
+		poller_pa_time_free (iter);
+	LIST_FOR_EACH (pa_defer_event, iter, data->defer_list)
+		poller_pa_defer_free (iter);
+
+	free (data);
+	free (api);
+}

liberty-tui.c

@@ -0,0 +1,270 @@
+/*
+ * liberty-tui.c: the ultimate C unlibrary: TUI
+ *
+ * Copyright (c) 2016 - 2017, Přemysl Eric Janouch <p@janouch.name>
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ */
+
+// This file includes some common stuff to build TUI applications with
+
+#include <ncurses.h>
+
+// It is surprisingly hard to find a good library to handle Unicode shenanigans,
+// and there's enough of those for it to be impractical to reimplement them.
+//
+//                         GLib          ICU     libunistring    utf8proc
+// Decently sized            .            .            x            x
+// Grapheme breaks           .            x            .            x
+// Character width           x            .            x            x
+// Locale handling           .            .            x            .
+// Liberal license           .            x            .            x
+//
+// Also note that the ICU API is icky and uses UTF-16 for its primary encoding.
+//
+// Currently we're chugging along with libunistring but utf8proc seems viable.
+// Non-Unicode locales can mostly be handled with simple iconv like in sdtui.
+// Similarly grapheme breaks can be guessed at using character width (a basic
+// test here is Zalgo text).
+//
+// None of this is ever going to work too reliably anyway because terminals
+// and Unicode don't go awfully well together.  In particular, character cell
+// devices have some problems with double-wide characters.
+
+#include <unistr.h>
+#include <uniwidth.h>
+#include <uniconv.h>
+#include <unicase.h>
+
+// --- Configurable display attributes -----------------------------------------
+
+struct attrs
+{
+	short fg;                           ///< Foreground colour index
+	short bg;                           ///< Background colour index
+	chtype attrs;                       ///< Other attributes
+};
+
+/// Decode attributes in the value using a subset of the git config format,
+/// ignoring all errors since it doesn't affect functionality
+static struct attrs
+attrs_decode (const char *value)
+{
+	struct strv v = strv_make ();
+	cstr_split (value, " ", true, &v);
+
+	int colors = 0;
+	struct attrs attrs = { -1, -1, 0 };
+	for (char **it = v.vector; *it; it++)
+	{
+		char *end = NULL;
+		long n = strtol (*it, &end, 10);
+		if (*it != end && !*end && n >= SHRT_MIN && n <= SHRT_MAX)
+		{
+			if (colors == 0) attrs.fg = n;
+			if (colors == 1) attrs.bg = n;
+			colors++;
+		}
+		else if (!strcmp (*it, "bold"))    attrs.attrs |= A_BOLD;
+		else if (!strcmp (*it, "dim"))     attrs.attrs |= A_DIM;
+		else if (!strcmp (*it, "ul"))      attrs.attrs |= A_UNDERLINE;
+		else if (!strcmp (*it, "blink"))   attrs.attrs |= A_BLINK;
+		else if (!strcmp (*it, "reverse")) attrs.attrs |= A_REVERSE;
+#ifdef A_ITALIC
+		else if (!strcmp (*it, "italic"))  attrs.attrs |= A_ITALIC;
+#endif  // A_ITALIC
+	}
+	strv_free (&v);
+	return attrs;
+}
+
+// --- Terminal output ---------------------------------------------------------
+
+// Necessary abstraction to simplify aligned, formatted character output
+
+// This callback you need to implement in the application
+static bool app_is_character_in_locale (ucs4_t ch);
+
+struct row_char
+{
+	ucs4_t c;                           ///< Unicode codepoint
+	chtype attrs;                       ///< Special attributes
+	int width;                          ///< How many cells this takes
+};
+
+struct row_buffer
+{
+	ARRAY (struct row_char, chars)      ///< Characters
+	int total_width;                    ///< Total width of all characters
+};
+
+static struct row_buffer
+row_buffer_make (void)
+{
+	struct row_buffer self = {};
+	ARRAY_INIT_SIZED (self.chars, 256);
+	return self;
+}
+
+static void
+row_buffer_free (struct row_buffer *self)
+{
+	free (self->chars);
+}
+
+/// Replace invalid chars and push all codepoints to the array w/ attributes.
+static void
+row_buffer_append (struct row_buffer *self, const char *str, chtype attrs)
+{
+	// The encoding is only really used internally for some corner cases
+	const char *encoding = locale_charset ();
+
+	// Note that this function is a hotspot, try to keep it decently fast
+	struct row_char current = { .attrs = attrs };
+	struct row_char invalid = { .attrs = attrs, .c = '?', .width = 1 };
+	const uint8_t *next = (const uint8_t *) str;
+	while ((next = u8_next (&current.c, next)))
+	{
+		current.width = uc_width (current.c, encoding);
+		if (current.width < 0 || !app_is_character_in_locale (current.c))
+			current = invalid;
+
+		ARRAY_RESERVE (self->chars, 1);
+		self->chars[self->chars_len++] = current;
+		self->total_width += current.width;
+	}
+}
+
+static void
+row_buffer_append_args (struct row_buffer *self, const char *s, ...)
+	ATTRIBUTE_SENTINEL;
+
+static void
+row_buffer_append_args (struct row_buffer *self, const char *s, ...)
+{
+	va_list ap;
+	va_start (ap, s);
+
+	while (s)
+	{
+		row_buffer_append (self, s, va_arg (ap, chtype));
+		s = va_arg (ap, const char *);
+	}
+	va_end (ap);
+}
+
+static void
+row_buffer_append_buffer (struct row_buffer *self, const struct row_buffer *rb)
+{
+	ARRAY_RESERVE (self->chars, rb->chars_len);
+	memcpy (self->chars + self->chars_len, rb->chars,
+		rb->chars_len * sizeof *rb->chars);
+
+	self->chars_len   += rb->chars_len;
+	self->total_width += rb->total_width;
+}
+
+/// Pop as many codepoints as needed to free up "space" character cells.
+/// Given the suffix nature of combining marks, this should work pretty fine.
+static int
+row_buffer_pop_cells (struct row_buffer *self, int space)
+{
+	int made = 0;
+	while (self->chars_len && made < space)
+		made += self->chars[--self->chars_len].width;
+	self->total_width -= made;
+	return made;
+}
+
+static void
+row_buffer_space (struct row_buffer *self, int width, chtype attrs)
+{
+	if (width < 0)
+		return;
+
+	ARRAY_RESERVE (self->chars, (size_t) width);
+
+	struct row_char space = { .attrs = attrs, .c = ' ', .width = 1 };
+	self->total_width += width;
+	while (width-- > 0)
+		self->chars[self->chars_len++] = space;
+}
+
+static void
+row_buffer_ellipsis (struct row_buffer *self, int target)
+{
+	if (self->total_width <= target
+	 || !row_buffer_pop_cells (self, self->total_width - target))
+		return;
+
+	// We use attributes from the last character we've removed,
+	// assuming that we don't shrink the array (and there's no real need)
+	ucs4_t ellipsis = 0x2026; // …
+	if (app_is_character_in_locale (ellipsis))
+	{
+		if (self->total_width >= target)
+			row_buffer_pop_cells (self, 1);
+		if (self->total_width + 1 <= target)
+			row_buffer_append (self, "…",   self->chars[self->chars_len].attrs);
+	}
+	else if (target >= 3)
+	{
+		if (self->total_width >= target)
+			row_buffer_pop_cells (self, 3);
+		if (self->total_width + 3 <= target)
+			row_buffer_append (self, "...", self->chars[self->chars_len].attrs);
+	}
+}
+
+static void
+row_buffer_align (struct row_buffer *self, int target, chtype attrs)
+{
+	row_buffer_ellipsis (self, target);
+	row_buffer_space (self, target - self->total_width, attrs);
+}
+
+static void
+row_buffer_print (uint32_t *ucs4, chtype attrs)
+{
+	// This assumes that we can reset the attribute set without consequences
+	char *str = u32_strconv_to_locale (ucs4);
+	if (str)
+	{
+		attrset (attrs);
+		addstr (str);
+		attrset (0);
+		free (str);
+	}
+}
+
+static void
+row_buffer_flush (struct row_buffer *self)
+{
+	if (!self->chars_len)
+		return;
+
+	// We only NUL-terminate the chunks because of the libunistring API
+	uint32_t chunk[self->chars_len + 1], *insertion_point = chunk;
+	for (size_t i = 0; i < self->chars_len; i++)
+	{
+		struct row_char *iter = self->chars + i;
+		if (i && iter[0].attrs != iter[-1].attrs)
+		{
+			row_buffer_print (chunk, iter[-1].attrs);
+			insertion_point = chunk;
+		}
+		*insertion_point++ = iter->c;
+		*insertion_point = 0;
+	}
+	row_buffer_print (chunk, self->chars[self->chars_len - 1].attrs);
+}

libertyconf.vim

@@ -0,0 +1,26 @@
+" Since the liberty configuration format is nearly indistinguishable,
+" this syntax highlight definition needs to be loaded with `set ft=libertyconf`
+if exists("b:current_syntax")
+	finish
+endif
+
+syn match libertyconfError "[^_[:alnum:][:space:]]\+"
+syn match libertyconfComment "#.*"
+syn match libertyconfSpecial "{\|}\|="
+syn match libertyconfNumber "[+-]\=\<\d\+\>"
+syn match libertyconfBoolean "\c\<\(true\|yes\|on\|false\|no\|off\)\>"
+syn match libertyconfNull "null"
+syn match libertyconfEscape display "\\\([xX]\x\{1,2}\|\o\{1,3}\|.\|$\)"
+	\ contained
+syn region libertyconfString start=+"+ skip=+\\\\\|\\"+ end=+"+
+	\ contains=libertyconfEscape
+
+let b:current_syntax = "libertyconf"
+hi def link libertyconfError Error
+hi def link libertyconfComment Comment
+hi def link libertyconfSpecial Special
+hi def link libertyconfNumber Number
+hi def link libertyconfBoolean Boolean
+hi def link libertyconfNull Constant
+hi def link libertyconfEscape SpecialChar
+hi def link libertyconfString String

libertyxdr.adoc

@@ -0,0 +1,108 @@
+libertyxdr(7)
+=============
+:doctype: manpage
+
+Name
+----
+LibertyXDR - an XDR-derived IDL and data serialization format
+
+Description
+-----------
+*LibertyXDR* is an interface description language, as well as a data
+serialization format, that has been largely derived from XDR, though notably
+simplified.
+
+Conventions
+~~~~~~~~~~~
+User-defined types should be named in *CamelCase*, field names in *snake_case*,
+and constants in *SCREAMING_SNAKE_CASE*.  Code generators will convert these to
+whatever is appropriate in their target language.
+
+Primitive data types
+~~~~~~~~~~~~~~~~~~~~
+Like in XDR, all data is serialized in the network byte order, i.e., big-endian.
+
+ * *void*: 0 bytes
++
+This is a dummy type that cannot be assigned a field name.
+
+ * *bool*: 1 byte 
++
+This is a boolean value: 0 means _false_, any other value means _true_.
+
+ * *u8*, *u16*, *u32*, *u64*: 1, 2, 4, and 8 bytes respectively
++
+These are unsigned integers.
+
+ * *i8*, *i16*, *i32*, *i64*: 1, 2, 4, and 8 bytes respectively
++
+These are signed integers in two's complement.
+
+ * *string*: implicitly prefixed by its length as a *u32*,
+   then immediately followed by its contents, with no trailing NUL byte
++
+This is a valid UTF-8 string without a byte order mark.  Note that strings are
+always unbounded, unlike in XDR.
+
+Constants
+~~~~~~~~~
+At the top level of a document, outside other definitions, you can define
+typeless integer constants:
+
+ const VERSION = 1;
+
+The value can be either a name of another previously defined constant,
+or an immediate decimal value, which may not contain leading zeros.
+
+Enumerations
+~~~~~~~~~~~~
+An *enum* is an *i8* with uniquely named values, in their own namespace.
+
+Values can be either specified explicitly, in the same way as with a constant,
+or they can be left implicit, in which case names assume a value that is one
+larger than their predecessor.  Zero is reserved for internal use, thus
+enumerations implicitly begin with a value of one.  For example, these form
+a sequence from one to three:
+
+ enum Vehicle { CAR, LORRY = 2, PLANE, };
+
+Structures
+~~~~~~~~~~
+A *struct* is a sequence of fields, specified by their type, and their chosen
+name.  You can add a *<>* suffix to change a field to an array, in which case
+it is implicitly preceded by a *u32* specifying its length in terms of its
+elements.
+
+Unlike in XDR, there is no padding between subsequent fields, and type
+definitions can be arbitrarily syntactically nested, as in C.
+
+ struct StockReport {
+   u8 version;       // Version of this report.
+   struct Item {
+     Vehicle kind;   // The vehicle in question.
+     i32 count;      // How many vehicle of that kind there are.
+   } items<>;        // Reported items.
+ };
+
+Unions
+~~~~~~
+A *union* is a kind of structure whose fields depend on the value of its first
+and always-present field, which must be a tag *enum*:
+
+ union VehicleDetails switch (Vehicle kind) {
+ case CAR:   void;
+ case LORRY: i8 axles;
+ case PLANE: i8 engines;
+ };
+
+All possible enumeration values must be named, and there is no *case*
+fall-through.
+
+Framing
+-------
+Unless this role is already filled by, e.g., WebSocket, _LibertyXDR_ structures
+should be prefixed by their byte length in the *u32* format, once serialized.
+
+See also
+--------
+_XDR: External Data Representation Standard_, RFC 4506

libertyxdr.vim

@@ -0,0 +1,21 @@
+" filetype.vim: au! BufNewFile,BufRead *.lxdr setf libertyxdr
+if exists("b:current_syntax")
+	finish
+endif
+
+syn match libertyxdrError "[^[:space:]:;,(){}<>=]\+"
+syn region libertyxdrBlockComment start=+/[*]+ end=+[*]/+
+syn match libertyxdrComment "//.*"
+syn match libertyxdrIdentifier "\<[[:alpha:]][[:alnum:]_]*\>"
+syn match libertyxdrNumber "\<0\>\|\(-\|\<\)[1-9][[:digit:]]*\>"
+syn keyword libertyxdrKeyword const enum struct union switch case
+syn keyword libertyxdrType bool u8 u16 u32 u64 i8 i16 i32 i64 string void
+
+let b:current_syntax = "libertyxdr"
+hi def link libertyxdrError Error
+hi def link libertyxdrBlockComment Comment
+hi def link libertyxdrComment Comment
+hi def link libertyxdrIdentifier Identifier
+hi def link libertyxdrNumber Number
+hi def link libertyxdrKeyword Statement
+hi def link libertyxdrType Type

meson/packaging/make-deb.sh

@@ -0,0 +1,28 @@
+#!/bin/sh -e
+cd "$MESON_BUILD_ROOT"
+. "$MESON_SUBDIR/meta"
+wd="`pwd`/`mktemp -d deb.XXXXXX`"
+trap "rm -rf '$wd'" INT QUIT TERM EXIT
+
+[ "$arch" = x86    ] && arch=i386
+[ "$arch" = x86_64 ] && arch=amd64
+target="$name-$version-$system-$arch.deb"
+
+echo 2.0 > "$wd/debian-binary"
+cat > "$wd/control" <<-EOF
+	Package: $name
+	Version: $version
+	Section: misc
+	Priority: optional
+	Architecture: $arch
+	Maintainer: $author
+	Description: $summary
+EOF
+fakeroot sh -e <<-EOF
+	DESTDIR="$wd/pkg" ninja install
+	cd "$wd/pkg" && tar cJf ../data.tar.xz .
+EOF
+
+(cd "$wd" && tar czf control.tar.gz ./control)
+ar rc "$target" "$wd/debian-binary" "$wd/control.tar.gz" "$wd/data.tar.xz"
+echo Written $target

meson/packaging/make-pacman.sh

@@ -0,0 +1,22 @@
+#!/bin/sh -e
+cd "$MESON_BUILD_ROOT"
+. "$MESON_SUBDIR/meta"
+wd="`pwd`/`mktemp -d pacman.XXXXXX`"
+trap "rm -rf '$wd'" INT QUIT TERM EXIT
+
+target="$name-$version-$arch.tar.xz"
+fakeroot sh -e <<-EOF
+	DESTDIR="$wd" ninja install
+	cat > "$wd/.PKGINFO" <<END
+	pkgname = $name
+	pkgver = $version-1
+	pkgdesc = $summary
+	url = $url
+	builddate = \`date -u +%s\`
+	packager = $author
+	size = \`du -sb | cut -f1\`
+	arch = $arch
+	END
+	cd "$wd" && tar cJf "../$target" .PKGINFO *
+	echo Written $target
+EOF

meson/packaging/meson.build

@@ -0,0 +1,11 @@
+# You need to prepare a configuration object with the required metadata
+packaging.set ('arch', target_machine.cpu_family ())
+packaging.set ('system', target_machine.system ())
+configure_file (input: 'meta.in', output: 'meta', configuration: packaging)
+
+# RPM is awful and I've given up on both manual generation (we'd have to either
+# include rpmrc data or generate fake noarch packages) and rpmbuild (just no)
+run_target ('deb',
+	command: [join_paths (meson.current_source_dir (), 'make-deb.sh')])
+run_target ('pacman',
+	command: [join_paths (meson.current_source_dir (), 'make-pacman.sh')])

meson/packaging/meta.in

@@ -0,0 +1,8 @@
+define() { [ -z "$2" ] && { echo $1 is undefined; exit 1; } || eval "$1='$2'"; }
+
+define name    "@name@"
+define version "@version@"
+define summary "@summary@"
+define author  "@author@"
+define arch    "@arch@"
+define system  "@system@"

siphash.c

@@ -61,13 +61,13 @@ siphash (const unsigned char key[16], const unsigned char *m, size_t len)
 
 	switch (len - blocks)
 	{
-	case 7: last7 |= (uint64_t) m[i + 6] << 48;
-	case 6: last7 |= (uint64_t) m[i + 5] << 40;
-	case 5: last7 |= (uint64_t) m[i + 4] << 32;
-	case 4: last7 |= (uint64_t) m[i + 3] << 24;
-	case 3: last7 |= (uint64_t) m[i + 2] << 16;
-	case 2: last7 |= (uint64_t) m[i + 1] <<  8;
-	case 1: last7 |= (uint64_t) m[i + 0]      ;
+	case 7: last7 |= (uint64_t) m[i + 6] << 48; // Fall-through
+	case 6: last7 |= (uint64_t) m[i + 5] << 40; // Fall-through
+	case 5: last7 |= (uint64_t) m[i + 4] << 32; // Fall-through
+	case 4: last7 |= (uint64_t) m[i + 3] << 24; // Fall-through
+	case 3: last7 |= (uint64_t) m[i + 2] << 16; // Fall-through
+	case 2: last7 |= (uint64_t) m[i + 1] <<  8; // Fall-through
+	case 1: last7 |= (uint64_t) m[i + 0]      ; // Fall-through
 	default:;
 	};
 	v3 ^= last7;
@@ -82,4 +82,3 @@ siphash (const unsigned char key[16], const unsigned char *m, size_t len)
 
 	return v0 ^ v1 ^ v2 ^ v3;
 }
-

tests/fuzz.c

@@ -0,0 +1,297 @@
+/*
+ * tests/fuzz.c
+ *
+ * Copyright (c) 2020, Přemysl Eric Janouch <p@janouch.name>
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ */
+
+#define PROGRAM_NAME "fuzz"
+#define PROGRAM_VERSION "0"
+
+#define LIBERTY_WANT_SSL
+// The MPD client is a full wrapper and needs the network
+#define LIBERTY_WANT_POLLER
+#define LIBERTY_WANT_ASYNC
+
+#define LIBERTY_WANT_PROTO_IRC
+#define LIBERTY_WANT_PROTO_HTTP
+#define LIBERTY_WANT_PROTO_SCGI
+#define LIBERTY_WANT_PROTO_FASTCGI
+#define LIBERTY_WANT_PROTO_WS
+#define LIBERTY_WANT_PROTO_MPD
+
+#include "../liberty.c"
+#include "../liberty-tui.c"
+
+static bool
+app_is_character_in_locale (ucs4_t ch)
+{
+	return ch < 128;
+}
+
+// --- UTF-8 -------------------------------------------------------------------
+
+static void
+test_utf8_validate (const uint8_t *data, size_t size)
+{
+	utf8_validate ((const char *) data, size);
+}
+
+// --- Base 64 -----------------------------------------------------------------
+
+static void
+test_base64_decode (const uint8_t *data, size_t size)
+{
+	struct str wrap = str_make ();
+	str_append_data (&wrap, data, size);
+
+	struct str out = str_make ();
+	base64_decode (wrap.str, true /* ignore_ws */, &out);
+	str_free (&out);
+
+	str_free (&wrap);
+}
+
+// --- IRC ---------------------------------------------------------------------
+
+static void
+test_irc_parse_message (const uint8_t *data, size_t size)
+{
+	struct str wrap = str_make ();
+	str_append_data (&wrap, data, size);
+
+	struct irc_message msg;
+	irc_parse_message (&msg, wrap.str);
+	irc_free_message (&msg);
+
+	str_free (&wrap);
+}
+
+// --- HTTP --------------------------------------------------------------------
+
+static void
+test_http_parse_media_type (const uint8_t *data, size_t size)
+{
+	struct str wrap = str_make ();
+	str_append_data (&wrap, data, size);
+
+	char *type = NULL;
+	char *subtype = NULL;
+	struct str_map parameters = str_map_make (free);
+	http_parse_media_type (wrap.str, &type, &subtype, &parameters);
+	free (type);
+	free (subtype);
+	str_map_free (&parameters);
+
+	str_free (&wrap);
+}
+
+static void
+test_http_parse_upgrade (const uint8_t *data, size_t size)
+{
+	struct str wrap = str_make ();
+	str_append_data (&wrap, data, size);
+
+	struct http_protocol *protocols = NULL;
+	http_parse_upgrade (wrap.str, &protocols);
+	LIST_FOR_EACH (struct http_protocol, iter, protocols)
+		http_protocol_destroy (iter);
+
+	str_free (&wrap);
+}
+
+// --- SCGI --------------------------------------------------------------------
+
+static bool
+test_scgi_parser_on_headers_read (void *user_data)
+{
+	(void) user_data;
+	return true;
+}
+
+static bool
+test_scgi_parser_on_content (void *user_data, const void *data, size_t len)
+{
+	(void) user_data;
+	(void) data;
+	(void) len;
+	return true;
+}
+
+static void
+test_scgi_parser_push (const uint8_t *data, size_t size)
+{
+	struct scgi_parser parser = scgi_parser_make ();
+	parser.on_headers_read = test_scgi_parser_on_headers_read;
+	parser.on_content      = test_scgi_parser_on_content;
+
+	scgi_parser_push (&parser, data, size, NULL);
+	scgi_parser_free (&parser);
+}
+
+// --- WebSockets --------------------------------------------------------------
+
+static bool
+test_ws_parser_on_frame_header (void *user_data, const struct ws_parser *self)
+{
+	(void) user_data;
+	(void) self;
+	return true;
+}
+
+static bool
+test_ws_parser_on_frame (void *user_data, const struct ws_parser *self)
+{
+	(void) user_data;
+	(void) self;
+	return true;
+}
+
+static void
+test_ws_parser_push (const uint8_t *data, size_t size)
+{
+	struct ws_parser parser = ws_parser_make ();
+	parser.on_frame_header = test_ws_parser_on_frame_header;
+	parser.on_frame        = test_ws_parser_on_frame;
+
+	ws_parser_push (&parser, data, size);
+	ws_parser_free (&parser);
+}
+
+// --- FastCGI -----------------------------------------------------------------
+
+static bool
+test_fcgi_parser_on_message (const struct fcgi_parser *parser, void *user_data)
+{
+	(void) parser;
+	(void) user_data;
+	return true;
+}
+
+static void
+test_fcgi_parser_push (const uint8_t *data, size_t size)
+{
+	struct fcgi_parser parser = fcgi_parser_make ();
+	parser.on_message = test_fcgi_parser_on_message;
+	fcgi_parser_push (&parser, data, size);
+	fcgi_parser_free (&parser);
+}
+
+static void
+test_fcgi_nv_parser_push (const uint8_t *data, size_t size)
+{
+	struct str_map values = str_map_make (free);
+	struct fcgi_nv_parser nv_parser = fcgi_nv_parser_make ();
+	nv_parser.output = &values;
+
+	fcgi_nv_parser_push (&nv_parser, data, size);
+	fcgi_nv_parser_free (&nv_parser);
+	str_map_free (&values);
+}
+
+// --- Config ------------------------------------------------------------------
+
+static void
+test_config_item_parse (const uint8_t *data, size_t size)
+{
+	struct config_item *item =
+		config_item_parse ((const char *) data, size, false, NULL);
+	if (item)
+		config_item_destroy (item);
+}
+
+// --- TUI ---------------------------------------------------------------------
+
+static void
+test_attrs_decode (const uint8_t *data, size_t size)
+{
+	struct str wrap = str_make ();
+	str_append_data (&wrap, data, size);
+
+	attrs_decode (wrap.str);
+
+	str_free (&wrap);
+}
+
+// --- MPD ---------------------------------------------------------------------
+
+static void
+test_mpd_client_process_input (const uint8_t *data, size_t size)
+{
+	struct poller poller;
+	poller_init (&poller);
+
+	struct mpd_client mpd = mpd_client_make (&poller);
+	str_append_data (&mpd.read_buffer, data, size);
+	mpd_client_process_input (&mpd);
+	mpd_client_free (&mpd);
+
+	poller_free (&poller);
+}
+
+// --- Main --------------------------------------------------------------------
+
+typedef void (*fuzz_test_fn) (const uint8_t *data, size_t size);
+static fuzz_test_fn generator = NULL;
+
+void
+LLVMFuzzerTestOneInput (const uint8_t *data, size_t size)
+{
+	generator (data, size);
+}
+
+int
+LLVMFuzzerInitialize (int *argcp, char ***argvp)
+{
+	struct str_map targets = str_map_make (NULL);
+#define REGISTER(name) str_map_set (&targets, #name, test_ ## name);
+	REGISTER (utf8_validate)
+	REGISTER (base64_decode)
+	REGISTER (irc_parse_message)
+	REGISTER (http_parse_media_type)
+	REGISTER (http_parse_upgrade)
+	REGISTER (scgi_parser_push)
+	REGISTER (ws_parser_push)
+	REGISTER (fcgi_parser_push)
+	REGISTER (fcgi_nv_parser_push)
+	REGISTER (config_item_parse)
+	REGISTER (attrs_decode)
+	REGISTER (mpd_client_process_input)
+
+	char **argv = *argvp, *option = "-test=", *name = NULL;
+	for (int i = 1; i < *argcp; i++)
+		if (!strncmp (argv[i], option, strlen (option)))
+		{
+			name = argv[i] + strlen (option);
+			memmove (argv + i, argv + i + 1, (*argcp - i) * sizeof *argv);
+			(*argcp)--;
+		}
+
+	if (!name)
+	{
+		struct str_map_iter iter = str_map_iter_make (&targets);
+		while (str_map_iter_next (&iter))
+			printf ("%s\n", iter.link->key);
+		exit (EXIT_FAILURE);
+	}
+
+	if (!(generator = str_map_find (&targets, name)))
+	{
+		fprintf (stderr, "Unknown test: %s\n", name);
+		exit (EXIT_FAILURE);
+	}
+
+	str_map_free (&targets);
+	return 0;
+}

tests/liberty.c

@@ -1,12 +1,10 @@
 /*
  * tests/liberty.c
  *
- * Copyright (c) 2015 - 2016, Přemysl Janouch <p.janouch@gmail.com>
- * All rights reserved.
+ * Copyright (c) 2015 - 2022, Přemysl Eric Janouch <p@janouch.name>
  *
  * Permission to use, copy, modify, and/or distribute this software for any
- * purpose with or without fee is hereby granted, provided that the above
- * copyright notice and this permission notice appear in all copies.
+ * purpose with or without fee is hereby granted.
  *
  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
@@ -157,31 +155,28 @@ test_list_with_tail (void)
 // --- Strings -----------------------------------------------------------------
 
 static void
-test_str_vector (void)
+test_strv (void)
 {
-	struct str_vector v;
-	str_vector_init (&v);
-
-	str_vector_add_owned (&v, xstrdup ("xkcd"));
-	str_vector_reset (&v);
+	struct strv v = strv_make ();
+	strv_append_owned (&v, xstrdup ("xkcd"));
+	strv_reset (&v);
 
 	const char *a[] =
 		{ "123", "456", "a", "bc", "def", "ghij", "klmno", "pqrstu" };
 
 	// Add the first two items via another vector
-	struct str_vector w;
-	str_vector_init (&w);
-	str_vector_add_args (&w, a[0], a[1], NULL);
-	str_vector_add_vector (&v, w.vector);
-	str_vector_free (&w);
+	struct strv w = strv_make ();
+	strv_append_args (&w, a[0], a[1], NULL);
+	strv_append_vector (&v, w.vector);
+	strv_free (&w);
 
 	// Add an item and delete it right after
-	str_vector_add (&v, "test");
-	str_vector_remove (&v, v.len - 1);
+	strv_append (&v, "test");
+	strv_remove (&v, v.len - 1);
 
 	// Add the rest of the list properly
 	for (int i = 2; i < (int) N_ELEMENTS (a); i++)
-		str_vector_add (&v, a[i]);
+		strv_append (&v, a[i]);
 
 	// Check the contents
 	soft_assert (v.len == N_ELEMENTS (a));
@@ -189,7 +184,7 @@ test_str_vector (void)
 		soft_assert (!strcmp (v.vector[i], a[i]));
 	soft_assert (v.vector[v.len] == NULL);
 
-	str_vector_free (&v);
+	strv_free (&v);
 }
 
 static void
@@ -197,15 +192,13 @@ test_str (void)
 {
 	uint8_t x[] = { 0x12, 0x34, 0x56, 0x78, 0x11, 0x22, 0x33, 0x44 };
 
-	struct str s;
-	str_init (&s);
-	str_ensure_space (&s, MEGA);
+	struct str s = str_make ();
+	str_reserve (&s, MEGA);
 	str_append_data (&s, x, sizeof x);
 	str_remove_slice (&s, 4, 4);
 	soft_assert (s.len == 4);
 
-	struct str t;
-	str_init (&t);
+	struct str t = str_make ();
 	str_append_str (&t, &s);
 	str_append (&t, "abc");
 	str_append_c (&t, 'd');
@@ -266,10 +259,8 @@ static void
 test_str_map (void)
 {
 	// Put two reference counted objects in the map under case-insensitive keys
-	struct str_map m;
-	str_map_init (&m);
+	struct str_map m = str_map_make (free_counter);
 	m.key_xfrm = tolower_ascii_strxfrm;
-	m.free = free_counter;
 
 	int *a = make_counter ();
 	int *b = make_counter ();
@@ -283,8 +274,7 @@ test_str_map (void)
 	soft_assert (str_map_find (&m, "DEFghi") == b);
 
 	// Check that we can iterate over both of them
-	struct str_map_iter iter;
-	str_map_iter_init (&iter, &m);
+	struct str_map_iter iter = str_map_iter_make (&m);
 
 	bool met_a = false;
 	bool met_b = false;
@@ -311,8 +301,7 @@ test_str_map (void)
 	free_counter (b);
 
 	// Iterator test with a high number of items
-	str_map_init (&m);
-	m.free = free;
+	m = str_map_make (free);
 
 	for (size_t i = 0; i < 100 * 100; i++)
 	{
@@ -320,8 +309,7 @@ test_str_map (void)
 		str_map_set (&m, x, x);
 	}
 
-	struct str_map_unset_iter unset_iter;
-	str_map_unset_iter_init (&unset_iter, &m);
+	struct str_map_unset_iter unset_iter = str_map_unset_iter_make (&m);
 	while ((str_map_unset_iter_next (&unset_iter)))
 	{
 		unsigned long x;
@@ -338,14 +326,21 @@ test_str_map (void)
 static void
 test_utf8 (void)
 {
-	const char valid  [] = "2H₂ + O₂ ⇌ 2H₂O, R = 4.7 kΩ, ⌀ 200 mm";
-	const char invalid[] = "\xf0\x90\x28\xbc";
-	soft_assert ( utf8_validate (valid,   sizeof valid));
-	soft_assert (!utf8_validate (invalid, sizeof invalid));
+	const char *full = "\xc5\x99", *partial = full, *empty = full;
+	soft_assert (utf8_decode (&full,    2) == 0x0159);
+	soft_assert (utf8_decode (&partial, 1) == -2);
+	soft_assert (utf8_decode (&empty,   0) == -1);
 
-	struct utf8_iter iter;
-	utf8_iter_init (&iter, "fóọ");
+	const char valid_1[] = "2H₂ + O₂ ⇌ 2H₂O, R = 4.7 kΩ, ⌀ 200 mm";
+	const char valid_2[] = "\xf0\x93\x82\xb9";
+	const char invalid_1[] = "\xf0\x90\x28\xbc";
+	const char invalid_2[] = "\xc0\x80";
+	soft_assert ( utf8_validate (valid_1,   sizeof valid_1));
+	soft_assert ( utf8_validate (valid_2,   sizeof valid_2));
+	soft_assert (!utf8_validate (invalid_1, sizeof invalid_1));
+	soft_assert (!utf8_validate (invalid_2, sizeof invalid_2));
 
+	struct utf8_iter iter = utf8_iter_make ("fóọ");
 	size_t ch_len;
 	hard_assert (utf8_iter_next (&iter, &ch_len) == 'f'    && ch_len == 1);
 	hard_assert (utf8_iter_next (&iter, &ch_len) == 0x00F3 && ch_len == 2);
@@ -359,8 +354,8 @@ test_base64 (void)
 	for (size_t i = 0; i < N_ELEMENTS (data); i++)
 		data[i] = i;
 
-	struct str encoded;  str_init (&encoded);
-	struct str decoded;  str_init (&decoded);
+	struct str encoded = str_make ();
+	struct str decoded = str_make ();
 
 	base64_encode (data, sizeof data, &encoded);
 	soft_assert (base64_decode (encoded.str, false, &decoded));
@@ -431,9 +426,9 @@ test_async (void)
 {
 	struct test_async_data data;
 	memset (&data, 0, sizeof data);
-	async_manager_init (&data.manager);
+	data.manager = async_manager_make ();
 
-	async_init (&data.busyloop, &data.manager);
+	data.busyloop = async_make (&data.manager);
 	data.busyloop.execute = on_busyloop_execute;
 	data.busyloop.destroy = on_busyloop_destroy;
 	async_run (&data.busyloop);
@@ -543,7 +538,7 @@ test_connector_fixture_init
 
 	// Make it so that we immediately accept all connections
 	poller_init (&self->poller);
-	poller_fd_init (&self->listening_event, &self->poller, self->listening_fd);
+	self->listening_event = poller_fd_make (&self->poller, self->listening_fd);
 	self->listening_event.dispatcher = test_connector_on_client;
 	self->listening_event.user_data = (poller_fd_fn) self;
 	poller_fd_set (&self->listening_event, POLLIN);
@@ -618,6 +613,81 @@ test_connector (const void *user_data, struct test_connector_fixture *self)
 	connector_free (&connector);
 }
 
+// --- Configuration -----------------------------------------------------------
+
+static void
+on_test_config_foo_change (struct config_item *item)
+{
+	*(bool *) item->user_data = item->value.boolean;
+}
+
+static bool
+test_config_validate_nonnegative
+	(const struct config_item *item, struct error **e)
+{
+	if (item->type == CONFIG_ITEM_NULL)
+		return true;
+
+	hard_assert (item->type == CONFIG_ITEM_INTEGER);
+	if (item->value.integer >= 0)
+		return true;
+
+	error_set (e, "must be non-negative");
+	return false;
+}
+
+static struct config_schema g_config_test[] =
+{
+	{ .name      = "foo",
+	  .comment   = "baz",
+	  .type      = CONFIG_ITEM_BOOLEAN,
+	  .default_  = "off",
+	  .on_change = on_test_config_foo_change },
+	{ .name      = "bar",
+	  .type      = CONFIG_ITEM_INTEGER,
+	  .validate  = test_config_validate_nonnegative,
+	  .default_  = "1" },
+	{ .name      = "foobar",
+	  .type      = CONFIG_ITEM_STRING,
+	  .default_  = "\"qux\\x01`\" \"\"`a`" },
+	{}
+};
+
+static void
+test_config_load (struct config_item *subtree, void *user_data)
+{
+	config_schema_apply_to_object (g_config_test, subtree, user_data);
+}
+
+static void
+test_config (void)
+{
+	struct config config = config_make ();
+
+	bool b = true;
+	config_register_module (&config, "top", test_config_load, &b);
+	config_load (&config, config_item_object ());
+	config_schema_call_changed (config.root);
+	hard_assert (b == false);
+
+	struct config_item *invalid = config_item_integer (-1);
+	hard_assert (!config_item_set_from (config_item_get (config.root,
+		"top.bar", NULL), invalid, NULL));
+	config_item_destroy (invalid);
+
+	hard_assert (!strcmp ("qux\001`a",
+		config_item_get (config.root, "top.foobar", NULL)->value.string.str));
+
+	struct str s = str_make ();
+	config_item_write (config.root, true, &s);
+	struct config_item *parsed = config_item_parse (s.str, s.len, false, NULL);
+	hard_assert (parsed);
+	config_item_destroy (parsed);
+	str_free (&s);
+
+	config_free (&config);
+}
+
 // --- Main --------------------------------------------------------------------
 
 int
@@ -629,13 +699,14 @@ main (int argc, char *argv[])
 	test_add_simple (&test, "/memory",         NULL, test_memory);
 	test_add_simple (&test, "/list",           NULL, test_list);
 	test_add_simple (&test, "/list-with-tail", NULL, test_list_with_tail);
-	test_add_simple (&test, "/str-vector",     NULL, test_str_vector);
+	test_add_simple (&test, "/strv",           NULL, test_strv);
 	test_add_simple (&test, "/str",            NULL, test_str);
 	test_add_simple (&test, "/error",          NULL, test_error);
 	test_add_simple (&test, "/str-map",        NULL, test_str_map);
 	test_add_simple (&test, "/utf-8",          NULL, test_utf8);
 	test_add_simple (&test, "/base64",         NULL, test_base64);
 	test_add_simple (&test, "/async",          NULL, test_async);
+	test_add_simple (&test, "/config",         NULL, test_config);
 
 	test_add (&test, "/connector", struct test_connector_fixture, NULL,
 		test_connector_fixture_init,

tests/lxdrgen.c

@@ -0,0 +1,123 @@
+/*
+ * tests/lxdrgen.c
+ *
+ * Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ */
+
+#define PROGRAM_NAME "test"
+#define PROGRAM_VERSION "0"
+
+#include "../liberty.c"
+#include "lxdrgen.lxdr.c"
+
+static void
+test_ser_deser_free (void)
+{
+	hard_assert (PROTO_GEN_VERSION == 1);
+
+	enum { CASES = 3 };
+
+	struct proto_gen_struct a = {}, b = {};
+	a.u = xcalloc ((a.u_len = CASES + rand () % 100), sizeof *a.u);
+	for (size_t i = 0; i < a.u_len; i++)
+	{
+		union proto_gen_union *u = a.u + i;
+		switch (i % CASES)
+		{
+		case 0:
+			u->tag = PROTO_GEN_ENUM_NUMBERS;
+			u->numbers.a = rand () % UINT8_MAX;
+			u->numbers.b = rand () % UINT16_MAX;
+			u->numbers.c = rand () % UINT32_MAX;
+			u->numbers.d = rand () % UINT64_MAX;
+			u->numbers.e = rand () % UINT8_MAX;
+			u->numbers.f = rand () % UINT16_MAX;
+			u->numbers.g = rand () % UINT32_MAX;
+			u->numbers.h = rand () % UINT64_MAX;
+			break;
+		case 1:
+			u->tag = PROTO_GEN_ENUM_OTHERS;
+			u->others.foo = rand () % 2;
+			u->others.bar = str_make ();
+			for (int i = rand () % 0x30; i > 0; i--)
+				str_append_c (&u->others.bar, 0x30 + i);
+			break;
+		case 2:
+			u->tag = PROTO_GEN_ENUM_NOTHING;
+			break;
+		default:
+			hard_assert (!"unhandled case");
+		}
+	}
+
+	struct str buf = str_make ();
+	hard_assert (proto_gen_struct_serialize (&a, &buf));
+	struct msg_unpacker r = msg_unpacker_make (buf.str, buf.len);
+	hard_assert (proto_gen_struct_deserialize (&b, &r));
+	hard_assert (!msg_unpacker_get_available (&r));
+	str_free (&buf);
+
+	hard_assert (a.u_len == b.u_len);
+	for (size_t i = 0; i < a.u_len; i++)
+	{
+		union proto_gen_union *ua = a.u + i;
+		union proto_gen_union *ub = b.u + i;
+		hard_assert (ua->tag == ub->tag);
+		switch (ua->tag)
+		{
+		case PROTO_GEN_ENUM_NUMBERS:
+			hard_assert (ua->numbers.a == ub->numbers.a);
+			hard_assert (ua->numbers.b == ub->numbers.b);
+			hard_assert (ua->numbers.c == ub->numbers.c);
+			hard_assert (ua->numbers.d == ub->numbers.d);
+			hard_assert (ua->numbers.e == ub->numbers.e);
+			hard_assert (ua->numbers.f == ub->numbers.f);
+			hard_assert (ua->numbers.g == ub->numbers.g);
+			hard_assert (ua->numbers.h == ub->numbers.h);
+			break;
+		case PROTO_GEN_ENUM_OTHERS:
+			hard_assert (ua->others.foo == ub->others.foo);
+			hard_assert (ua->others.bar.len == ub->others.bar.len);
+			hard_assert (!memcmp (ua->others.bar.str, ub->others.bar.str,
+				ua->others.bar.len));
+			break;
+		case PROTO_GEN_ENUM_NOTHING:
+			break;
+		default:
+			hard_assert (!"unexpected case");
+		}
+	}
+
+	// Emulate partially deserialized data to test disposal of that.
+	for (size_t i = b.u_len - CASES; i < b.u_len; i++)
+	{
+		proto_gen_union_free (&b.u[i]);
+		memset (&b.u[i], 0, sizeof b.u[i]);
+	}
+
+	proto_gen_struct_free (&a);
+	proto_gen_struct_free (&b);
+}
+
+int
+main (int argc, char *argv[])
+{
+	struct test test;
+	test_init (&test, argc, argv);
+
+	test_add_simple (&test, "/ser-deser-free", NULL, test_ser_deser_free);
+
+	return test_run (&test);
+}

tests/lxdrgen.lxdr

@@ -0,0 +1,23 @@
+/*
+ * tests/lxdrgen.lxdr: a test protocol for the generator
+ */
+const VERSION = 1;
+const NOISREV = -1;
+
+// TODO: Test failure paths, and in general go for full coverage.
+struct Struct {
+	union Union switch (enum Enum {
+		NUMBERS = VERSION,
+		OTHERS = 2,
+		NOTHING,
+	} tag) {
+	case NUMBERS:
+		i8 a; i16 b; i32 c; i64 d;
+		u8 e; u16 f; u32 g; u64 h;
+	case OTHERS:
+		bool foo;
+		string bar;
+	case NOTHING:
+		void;
+	} u<>;
+};

tests/proto.c

@@ -1,12 +1,10 @@
 /*
  * tests/proto.c
  *
- * Copyright (c) 2015, Přemysl Janouch <p.janouch@gmail.com>
- * All rights reserved.
+ * Copyright (c) 2015, Přemysl Eric Janouch <p@janouch.name>
  *
  * Permission to use, copy, modify, and/or distribute this software for any
- * purpose with or without fee is hereby granted, provided that the above
- * copyright notice and this permission notice appear in all copies.
+ * purpose with or without fee is hereby granted.
  *
  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
@@ -44,8 +42,7 @@ test_irc (void)
 	irc_parse_message (&msg, "@first=a\\:\\s\\r\\n\\\\;2nd "
 		":srv hi there :good m8 :how are you?");
 
-	struct str_map_iter iter;
-	str_map_iter_init (&iter, &msg.tags);
+	struct str_map_iter iter = str_map_iter_make (&msg.tags);
 	soft_assert (msg.tags.len == 2);
 
 	char *value;
@@ -80,8 +77,7 @@ test_irc (void)
 static void
 test_http_parser (void)
 {
-	struct str_map parameters;
-	str_map_init (&parameters);
+	struct str_map parameters = str_map_make (free);
 	parameters.key_xfrm = tolower_ascii_strxfrm;
 
 	char *type = NULL;
@@ -92,9 +88,11 @@ test_http_parser (void)
 	soft_assert (!strcasecmp_ascii (subtype, "html"));
 	soft_assert (parameters.len == 1);
 	soft_assert (!strcmp (str_map_find (&parameters, "charset"), "utf-8"));
+	free (type);
+	free (subtype);
 	str_map_free (&parameters);
 
-	struct http_protocol *protocols;
+	struct http_protocol *protocols = NULL;
 	soft_assert (http_parse_upgrade ("websocket, HTTP/2.0, , ", &protocols));
 
 	soft_assert (!strcmp (protocols->name, "websocket"));
@@ -109,10 +107,20 @@ test_http_parser (void)
 		http_protocol_destroy (iter);
 }
 
+struct scgi_fixture
+{
+	struct scgi_parser parser;
+	bool seen_headers;
+	bool seen_content;
+};
+
 static bool
 test_scgi_parser_on_headers_read (void *user_data)
 {
-	struct scgi_parser *parser = user_data;
+	struct scgi_fixture *fixture = user_data;
+	struct scgi_parser *parser = &fixture->parser;
+	fixture->seen_headers = true;
+
 	soft_assert (parser->headers.len == 4);
 	soft_assert (!strcmp (str_map_find (&parser->headers,
 		"CONTENT_LENGTH"), "27"));
@@ -128,7 +136,9 @@ test_scgi_parser_on_headers_read (void *user_data)
 static bool
 test_scgi_parser_on_content (void *user_data, const void *data, size_t len)
 {
-	(void) user_data;
+	struct scgi_fixture *fixture = user_data;
+	fixture->seen_content = true;
+
 	soft_assert (!strncmp (data, "What is the answer to life?", len));
 	return true;
 }
@@ -136,11 +146,12 @@ test_scgi_parser_on_content (void *user_data, const void *data, size_t len)
 static void
 test_scgi_parser (void)
 {
-	struct scgi_parser parser;
-	scgi_parser_init (&parser);
-	parser.on_headers_read = test_scgi_parser_on_headers_read;
-	parser.on_content      = test_scgi_parser_on_content;
-	parser.user_data       = &parser;
+	struct scgi_fixture fixture = { scgi_parser_make(), false, false };
+	struct scgi_parser *parser = &fixture.parser;
+
+	parser->on_headers_read = test_scgi_parser_on_headers_read;
+	parser->on_content      = test_scgi_parser_on_content;
+	parser->user_data       = &fixture;
 
 	// This is an example straight from the specification
 	const char example[] =
@@ -152,8 +163,9 @@ test_scgi_parser (void)
 		","
 		"What is the answer to life?";
 
-	soft_assert (scgi_parser_push (&parser, example, sizeof example, NULL));
-	scgi_parser_free (&parser);
+	soft_assert (scgi_parser_push (parser, example, sizeof example, NULL));
+	soft_assert (fixture.seen_headers && fixture.seen_content);
+	scgi_parser_free (parser);
 }
 
 static bool
@@ -182,8 +194,7 @@ test_websockets (void)
 	soft_assert (!strcmp (accept, "s3pPLMBiTxaQ9kYGzzhZRbK+xOo="));
 	free (accept);
 
-	struct ws_parser parser;
-	ws_parser_init (&parser);
+	struct ws_parser parser = ws_parser_make ();
 	parser.on_frame_header = test_websockets_on_frame_header;
 	parser.on_frame        = test_websockets_on_frame;
 	parser.user_data       = &parser;

tests/pulse.c

@@ -0,0 +1,127 @@
+/*
+ * tests/pulse.c
+ *
+ * Copyright (c) 2021, Přemysl Eric Janouch <p@janouch.name>
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ */
+
+#define PROGRAM_NAME "test"
+#define PROGRAM_VERSION "0"
+
+#define LIBERTY_WANT_POLLER
+
+#include "../liberty.c"
+#include "../liberty-pulse.c"
+
+// --- Tests -------------------------------------------------------------------
+
+enum
+{
+	EVENT_IO    =  1 << 0,
+	EVENT_TIME  =  1 << 1,
+	EVENT_DEFER =  1 << 2,
+	EVENT_ALL   = (1 << 3) - 1
+};
+
+static intptr_t g_events = 0;
+static intptr_t g_destroys = 0;
+
+static void
+io_event_cb (pa_mainloop_api *a,
+	pa_io_event *e, int fd, pa_io_event_flags_t events, void *userdata)
+{
+	(void) a; (void) e; (void) fd; (void) events;
+	g_events |= (intptr_t) userdata;
+}
+
+static void
+io_event_destroy_cb (pa_mainloop_api *a, pa_io_event *e, void *userdata)
+{
+	(void) a; (void) e;
+	g_destroys += (intptr_t) userdata;
+}
+
+static void
+time_event_cb (pa_mainloop_api *a,
+	pa_time_event *e, const struct timeval *tv, void *userdata)
+{
+	(void) a; (void) e; (void) tv;
+	g_events |= (intptr_t) userdata;
+}
+
+static void
+time_event_destroy_cb (pa_mainloop_api *a, pa_time_event *e, void *userdata)
+{
+	(void) a; (void) e;
+	g_destroys += (intptr_t) userdata;
+}
+
+static void
+defer_event_cb (pa_mainloop_api *a, pa_defer_event *e, void *userdata)
+{
+	(void) a; (void) e;
+	g_events |= (intptr_t) userdata;
+}
+
+static void
+defer_event_destroy_cb (pa_mainloop_api *a, pa_defer_event *e, void *userdata)
+{
+	(void) a; (void) e;
+	g_destroys += (intptr_t) userdata;
+}
+
+static void
+test_pulse (void)
+{
+	struct poller poller;
+	poller_init (&poller);
+
+	// Let's just get this over with, not aiming for high test coverage here
+	pa_mainloop_api *api = poller_pa_new (&poller);
+
+	pa_io_event *ie = api->io_new (api, STDOUT_FILENO, PA_IO_EVENT_OUTPUT,
+		io_event_cb, (void *) EVENT_IO);
+	api->io_set_destroy (ie, io_event_destroy_cb);
+
+	const struct timeval tv = poller_pa_get_current_time ();
+	pa_time_event *te = api->time_new (api, &tv,
+		time_event_cb, (void *) EVENT_TIME);
+	api->time_set_destroy (te, time_event_destroy_cb);
+	api->time_restart (te, &tv);
+
+	pa_defer_event *de = api->defer_new (api,
+		defer_event_cb, (void *) EVENT_DEFER);
+	api->defer_set_destroy (de, defer_event_destroy_cb);
+	api->defer_enable (api->defer_new (api,
+		defer_event_cb, (void *) EVENT_DEFER), false);
+
+	alarm (1);
+	while (g_events != EVENT_ALL)
+		poller_run (&poller);
+
+	poller_pa_destroy (api);
+	soft_assert (g_destroys == EVENT_ALL);
+	poller_free (&poller);
+}
+
+// --- Main --------------------------------------------------------------------
+
+int
+main (int argc, char *argv[])
+{
+	struct test test;
+	test_init (&test, argc, argv);
+	test_add_simple (&test, "/pulse", NULL, test_pulse);
+	return test_run (&test);
+}

tools/asciiman.awk

@@ -0,0 +1,245 @@
+# asciiman.awk: stupid AsciiDoc to manual page converter
+#
+# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+#
+# This is not intended to produce great output, merely useful output.
+# As such, input documents should restrict themselves as follows:
+#
+#  - In-line formatting sequences must not overlap,
+#    cannot be escaped, and cannot span lines.
+#  - Heading underlines must match in byte length exactly.
+#  - Only a small subset of syntax is supported overall.
+#
+# Also beware that the output has only been tested with GNU troff.
+# Attributes can be passed via environment variables starting with "asciidoc-".
+
+function fatal(message) {
+	print ".\\\" " FILENAME ":" FNR ": fatal error: " message
+	print FILENAME ":" FNR ": fatal error: " message > "/dev/stderr"
+	exit 1
+}
+
+function getattribute(name) {
+	if (!(name in Attrs) && ("asciidoc-" name) in ENVIRON)
+		Attrs[name] = ENVIRON["asciidoc-" name]
+	return Attrs[name]
+}
+
+function expand(s,   attr, v) {
+	# TODO: This should not expand unknown attribute names.
+	while (match(s, /[{][^{}]*[}]/)) {
+		s = substr(s, 1, RSTART - 1) \
+			getattribute(substr(s, RSTART + 1, RLENGTH - 2)) \
+			substr(s, RSTART + RLENGTH)
+	}
+	return s
+}
+
+function escape(s) {
+	gsub(/\\/, "\\\\", s)
+	gsub(/-/, "\\-", s)
+	sub(/^[.']/, "\\\\\\&&", s)
+	return s
+}
+
+function readattribute(line,    attrname, attrvalue) {
+	if (match(line, /^:[^:]*: /)) {
+		attrname = substr(line, RSTART + 1, RLENGTH - 3)
+		attrvalue = substr(line, RSTART + RLENGTH)
+		Attrs[attrname] = expand(attrvalue)
+		return 1
+	}
+}
+
+NR == 1 {
+	nameline = $0
+	if (match(nameline, /[(][[:digit:]][)]$/)) {
+		name = substr(nameline, 1, RSTART - 1)
+		section = substr(nameline, RSTART + 1, RLENGTH - 2)
+	} else {
+		fatal("invalid header line")
+	}
+
+	getline
+	if (length(nameline) != length($0) || /[^=]/)
+		fatal("invalid header underline")
+
+	getline
+	while (readattribute($0))
+		getline
+	if ($0)
+		fatal("expected an empty line after the header")
+
+	# Requesting tbl(1), even though we currently do not support tables.
+	print "'\\\" t"
+	printf ".TH \"%s\" \"%s\" \"\" \"%s\"",
+		toupper(name), section, getattribute("mansource")
+	if (getattribute("manmanual"))
+		printf " \"%s\"", getattribute("manmanual")
+	print ""
+
+	# Hyphenation is indeed rather annoying, in particular with long links.
+	print ".nh"
+}
+
+function format(line,    v) {
+	# Pass-through, otherwise useful for hacks, is a bit of a lie here,
+	# and formatting doesn't fully respect word boundaries.
+	while (line) {
+		if (match(line, /^[+][+][+][^+]+[+][+][+]/)) {
+			v = v substr(line, RSTART + 3, RLENGTH - 6)
+		} else if (match(line, /^__[^_]+__/)) {
+			v = v "\\fI" substr(line, RSTART + 2, RLENGTH - 4) "\\fP"
+		} else if (match(line, /^[*][*][^*]+[*][*]/)) {
+			v = v "\\fB" substr(line, RSTART + 2, RLENGTH - 4) "\\fP"
+		} else if (match(line, /^_[^_]+_/) &&
+			substr(line, RSTART + RLENGTH) !~ /^[[:alnum:]]/) {
+			v = v "\\fI" substr(line, RSTART + 1, RLENGTH - 2) "\\fP"
+		} else if (match(line, /^[*][^*]+[*]/) &&
+			substr(line, RSTART + RLENGTH) !~ /^[[:alnum:]]/) {
+			v = v "\\fB" substr(line, RSTART + 1, RLENGTH - 2) "\\fP"
+		} else {
+			v = v substr(line, 1, 1)
+			line = substr(line, 2)
+			continue
+		}
+		line = substr(line, RSTART + RLENGTH)
+	}
+	return v
+}
+
+function inline(line) {
+	if (!line) {
+		print ".sp"
+		return
+	}
+
+	line = format(escape(expand(line)))
+
+	# Strip empty URL descriptions, otherwise useful for demarking the end.
+	while (match(line, /[^[:space:]]+\[\]/)) {
+		line = substr(line, 1, RSTART + RLENGTH - 3) \
+			 substr(line, RSTART + RLENGTH)
+	}
+
+	# Enable double-spacing after the end of a sentence.
+	gsub(/[.][[:space:]]+/, ".\n", line)
+	gsub(/[!][[:space:]]+/, "!\n", line)
+	gsub(/[?][[:space:]]+/, "?\n", line)
+
+	# Quote commands resulting from that, as well as from expand().
+	gsub(/\n[.]/, "\n\\\\\\&.", line)
+	gsub(/\n[']/, "\n\\\\\\&'", line)
+
+	sub(/[[:space:]]+[+]$/, "\n.br", line)
+	print line
+}
+
+# Returns 1 iff the left-over $0 should be processed further.
+function process(firstline) {
+	if (readattribute(firstline))
+		return 0
+	if (getline <= 0) {
+		inline(firstline)
+		return 0
+	}
+
+	if (length(firstline) == length($0) && /^-+$/) {
+		print ".SH \"" escape(toupper(expand(firstline))) "\""
+		return 0
+	}
+	if (length(firstline) == length($0) && /^~+$/) {
+		print ".SS \"" escape(expand(firstline)) "\""
+		return 0
+	}
+	if (firstline ~ /^(-{4,}|[.]{4,})$/) {
+		print ".if n .RS 4"
+		print ".nf"
+		print ".fam C"
+		do {
+			print escape($0)
+		} while (getline > 0 && $0 != firstline)
+		print ".fam"
+		print ".fi"
+		print ".if n .RE"
+		return 0
+	}
+	if (firstline ~ /^\/{4,}$/) {
+		do {
+			print ".\\\" " $0
+		} while (getline > 0 && $0 != firstline)
+		return 0
+	}
+	if (match(firstline, /^\/\//)) {
+		print ".\\\" " firstline
+		return 1
+	}
+
+	# We generally assume these block end with a blank line.
+	if (match(firstline, /^[[:space:]]*[*][[:space:]]+/)) {
+		# Bullet magic copied over from AsciiDoc/Asciidoctor generators.
+		print ".RS 4"
+		print ".ie n \\{\\"
+		print "\\h'-04'\\(bu\\h'+03'\\c"
+		print ".\\}"
+		print ".el \\{\\"
+		print ".sp -1"
+		print ".IP \\(bu 2.3"
+		print ".\\}"
+
+		inline(substr(firstline, RSTART + RLENGTH))
+		while ($0) {
+			sub(/^[[:space:]]+/, "")
+			sub(/^[+]$/, "")
+			if (!process($0) && getline <= 0)
+				fatal("unexpected EOF")
+			if (match($0, /^[[:space:]]*[*][[:space:]]+/))
+				break
+		}
+		print ".RE"
+		print ".sp"
+		return !!$0
+	}
+	if (match(firstline, /^[[:space:]]+/)) {
+		print ".if n .RS 4"
+		print ".nf"
+		print ".fam C"
+		do {
+			print escape(substr(firstline, RLENGTH + 1))
+			firstline = $0
+		} while ($0 && getline > 0)
+		print ".fam"
+		print ".fi"
+		print ".if n .RE"
+		return 1
+	}
+	if (match(firstline, /::$/)) {
+		inline(substr(firstline, 1, RSTART - 1))
+		while (match($0, /::$/)) {
+			print ".br"
+			inline(substr($0, 1, RSTART - 1))
+			if (getline <= 0)
+				fatal("unexpected EOF")
+		}
+
+		print ".RS 4"
+		while ($0) {
+			sub(/^[[:space:]]+/, "")
+			sub(/^[+]$/, "")
+			if (!process($0) && getline <= 0)
+				fatal("unexpected EOF")
+			if (match($0, /::$/))
+				break
+		}
+		print ".RE"
+		print ".sp"
+		return !!$0
+	}
+	inline(firstline)
+	return 1
+}
+
+{
+	while (process($0)) {}
+}

tools/cmake-dump.awk

@@ -0,0 +1,24 @@
+# cmake-dump.awk: dump parsed CMake scripts as tables
+#
+# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+#
+# Parsed scripts are output in a table, with commands separated using ASCII
+# Record Separators, and arguments using Unit Separators.
+#
+# Example usage: awk -f cmake-parser.awk -f cmake-dump.awk CMakeLists.txt \
+#  | sed 'y/\x1F\x1E\t\n/\t\n  /' \
+#  | sed -n '/^project\t\([^\t]*\).*\tVERSION\t\([^\t]*\).*/{s//\1 \2/p;q;}'
+
+function sanitize(s) {
+	if (s ~ /[\x1E\x1F]/)
+		fatal("conflicting ASCII control characters found in source")
+	return s
+}
+
+Command {
+	out = sanitize(Command)
+	for (i in Args)
+		out = out "\x1F" sanitize(Args[i])
+	printf "%s\x1E", out
+}

tools/cmake-parser.awk

@@ -0,0 +1,252 @@
+# cmake-parser.awk: rudimentary CMake script parser
+#
+# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+#
+# Implemented roughly according to the grammar described in cmake-language(7),
+# which is self-conflicting, and not an accurate description.
+#
+# The result of parsing is stored in the case-normalized Command variable,
+# and the Args array.  These can be used by subsequent scripts.
+
+function warning(message) {
+	print FILENAME ":" FNR ": warning: " message > "/dev/stderr"
+}
+
+function fatal(message) {
+	print FILENAME ":" FNR ": fatal error: " message > "/dev/stderr"
+	exit 1
+}
+
+function expect(v) {
+	if (!v && v == 0)
+		fatal("broken expectations at `" $0 "'")
+	return v
+}
+
+function literal(v) {
+	if (substr($0, 1, length(v)) != v)
+		return 0
+	$0 = substr($0, length(v) + 1)
+	return 1
+}
+
+function regexp(re) {
+	if (!match($0, "^" re))
+		return 0
+	$0 = substr($0, RLENGTH + 1)
+	return 1
+}
+
+function space() {
+	return regexp("[ \t]+")
+}
+
+function unbracket(len,    v) {
+	do {
+		if (match($0, "]={" len "}]")) {
+			v = v substr($0, 1, RSTART - 1)
+			$0 = substr($0, RSTART + RLENGTH)
+			return v
+		}
+		v = v $0 RS
+	} while (getline > 0)
+	fatal("unterminated bracket")
+}
+
+function bracket_comment() {
+	if (!match($0, /^#\[=*\[/))
+		return 0
+	$0 = substr($0, RSTART + RLENGTH)
+	unbracket(RLENGTH - 3)
+	return 1
+}
+
+function line_ending() {
+	while (space() || bracket_comment()) {}
+	if (/^#/)
+		$0 = ""
+	return !$0
+}
+
+# ------------------------------------------------------------------------------
+
+# While elementary expansion of previously set variables is implementable,
+# it doesn't seem to be worth the effort.
+function expand(s,    v) {
+	v = s
+	while (match(v, /\\*[$](|ENV|CACHE)[{]/)) {
+		if (index(substr(v, RSTART), "$") % 2 != 0) {
+			warning("variable expansion is not supported: " s)
+			return s
+		}
+		v = substr(v, RSTART + RLENGTH)
+	}
+	return s
+}
+
+function escape_sequence(    v) {
+	if (!literal("\\"))
+		return 0
+
+	if (literal("t")) return "\t"
+	if (literal("r")) return "\r"
+	if (literal("n")) return "\n"
+
+	# escape_semicolon isn't treated any specially here.
+	if (regexp("[A-Za-z0-9]"))
+		fatal("unsupported escape sequence")
+
+	if ($0) {
+		v = substr($0, 1, 1)
+		$0 = substr($0, 2)
+		return v
+	}
+	if (getline > 0)
+		return ""
+	fatal("premature end of file")
+}
+
+function quoted_argument(    v, unescaped) {
+	if (!literal("\""))
+		return 0
+
+	v = ""
+	while (!literal("\"")) {
+		if (!$0) {
+			if (getline <= 0)
+				fatal("premature end of file")
+			v = v RS
+		} else if ((unescaped = escape_sequence())) {
+			if (unescaped == "\\" || unescaped == "$")
+				v = v "\\"
+			else if (unescaped == ";")
+				v = v "\\\\"
+			v = v unescaped
+		} else if (unescaped == "") {
+			# quoted_continuation
+		} else {
+			v = v substr($0, 1, 1)
+			$0 = substr($0, 2)
+		}
+	}
+	return v
+}
+
+function finalize_quoted(expanded,    v) {
+	while (match(expanded, /\\./)) {
+		v = v substr(expanded, 1, RSTART - 1) \
+			substr(expanded, RSTART + 1, 1)
+		expanded = substr(expanded, RSTART + RLENGTH)
+	}
+	Args[++N] = v expanded
+}
+
+function unquoted_argument(    v, unescaped) {
+	while (1) {
+		if (match($0, /^[^[:space:]()#"\\]+/)) {
+			v = v substr($0, RSTART, RLENGTH)
+			$0 = substr($0, RSTART + RLENGTH)
+		} else if ((unescaped = escape_sequence())) {
+			if (unescaped == "\\" || unescaped == "$" || unescaped == ";")
+				v = v "\\"
+			v = v unescaped
+		} else if (unescaped == "") {
+			fatal("unexpected backslash in an unquoted argument")
+		} else {
+			# unquoted_legacy is not supported.
+			return v
+		}
+	}
+}
+
+function finalize_unquoted(expanded,    v) {
+	while (expanded) {
+		if (expanded ~ /^;/) {
+			if (v)
+				Args[++N] = v
+			v = ""
+			expanded = substr(expanded, 2)
+		} else if (expanded ~ /^\\./) {
+			v = v substr(expanded, 2, 1)
+			expanded = substr(expanded, 3)
+		} else {
+			v = v substr(expanded, 1, 1)
+			expanded = substr(expanded, 2)
+		}
+	}
+	if (v)
+		Args[++N] = v
+}
+
+# We keep and reprocess some escape sequences in here.
+function argument(    arg, expanded, v) {
+	if (regexp("\\[=*\\["))
+		Args[++N] = unbracket(RLENGTH - 2)
+	else if ((arg = quoted_argument()) || arg == "")
+		finalize_quoted(expand(arg))
+	else if ((arg = unquoted_argument()))
+		finalize_unquoted(expand(arg))
+	else
+		return 0
+	return 1
+}
+
+# ------------------------------------------------------------------------------
+
+function identifier(    v) {
+	if (!match($0, /^[A-Za-z_][A-Za-z0-9_]*/))
+		return 0
+	v = substr($0, 1, RLENGTH)
+	$0 = substr($0, RLENGTH + 1)
+	return v
+}
+
+function separation() {
+	if (space() || bracket_comment())
+		return 1
+
+	if (!line_ending())
+		return 0
+	if (getline > 0)
+		return 1
+	fatal("premature end of file")
+}
+
+function command_invocation(    level) {
+	while (space()) {}
+	Command = identifier()
+	if (!Command)
+		return 0
+	while (space()) {}
+
+	Command = tolower(Command)
+	for (N in Args)
+		delete Args[N]
+
+	N = 0
+	expect(literal("("))
+	while (1) {
+		while (separation()) {}
+		if (literal(")")) {
+			if (!level--)
+				break
+			Args[++N] = ")"
+			continue
+		}
+		if (literal("(")) {
+			level++
+			Args[++N] = "("
+			continue
+		}
+		expect(argument())
+		if (!/^[()]/)
+			expect(separation())
+	}
+	return 1
+}
+
+{
+	command_invocation()
+	expect(line_ending())
+}

tools/lxdrgen-c.awk

@@ -0,0 +1,324 @@
+# lxdrgen-c.awk: C backend for lxdrgen.awk.
+#
+# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+#
+# Neither *_new() nor *_destroy() functions are provided, because they'd only
+# be useful for top-levels, and are merely extra malloc()/free() calls.
+# Users are expected to reuse buffers.
+#
+# Similarly, no constructors are produced--those are easy to write manually.
+#
+# All arrays are deserialized zero-terminated, so u8<> and i8<> can be directly
+# used as C strings.
+#
+# All types must be able to dispose partially zero values going from the back,
+# i.e., in the reverse order of deserialization.
+
+function define_internal(name, ctype) {
+	Types[name] = "internal"
+	CodegenCType[name] = ctype
+}
+
+function define_int(shortname, ctype) {
+	define_internal(shortname, ctype)
+	CodegenSerialize[shortname] = \
+		"\tstr_pack_" shortname "(w, %s);\n"
+	CodegenDeserialize[shortname] = \
+		"\tif (!msg_unpacker_" shortname "(r, &%s))\n" \
+		"\t\treturn false;\n"
+}
+
+function define_sint(size) { define_int("i" size, "int" size "_t") }
+function define_uint(size) { define_int("u" size, "uint" size "_t") }
+
+function codegen_begin() {
+	define_sint("8")
+	define_sint("16")
+	define_sint("32")
+	define_sint("64")
+	define_uint("8")
+	define_uint("16")
+	define_uint("32")
+	define_uint("64")
+
+	define_internal("string", "struct str")
+	CodegenDispose["string"] = "\tstr_free(&%s);\n"
+	CodegenSerialize["string"] = \
+		"\tif (!proto_string_serialize(&%s, w))\n" \
+		"\t\treturn false;\n"
+	CodegenDeserialize["string"] = \
+		"\tif (!proto_string_deserialize(&%s, r))\n" \
+		"\t\treturn false;\n"
+
+	define_internal("bool", "bool")
+	CodegenSerialize["bool"] = \
+		"\tstr_pack_u8(w, !!%s);\n"
+	CodegenDeserialize["bool"] = \
+		"\t{\n" \
+		"\t\tuint8_t v = 0;\n" \
+		"\t\tif (!msg_unpacker_u8(r, &v))\n" \
+		"\t\t\treturn false;\n" \
+		"\t\t%s = !!v;\n" \
+		"\t}\n"
+
+	print "// Code generated from " FILENAME ". DO NOT EDIT."
+	print "// This file directly depends on liberty.c, but doesn't include it."
+	print ""
+	print "static bool"
+	print "proto_string_serialize(const struct str *s, struct str *w) {"
+	print "\tif (s->len > UINT32_MAX)"
+	print "\t\treturn false;"
+	print "\tstr_pack_u32(w, s->len);"
+	print "\tstr_append_str(w, s);"
+	print "\treturn true;"
+	print "}"
+	print ""
+	print "static bool"
+	print "proto_string_deserialize(struct str *s, struct msg_unpacker *r) {"
+	print "\tuint32_t len = 0;"
+	print "\tif (!msg_unpacker_u32(r, &len))"
+	print "\t\treturn false;"
+	print "\tif (msg_unpacker_get_available(r) < len)"
+	print "\t\treturn false;"
+	print "\t*s = str_make();"
+	print "\tstr_append_data(s, r->data + r->offset, len);"
+	print "\tr->offset += len;"
+	print "\tif (!utf8_validate (s->str, s->len))"
+	print "\t\treturn false;"
+	print "\treturn true;"
+	print "}"
+}
+
+function codegen_constant(name, value) {
+	print ""
+	print "enum { " PrefixUpper name " = " value " };"
+}
+
+function codegen_enum_value(name, subname, value, cg) {
+	append(cg, "fields",
+		"\t" PrefixUpper toupper(cameltosnake(name)) "_" subname \
+		" = " value ",\n")
+}
+
+function codegen_enum(name, cg,    ctype) {
+	ctype = "enum " PrefixLower cameltosnake(name)
+	print ""
+	print ctype " {"
+	print cg["fields"] "};"
+
+	# XXX: This should also check if it isn't out-of-range for any reason,
+	# but our usage of sprintf() stands in the way a bit.
+	CodegenSerialize[name] = "\tstr_pack_i8(w, %s);\n"
+	CodegenDeserialize[name] = \
+		"\t{\n" \
+		"\t\tint8_t v = 0;\n" \
+		"\t\tif (!msg_unpacker_i8(r, &v) || !v)\n" \
+		"\t\t\treturn false;\n" \
+		"\t\t%s = v;\n" \
+		"\t}\n"
+
+	CodegenCType[name] = ctype
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_struct_tag(d, cg,    f) {
+	f = "self->" d["name"]
+	append(cg, "fields", "\t" CodegenCType[d["type"]] " " d["name"] ";\n")
+	append(cg, "dispose", sprintf(CodegenDispose[d["type"]], f))
+	append(cg, "serialize", sprintf(CodegenSerialize[d["type"]], f))
+	# Do not deserialize here, that would be out of order.
+}
+
+function codegen_struct_field(d, cg,    f, dispose, serialize, deserialize) {
+	f = "self->" d["name"]
+	dispose = CodegenDispose[d["type"]]
+	serialize = CodegenSerialize[d["type"]]
+	deserialize = CodegenDeserialize[d["type"]]
+	if (!d["isarray"]) {
+		append(cg, "fields", "\t" CodegenCType[d["type"]] " " d["name"] ";\n")
+		append(cg, "dispose", sprintf(dispose, f))
+		append(cg, "serialize", sprintf(serialize, f))
+		append(cg, "deserialize", sprintf(deserialize, f))
+		return
+	}
+
+	append(cg, "fields",
+		"\t" CodegenCType["u32"] " " d["name"] "_len;\n" \
+		"\t" CodegenCType[d["type"]] " *" d["name"] ";\n")
+
+	if (dispose)
+		append(cg, "dispose", "\tif (" f ")\n" \
+			"\t\tfor (size_t i = 0; i < " f "_len; i++)\n" \
+			indent(indent(sprintf(dispose, f "[i]"))))
+	append(cg, "dispose", "\tfree(" f ");\n")
+
+	append(cg, "serialize", sprintf(CodegenSerialize["u32"], f "_len"))
+	if (d["type"] == "u8" || d["type"] == "i8") {
+		append(cg, "serialize",
+			"\tstr_append_data(w, " f ", " f "_len);\n")
+	} else if (serialize) {
+		append(cg, "serialize",
+			"\tfor (size_t i = 0; i < " f "_len; i++)\n" \
+			indent(sprintf(serialize, f "[i]")))
+	}
+
+	append(cg, "deserialize", sprintf(CodegenDeserialize["u32"], f "_len") \
+		"\tif (!(" f " = calloc(" f "_len + 1, sizeof *" f ")))\n" \
+		"\t\treturn false;\n")
+	if (d["type"] == "u8" || d["type"] == "i8") {
+		append(cg, "deserialize",
+			"\tif (msg_unpacker_get_available(r) < " f "_len)\n" \
+			"\t\treturn false;\n" \
+			"\tmemcpy(" f ", r->data + r->offset, " f "_len);\n" \
+			"\tr->offset += " f "_len;\n")
+	} else if (deserialize) {
+		append(cg, "deserialize",
+			"\tfor (size_t i = 0; i < " f "_len; i++)\n" \
+			indent(sprintf(deserialize, f "[i]")))
+	}
+}
+
+function codegen_struct(name, cg,    ctype, funcname) {
+	ctype = "struct " PrefixLower cameltosnake(name)
+	print ""
+	print ctype " {"
+	print cg["fields"] "};"
+
+	if (cg["dispose"]) {
+		funcname = PrefixLower cameltosnake(name) "_free"
+		print ""
+		print "static void\n" funcname "(" ctype " *self) {"
+		print cg["dispose"] "}"
+
+		CodegenDispose[name] = "\t" funcname "(&%s);\n"
+	}
+	if (cg["serialize"]) {
+		funcname = PrefixLower cameltosnake(name) "_serialize"
+		print ""
+		print "static bool\n" \
+			  funcname "(\n\t\tconst " ctype " *self, struct str *w) {"
+		print cg["serialize"] "\treturn true;"
+		print "}"
+
+		CodegenSerialize[name] = "\tif (!" funcname "(&%s, w))\n" \
+			"\t\treturn false;\n"
+	}
+	if (cg["deserialize"]) {
+		funcname = PrefixLower cameltosnake(name) "_deserialize"
+		print ""
+		print "static bool\n" \
+			  funcname "(\n\t\t" ctype " *self, struct msg_unpacker *r) {"
+		print cg["deserialize"] "\treturn true;"
+		print "}"
+
+		CodegenDeserialize[name] = "\tif (!" funcname "(&%s, r))\n" \
+			"\t\treturn false;\n"
+	}
+
+	CodegenCType[name] = ctype
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_union_tag(d, cg) {
+	cg["tagtype"] = d["type"]
+	cg["tagname"] = d["name"]
+	append(cg, "fields", "\t" CodegenCType[d["type"]] " " d["name"] ";\n")
+}
+
+function codegen_union_struct( \
+		name, casename, cg, scg,     structname, fieldname, fullcasename) {
+	# Don't generate obviously useless structs.
+	fullcasename = toupper(cameltosnake(cg["tagtype"])) "_" casename
+	if (!scg["dispose"] && !scg["deserialize"]) {
+		append(cg, "structless", "\tcase " PrefixUpper fullcasename ":\n")
+		for (i in scg)
+			delete scg[i]
+		return
+	}
+
+	# And thus not all generated structs are present in Types.
+	structname = name "_" casename
+	fieldname = tolower(casename)
+	codegen_struct(structname, scg)
+
+	append(cg, "fields", "\t" CodegenCType[structname] " " fieldname ";\n")
+	if (CodegenDispose[structname])
+		append(cg, "dispose", "\tcase " PrefixUpper fullcasename ":\n" \
+			indent(sprintf(CodegenDispose[structname], "self->" fieldname)) \
+			"\t\tbreak;\n")
+
+	# With no de/serialization code, this will simply recognize the tag.
+	append(cg, "serialize", "\tcase " PrefixUpper fullcasename ":\n" \
+		indent(sprintf(CodegenSerialize[structname], "self->" fieldname)) \
+		"\t\tbreak;\n")
+	append(cg, "deserialize", "\tcase " PrefixUpper fullcasename ":\n" \
+		indent(sprintf(CodegenDeserialize[structname], "self->" fieldname)) \
+		"\t\tbreak;\n")
+}
+
+function codegen_union(name, cg,    f, ctype, funcname) {
+	ctype = "union " PrefixLower cameltosnake(name)
+	print ""
+	print ctype " {"
+	print cg["fields"] "};"
+
+	f = "self->" cg["tagname"]
+	if (cg["dispose"]) {
+		funcname = PrefixLower cameltosnake(name) "_free"
+		print ""
+		print "static void\n" funcname "(" ctype " *self) {"
+		print "\tswitch (" f ") {"
+		if (cg["structless"])
+			print cg["structless"] \
+				indent(sprintf(CodegenDispose[cg["tagtype"]], f)) "\t\tbreak;"
+		print cg["dispose"] "\tdefault:"
+		print "\t\tbreak;"
+		print "\t}"
+		print "}"
+
+		CodegenDispose[name] = "\t" funcname "(&%s);\n"
+	}
+	if (cg["serialize"]) {
+		funcname = PrefixLower cameltosnake(name) "_serialize"
+		print ""
+		print "static bool\n" \
+			  funcname "(\n\t\tconst " ctype " *self, struct str *w) {"
+		print "\tswitch (" f ") {"
+		if (cg["structless"])
+			print cg["structless"] \
+				indent(sprintf(CodegenSerialize[cg["tagtype"]], f)) "\t\tbreak;"
+		print cg["serialize"] "\tdefault:"
+		print "\t\treturn false;"
+		print "\t}"
+		print "\treturn true;"
+		print "}"
+
+		CodegenSerialize[name] = "\tif (!" funcname "(&%s, w))\n" \
+			"\t\treturn false;\n"
+	}
+	if (cg["deserialize"]) {
+		funcname = PrefixLower cameltosnake(name) "_deserialize"
+		print ""
+		print "static bool\n" \
+			  funcname "(\n\t\t" ctype " *self, struct msg_unpacker *r) {"
+		print sprintf(CodegenDeserialize[cg["tagtype"]], f)
+		print "\tswitch (" f ") {"
+		if (cg["structless"])
+			print cg["structless"] "\t\tbreak;"
+		print cg["deserialize"] "\tdefault:"
+		print "\t\treturn false;"
+		print "\t}"
+		print "\treturn true;"
+		print "}"
+
+		CodegenDeserialize[name] = "\tif (!" funcname "(&%s, r))\n" \
+			"\t\treturn false;\n"
+	}
+
+	CodegenCType[name] = ctype
+	for (i in cg)
+		delete cg[i]
+}

tools/lxdrgen-go.awk

@@ -0,0 +1,541 @@
+# lxdrgen-go.awk: Go backend for lxdrgen.awk.
+#
+# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+#
+# This backend also enables proxying to other endpoints using JSON.
+
+function define_internal(name, gotype) {
+	Types[name] = "internal"
+	CodegenGoType[name] = gotype
+}
+
+function define_sint(size,    shortname, gotype) {
+	shortname = "i" size
+	gotype = "int" size
+	define_internal(shortname, gotype)
+
+	CodegenAppendJSON[shortname] = \
+		"\tb = strconv.AppendInt(b, int64(%s), 10)\n"
+	if (size == 8) {
+		CodegenSerialize[shortname] = "\tdata = append(data, uint8(%s))\n"
+		CodegenDeserialize[shortname] = \
+			"\tif len(data) >= 1 {\n" \
+			"\t\t%s, data = int8(data[0]), data[1:]\n" \
+			"\t} else {\n" \
+			"\t\treturn nil, false\n" \
+			"\t}\n"
+		return
+	}
+
+	CodegenSerialize[shortname] = \
+		"\tdata = binary.BigEndian.AppendUint" size "(data, uint" size "(%s))\n"
+	CodegenDeserialize[shortname] = \
+		"\tif len(data) >= " (size / 8) " {\n" \
+		"\t\t%s = " gotype "(binary.BigEndian.Uint" size "(data))\n" \
+		"\t\tdata = data[" (size / 8) ":]\n" \
+		"\t} else {\n" \
+		"\t\treturn nil, false\n" \
+		"\t}\n"
+}
+
+function define_uint(size,    shortname, gotype) {
+	# Both []byte and []uint8 luckily marshal as base64-encoded JSON strings,
+	# so there's no need to rename the type as an exception.
+	shortname = "u" size
+	gotype = "uint" size
+	define_internal(shortname, gotype)
+
+	CodegenAppendJSON[shortname] = \
+		"\tb = strconv.AppendUint(b, uint64(%s), 10)\n"
+	if (size == 8) {
+		CodegenSerialize[shortname] = "\tdata = append(data, %s)\n"
+		CodegenDeserialize[shortname] = \
+			"\tif len(data) >= 1 {\n" \
+			"\t\t%s, data = data[0], data[1:]\n" \
+			"\t} else {\n" \
+			"\t\treturn nil, false\n" \
+			"\t}\n"
+		return
+	}
+
+	CodegenSerialize[shortname] = \
+		"\tdata = binary.BigEndian.AppendUint" size "(data, %s)\n"
+	CodegenDeserialize[shortname] = \
+		"\tif len(data) >= " (size / 8) " {\n" \
+		"\t\t%s = binary.BigEndian.Uint" size "(data)\n" \
+		"\t\tdata = data[" (size / 8) ":]\n" \
+		"\t} else {\n" \
+		"\t\treturn nil, false\n" \
+		"\t}\n"
+}
+
+# Currently two outputs cannot coexist within the same package.
+function codegen_private(name) {
+	return "proto" name
+}
+
+function codegen_begin(    funcname) {
+	define_sint("8")
+	define_sint("16")
+	define_sint("32")
+	define_sint("64")
+	define_uint("8")
+	define_uint("16")
+	define_uint("32")
+	define_uint("64")
+	define_internal("bool", "bool")
+	define_internal("string", "string")
+
+	# Cater to "go generate", for what it's worth.
+	CodegenPackage = ENV["GOPACKAGE"]
+	if (!CodegenPackage)
+		CodegenPackage = "main"
+
+	print "// Code generated from " FILENAME ". DO NOT EDIT."
+	print ""
+	print "package " CodegenPackage
+	print ""
+	print "import ("
+	print "\t`encoding/base64`"
+	print "\t`encoding/binary`"
+	print "\t`encoding/json`"
+	print "\t`errors`"
+	print "\t`math`"
+	print "\t`strconv`"
+	print "\t`unicode/utf8`"
+	print ")"
+	print ""
+	print "// This is a hack to always use the base64 import."
+	print "var _ = base64.StdEncoding"
+	print ""
+
+	CodegenAppendJSON["bool"] = \
+		"\tb = strconv.AppendBool(b, %s)\n"
+	CodegenSerialize["bool"] = \
+		"\tif %s {\n" \
+		"\t\tdata = append(data, 1)\n" \
+		"\t} else {\n" \
+		"\t\tdata = append(data, 0)\n" \
+		"\t}\n"
+
+	funcname = codegen_private("ConsumeBoolFrom")
+	print "// " funcname " tries to deserialize a boolean value"
+	print "// from the beginning of a byte stream. When successful,"
+	print "// it returns a subslice with any data that might follow."
+	print "func " funcname "(data []byte, b *bool) ([]byte, bool) {"
+	print "\tif len(data) < 1 {"
+	print "\t\treturn nil, false"
+	print "\t}"
+	print "\tif data[0] != 0 {"
+	print "\t\t*b = true"
+	print "\t} else {"
+	print "\t\t*b = false"
+	print "\t}"
+	print "\treturn data[1:], true"
+	print "}"
+	print ""
+
+	CodegenDeserialize["bool"] = \
+		"\tif data, ok = " funcname "(data, &%s); !ok {\n" \
+		"\t\treturn nil, ok\n" \
+		"\t}\n"
+
+	funcname = codegen_private("AppendStringTo")
+	print "// " funcname " tries to serialize a string value,"
+	print "// appending it to the end of a byte stream."
+	print "func " funcname "(data []byte, s string) ([]byte, bool) {"
+	print "\tif len(s) > math.MaxUint32 {"
+	print "\t\treturn nil, false"
+	print "\t}"
+	print "\tdata = binary.BigEndian.AppendUint32(data, uint32(len(s)))"
+	print "\treturn append(data, s...), true"
+	print "}"
+	print ""
+
+	CodegenSerialize["string"] = \
+		"\tif data, ok = " funcname "(data, %s); !ok {\n" \
+		"\t\treturn nil, ok\n" \
+		"\t}\n"
+
+	funcname = codegen_private("ConsumeStringFrom")
+	print "// " funcname " tries to deserialize a string value"
+	print "// from the beginning of a byte stream. When successful,"
+	print "// it returns a subslice with any data that might follow."
+	print "func " funcname "(data []byte, s *string) ([]byte, bool) {"
+	print "\tif len(data) < 4 {"
+	print "\t\treturn nil, false"
+	print "\t}"
+	print "\tlength := binary.BigEndian.Uint32(data)"
+	print "\tif data = data[4:]; uint64(len(data)) < uint64(length) {"
+	print "\t\treturn nil, false"
+	print "\t}"
+	print "\t*s = string(data[:length])"
+	print "\tif !utf8.ValidString(*s) {"
+	print "\t\treturn nil, false"
+	print "\t}"
+	print "\treturn data[length:], true"
+	print "}"
+	print ""
+
+	CodegenDeserialize["string"] = \
+		"\tif data, ok = " funcname "(data, &%s); !ok {\n" \
+		"\t\treturn nil, ok\n" \
+		"\t}\n"
+
+	funcname = codegen_private("UnmarshalEnumJSON")
+	print "// " funcname " converts a JSON fragment to an integer,"
+	print "// ensuring that it's within the expected range of enum values."
+	print "func " funcname "(data []byte) (int64, error) {"
+	print "\tvar n int64"
+	print "\tif err := json.Unmarshal(data, &n); err != nil {"
+	print "\t\treturn 0, err"
+	print "\t} else if n > math.MaxInt8 || n < math.MinInt8 {"
+	print "\t\treturn 0, errors.New(`integer out of range`)"
+	print "\t} else {"
+	print "\t\treturn n, nil"
+	print "\t}"
+	print "}"
+	print ""
+}
+
+function codegen_constant(name, value) {
+	print "const " PrefixCamel snaketocamel(name) " = " value
+	print ""
+}
+
+function codegen_enum_value(name, subname, value, cg,    goname) {
+	goname = PrefixCamel name snaketocamel(subname)
+	append(cg, "fields",
+		"\t" goname " = " value "\n")
+	append(cg, "stringer",
+		"\tcase " goname ":\n" \
+		"\t\treturn `" snaketocamel(subname) "`\n")
+	append(cg, "marshal",
+		goname ",\n")
+	append(cg, "unmarshal",
+		"\tcase `" snaketocamel(subname) "`:\n" \
+		"\t\t*v = " goname "\n")
+}
+
+function codegen_enum(name, cg,    gotype, fields, funcname) {
+	gotype = PrefixCamel name
+	print "type " gotype " int8"
+	print ""
+
+	print "const ("
+	print cg["fields"] ")"
+	print ""
+
+	print "func (v " gotype ") String() string {"
+	print "\tswitch v {"
+	print cg["stringer"] "\tdefault:"
+	print "\t\treturn strconv.Itoa(int(v))"
+	print "\t}"
+	print "}"
+	print ""
+
+	CodegenIsMarshaler[name] = 1
+	fields = cg["marshal"]
+	sub(/,\n$/, ":", fields)
+	gsub(/\n/, "\n\t", fields)
+	print "func (v " gotype ") MarshalJSON() ([]byte, error) {"
+	print "\tswitch v {"
+	print indent("case " fields)
+	print "\t\treturn []byte(`\"` + v.String() + `\"`), nil"
+	print "\t}"
+	print "\treturn json.Marshal(int(v))"
+	print "}"
+	print ""
+
+	funcname = codegen_private("UnmarshalEnumJSON")
+	print "func (v *" gotype ") UnmarshalJSON(data []byte) error {"
+	print "\tvar s string"
+	print "\tif json.Unmarshal(data, &s) == nil {"
+	print "\t\t// Handled below."
+	print "\t} else if n, err := " funcname "(data); err != nil {"
+	print "\t\treturn err"
+	print "\t} else {"
+	print "\t\t*v = " gotype "(n)"
+	print "\t\treturn nil"
+	print "\t}"
+	print ""
+	print "\tswitch s {"
+	print cg["unmarshal"] "\tdefault:"
+	print "\t\treturn errors.New(`unrecognized value: ` + s)"
+	print "\t}"
+	print "\treturn nil"
+	print "}"
+	print ""
+
+	# XXX: This should also check if it isn't out-of-range for any reason,
+	# but our usage of sprintf() stands in the way a bit.
+	CodegenSerialize[name] = "\tdata = append(data, uint8(%s))\n"
+	CodegenDeserialize[name] = \
+		"\tif len(data) >= 1 {\n" \
+		"\t\t%s, data = " gotype "(data[0]), data[1:]\n" \
+		"\t} else {\n" \
+		"\t\treturn nil, false\n" \
+		"\t}\n"
+
+	CodegenGoType[name] = gotype
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_marshal(type, f,    marshal) {
+	if (CodegenAppendJSON[type])
+		return sprintf(CodegenAppendJSON[type], f)
+
+	# Complex types are json.Marshalers, there's no need to json.Marshal(&f).
+	if (CodegenIsMarshaler[type])
+		marshal = f ".MarshalJSON()"
+	else
+		marshal = "json.Marshal(" f ")"
+
+	return \
+		"\tif j, err := " marshal "; err != nil {\n" \
+		"\t\treturn nil, err\n" \
+		"\t} else {\n" \
+		"\t\tb = append(b, j...)\n" \
+		"\t}\n"
+}
+
+function codegen_struct_field_marshal(d, cg,    camel, f, marshal) {
+	camel = snaketocamel(d["name"])
+	f = "s." camel
+	if (!d["isarray"]) {
+		append(cg, "marshal",
+			"\tb = append(b, `,\"" decapitalize(camel) "\":`...)\n" \
+			codegen_marshal(d["type"], f))
+		return
+	}
+
+	# Note that we do not produce `null` for nil slices, unlike encoding/json.
+	# And arrays never get deserialized as such.
+	if (d["type"] == "u8") {
+		append(cg, "marshal",
+			"\tb = append(b, `,\"" decapitalize(camel) "\":\"`...)\n" \
+			"\tb = append(b, base64.StdEncoding.EncodeToString(" f ")...)\n" \
+			"\tb = append(b, '\"')\n")
+		return
+	}
+
+	append(cg, "marshal",
+		"\tb = append(b, `,\"" decapitalize(camel) "\":[`...)\n" \
+		"\tfor i := 0; i < len(" f "); i++ {\n" \
+		"\t\tif i > 0 {\n" \
+		"\t\t\tb = append(b, ',')\n" \
+		"\t\t}\n" \
+		indent(codegen_marshal(d["type"], f "[i]")) \
+		"\t}\n" \
+		"\tb = append(b, ']')\n")
+}
+
+function codegen_struct_field(d, cg,    camel, f, serialize, deserialize) {
+	codegen_struct_field_marshal(d, cg)
+
+	camel = snaketocamel(d["name"])
+	f = "s." camel
+	serialize = CodegenSerialize[d["type"]]
+	deserialize = CodegenDeserialize[d["type"]]
+	if (!d["isarray"]) {
+		append(cg, "fields", "\t" camel " " CodegenGoType[d["type"]] \
+			" `json:\"" decapitalize(camel) "\"`\n")
+		append(cg, "serialize", sprintf(serialize, f))
+		append(cg, "deserialize", sprintf(deserialize, f))
+		return
+	}
+
+	append(cg, "fields", "\t" camel " []" CodegenGoType[d["type"]] \
+		" `json:\"" decapitalize(camel) "\"`\n")
+
+	# XXX: This should also check if it isn't out-of-range for any reason.
+	append(cg, "serialize",
+		sprintf(CodegenSerialize["u32"], "uint32(len(" f "))"))
+	if (d["type"] == "u8") {
+		append(cg, "serialize",
+			"\tdata = append(data, " f "...)\n")
+	} else {
+		append(cg, "serialize",
+			"\tfor i := 0; i < len(" f "); i++ {\n" \
+			indent(sprintf(serialize, f "[i]")) \
+			"\t}\n")
+	}
+
+	append(cg, "deserialize",
+		"\t{\n" \
+		"\t\tvar length uint32\n" \
+		indent(sprintf(CodegenDeserialize["u32"], "length")))
+	if (d["type"] == "u8") {
+		append(cg, "deserialize",
+			"\t\tif uint64(len(data)) < uint64(length) {\n" \
+			"\t\t\treturn nil, false\n" \
+			"\t\t}\n" \
+			"\t\t" f ", data = data[:length], data[length:]\n" \
+			"\t}\n")
+	} else {
+		append(cg, "deserialize",
+			"\t\t" f " = make([]" CodegenGoType[d["type"]] ", length)\n" \
+			"\t}\n" \
+			"\tfor i := 0; i < len(" f "); i++ {\n" \
+			indent(sprintf(deserialize, f "[i]")) \
+			"\t}\n")
+	}
+}
+
+function codegen_struct_tag(d, cg,    camel, f) {
+	codegen_struct_field_marshal(d, cg)
+
+	camel = snaketocamel(d["name"])
+	f = "s." camel
+	append(cg, "fields", "\t" camel " " CodegenGoType[d["type"]] \
+		" `json:\"" decapitalize(camel) "\"`\n")
+	append(cg, "serialize", sprintf(CodegenSerialize[d["type"]], f))
+	# Do not deserialize here, that is already done by the containing union.
+}
+
+function codegen_struct(name, cg,    gotype) {
+	gotype = PrefixCamel name
+	print "type " gotype " struct {\n" cg["fields"] "}\n"
+
+	if (cg["marshal"]) {
+		CodegenIsMarshaler[name] = 1
+		print "func (s *" gotype ") MarshalJSON() ([]byte, error) {"
+		print "\tb := []byte{}"
+		print cg["marshal"] "\tb[0] = '{'"
+		print "\treturn append(b, '}'), nil"
+		print "}"
+		print ""
+	}
+
+	if (cg["serialize"]) {
+		print "func (s *" gotype ") AppendTo(data []byte) ([]byte, bool) {"
+		print "\tok := true"
+		print cg["serialize"] "\treturn data, ok"
+		print "}"
+		print ""
+
+		CodegenSerialize[name] = \
+			"\tif data, ok = %s.AppendTo(data); !ok {\n" \
+			"\t\treturn nil, ok\n" \
+			"\t}\n"
+	}
+	if (cg["deserialize"]) {
+		print "func (s *" gotype ") ConsumeFrom(data []byte) ([]byte, bool) {"
+		print "\tok := true"
+		print cg["deserialize"] "\treturn data, ok"
+		print "}"
+		print ""
+
+		CodegenDeserialize[name] = \
+			"\tif data, ok = %s.ConsumeFrom(data); !ok {\n" \
+			"\t\treturn nil, ok\n" \
+			"\t}\n"
+	}
+
+	CodegenGoType[name] = gotype
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_union_tag(d, cg) {
+	cg["tagtype"] = d["type"]
+	cg["tagname"] = d["name"]
+	# The tag is implied from the type of struct stored in the interface.
+}
+
+function codegen_union_struct(name, casename, cg, scg,     structname, init) {
+	# And thus not all generated structs are present in Types.
+	structname = name snaketocamel(casename)
+	codegen_struct(structname, scg)
+
+	init = CodegenGoType[structname] "{" snaketocamel(cg["tagname"]) \
+		": " decapitalize(snaketocamel(cg["tagname"])) "}"
+	append(cg, "unmarshal",
+		"\tcase " CodegenGoType[cg["tagtype"]] snaketocamel(casename) ":\n" \
+		"\t\ts := " init "\n" \
+		"\t\terr = json.Unmarshal(data, &s)\n" \
+		"\t\tu.Interface = &s\n")
+	append(cg, "serialize",
+		"\tcase *" CodegenGoType[structname] ":\n" \
+		indent(sprintf(CodegenSerialize[structname], "union")))
+	append(cg, "deserialize",
+		"\tcase " CodegenGoType[cg["tagtype"]] snaketocamel(casename) ":\n" \
+		"\t\ts := " init "\n" \
+		indent(sprintf(CodegenDeserialize[structname], "s")) \
+		"\t\tu.Interface = &s\n")
+}
+
+function codegen_union(name, cg,    gotype, tagfield, tagvar) {
+	gotype = PrefixCamel name
+	print "type " gotype " struct {"
+	print "\tInterface any"
+	print "}"
+	print ""
+
+	# This cannot be a pointer method, it wouldn't work recursively.
+	CodegenIsMarshaler[name] = 1
+	print "func (u " gotype ") MarshalJSON() ([]byte, error) {"
+	print "\treturn u.Interface.(json.Marshaler).MarshalJSON()"
+	print "}"
+	print ""
+
+	tagfield = snaketocamel(cg["tagname"])
+	tagvar = decapitalize(tagfield)
+	print "func (u *" gotype ") UnmarshalJSON(data []byte) (err error) {"
+	print "\tvar t struct {"
+	print "\t\t" tagfield " " CodegenGoType[cg["tagtype"]] \
+		" `json:\"" tagvar "\"`"
+	print "\t}"
+	print "\tif err := json.Unmarshal(data, &t); err != nil {"
+	print "\t\treturn err"
+	print "\t}"
+	print ""
+	print "\tswitch " tagvar " := t." tagfield "; " tagvar " {"
+	print cg["unmarshal"] "\tdefault:"
+	print "\t\terr = errors.New(`unsupported value: ` + " tagvar ".String())"
+	print "\t}"
+	print "\treturn err"
+	print "}"
+	print ""
+
+	# XXX: Consider changing the interface into an AppendTo/ConsumeFrom one,
+	# that would eliminate these type case switches entirely.
+	# On the other hand, it would make it possible to send unsuitable structs.
+	print "func (u *" gotype ") AppendTo(data []byte) ([]byte, bool) {"
+	print "\tok := true"
+	print "\tswitch union := u.Interface.(type) {"
+	print cg["serialize"] "\tdefault:"
+	print "\t\treturn nil, false"
+	print "\t}"
+	print "\treturn data, ok"
+	print "}"
+	print ""
+
+	CodegenSerialize[name] = \
+		"\tif data, ok = %s.AppendTo(data); !ok {\n" \
+		"\t\treturn nil, ok\n" \
+		"\t}\n"
+
+	print "func (u *" gotype ") ConsumeFrom(data []byte) ([]byte, bool) {"
+	print "\tok := true"
+	print "\tvar " tagvar " " CodegenGoType[cg["tagtype"]]
+	print sprintf(CodegenDeserialize[cg["tagtype"]], tagvar)
+	print "\tswitch " tagvar " {"
+	print cg["deserialize"] "\tdefault:"
+	print "\t\treturn nil, false"
+	print "\t}"
+	print "\treturn data, ok"
+	print "}"
+	print ""
+
+	CodegenDeserialize[name] = \
+		"\tif data, ok = %s.ConsumeFrom(data); !ok {\n" \
+		"\t\treturn nil, ok\n" \
+		"\t}\n"
+
+	CodegenGoType[name] = gotype
+	for (i in cg)
+		delete cg[i]
+}

tools/lxdrgen-mjs.awk

@@ -0,0 +1,226 @@
+# lxdrgen-mjs.awk: Javascript backend for lxdrgen.awk.
+#
+# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+#
+# This backend is currently for decoding the binary format only.
+# (JSON is way too expensive to process and transfer.)
+#
+# Import the resulting script as a Javascript module.
+# Identifiers intentionally aren't prefixed.
+
+function define_internal(name) {
+	Types[name] = "internal"
+}
+
+function define_sint(size,    shortname) {
+	shortname = "i" size
+	define_internal(shortname)
+	CodegenDeserialize[shortname] = "\t%s = r." shortname "()\n"
+
+	print ""
+	print "\t" shortname "() {"
+	if (size == "64") {
+		# XXX: 2^53 - 1 must be enough for anyone.  BigInts are a PITA.
+		print "\t\tconst " shortname \
+			" = Number(this.getBigInt" size "(this.offset))"
+	} else {
+		print "\t\tconst " shortname " = this.getInt" size "(this.offset)"
+	}
+	print "\t\tthis.offset += " (size / 8)
+	print "\t\treturn " shortname
+	print "\t}"
+}
+
+function define_uint(size,    shortname) {
+	shortname = "u" size
+	define_internal(shortname)
+	CodegenDeserialize[shortname] = "\t%s = r." shortname "()\n"
+
+	print ""
+	print "\t" shortname "() {"
+	if (size == "64") {
+		# XXX: 2^53 - 1 must be enough for anyone.  BigInts are a PITA.
+		print "\t\tconst " shortname \
+			" = Number(this.getBigUint" size "(this.offset))"
+	} else {
+		print "\t\tconst " shortname " = this.getUint" size "(this.offset)"
+	}
+	print "\t\tthis.offset += " (size / 8)
+	print "\t\treturn " shortname
+	print "\t}"
+}
+
+function codegen_begin() {
+	print "// Code generated from " FILENAME ". DO NOT EDIT."
+	print ""
+	print "export class Reader extends DataView {"
+	print "\tconstructor() {"
+	print "\t\tsuper(...arguments)"
+	print "\t\tthis.offset = 0"
+	print "\t\tthis.decoder = new TextDecoder('utf-8', {fatal: true})"
+	print "\t}"
+	print ""
+	print "\tget empty() {"
+	print "\t\treturn this.byteLength <= this.offset"
+	print "\t}"
+	print ""
+	print "\trequire(len) {"
+	print "\t\tif (this.byteLength - this.offset < len)"
+	print "\t\t\tthrow `Premature end of data`"
+	print "\t\treturn this.byteOffset + this.offset"
+	print "\t}"
+
+	define_internal("string")
+	CodegenDeserialize["string"] = "\t%s = r.string()\n"
+
+	print ""
+	print "\tstring() {"
+	print "\t\tconst len = this.getUint32(this.offset)"
+	print "\t\tthis.offset += 4"
+	print "\t\tconst array = new Uint8Array("
+	print "\t\t\tthis.buffer, this.require(len), len)"
+	print "\t\tthis.offset += len"
+	print "\t\treturn this.decoder.decode(array)"
+	print "\t}"
+
+	define_internal("bool")
+	CodegenDeserialize["bool"] = "\t%s = r.bool()\n"
+
+	print ""
+	print "\tbool() {"
+	print "\t\tconst u8 = this.getUint8(this.offset)"
+	print "\t\tthis.offset += 1"
+	print "\t\treturn u8 != 0"
+	print "\t}"
+
+	define_sint("8")
+	define_sint("16")
+	define_sint("32")
+	define_sint("64")
+	define_uint("8")
+	define_uint("16")
+	define_uint("32")
+	define_uint("64")
+
+	print "}"
+}
+
+function codegen_constant(name, value) {
+	print ""
+	print "export const " decapitalize(snaketocamel(name)) " = " value
+}
+
+function codegen_enum_value(name, subname, value, cg) {
+	append(cg, "fields", "\t" snaketocamel(subname) ": " value ",\n")
+}
+
+function codegen_enum(name, cg) {
+	print ""
+	print "export const " name " = Object.freeze({"
+	print cg["fields"] "})"
+
+	CodegenDeserialize[name] = "\t%s = r.i8()\n"
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_struct_field(d, cg,    camel, f, deserialize) {
+	camel = decapitalize(snaketocamel(d["name"]))
+	f = "s." camel
+	append(cg, "fields", "\t" camel "\n")
+
+	deserialize = CodegenDeserialize[d["type"]]
+	if (!d["isarray"]) {
+		append(cg, "deserialize", sprintf(deserialize, f))
+		return
+	}
+
+	append(cg, "deserialize",
+		"\t{\n" \
+		indent(sprintf(CodegenDeserialize["u32"], "const len")))
+	if (d["type"] == "u8") {
+		append(cg, "deserialize",
+			"\t\t" f " = new Uint8Array(\n" \
+			"\t\t\tr.buffer, r.require(len), len)\n" \
+			"\t\tr.offset += len\n" \
+			"\t}\n")
+		return
+	}
+	if (d["type"] == "i8") {
+		append(cg, "deserialize",
+			"\t\t" f " = new Int8Array(\n" \
+			"\t\t\tr.buffer, r.require(len), len)\n" \
+			"\t\tr.offset += len\n" \
+			"\t}\n")
+		return
+	}
+
+	append(cg, "deserialize",
+		"\t\t" f " = new Array(len)\n" \
+		"\t}\n" \
+		"\tfor (let i = 0; i < " f ".length; i++)\n" \
+		indent(sprintf(deserialize, f "[i]")))
+}
+
+function codegen_struct_tag(d, cg) {
+	append(cg, "fields", "\t" decapitalize(snaketocamel(d["name"])) "\n")
+	# Do not deserialize here, that is already done by the containing union.
+}
+
+function codegen_struct(name, cg) {
+	print ""
+	print "export class " name " {"
+	print cg["fields"] cg["methods"]
+	print "\tstatic deserialize(r) {"
+	print "\t\tconst s = new " name "()"
+	print indent(cg["deserialize"]) "\t\treturn s"
+	print "\t}"
+	print "}"
+
+	CodegenDeserialize[name] = "\t%s = " name ".deserialize(r)\n"
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_union_tag(d, cg) {
+	cg["tagtype"] = d["type"]
+	cg["tagname"] = d["name"]
+}
+
+function codegen_union_struct(name, casename, cg, scg,     structname) {
+	append(scg, "methods",
+		"\n" \
+		"\tconstructor() {\n" \
+		"\t\tthis." decapitalize(snaketocamel(cg["tagname"])) \
+			" = " cg["tagtype"] "." snaketocamel(casename) "\n" \
+		"\t}\n")
+
+	# And thus not all generated structs are present in Types.
+	structname = name snaketocamel(casename)
+	codegen_struct(structname, scg)
+
+	append(cg, "deserialize",
+		"\tcase " cg["tagtype"] "." snaketocamel(casename) ":\n" \
+		"\t{\n" \
+		indent(sprintf(CodegenDeserialize[structname], "const s")) \
+		"\t\treturn s\n" \
+		"\t}\n")
+}
+
+function codegen_union(name, cg,    tagvar) {
+	tagvar = decapitalize(snaketocamel(cg["tagname"]))
+
+	print ""
+	print "export function deserialize" name "(r) {"
+	print sprintf(CodegenDeserialize[cg["tagtype"]], "const " tagvar) \
+		"\tswitch (" tagvar ") {"
+	print cg["deserialize"] "\tdefault:"
+	print "\t\tthrow `Unknown " cg["tagtype"] " (${tagvar})`"
+	print "\t}"
+	print "}"
+
+	CodegenDeserialize[name] = "\t%s = deserialize" name "(r)\n"
+	for (i in cg)
+		delete cg[i]
+}

tools/lxdrgen.awk

@@ -0,0 +1,291 @@
+# lxdrgen.awk: an XDR-derived code generator for network protocols.
+#
+# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+#
+# Usage: env LC_ALL=C awk -f lxdrgen.awk -f lxdrgen-{c,go,mjs}.awk \
+#  -v PrefixCamel=Foo foo.lxdr > foo.{c,go,mjs} | {clang-format,gofmt,...}
+
+# --- Utilities ----------------------------------------------------------------
+
+function cameltosnake(s) {
+	while (match(s, /[[:lower:]][[:upper:]]/)) {
+		s = substr(s, 1, RSTART) "_" \
+			tolower(substr(s, RSTART + 1, RLENGTH - 1)) \
+			substr(s, RSTART + RLENGTH)
+	}
+	return tolower(s)
+}
+
+function snaketocamel(s) {
+	s = toupper(substr(s, 1, 1)) tolower(substr(s, 2))
+	while (match(s, /_[[:alnum:]]/)) {
+		s = substr(s, 1, RSTART - 1) \
+			toupper(substr(s, RSTART + 1, RLENGTH - 1)) \
+			substr(s, RSTART + RLENGTH)
+	}
+	return s
+}
+
+function decapitalize(s) {
+	if (match(s, /[[:upper:]][[:lower:]]/)) {
+		return tolower(substr(s, 1, 1)) substr(s, 2)
+	}
+	return s
+}
+
+function indent(s) {
+	if (!s)
+		return s
+
+	gsub(/\n/, "\n\t", s)
+	sub(/\t*$/, "", s)
+	return "\t" s
+}
+
+function append(a, key, value) {
+	a[key] = a[key] value
+}
+
+# --- Parsing ------------------------------------------------------------------
+
+function fatal(message) {
+	print "// " FILENAME ":" FNR ": fatal error: " message
+	print FILENAME ":" FNR ": fatal error: " message > "/dev/stderr"
+	exit 1
+}
+
+function skipcomment() {
+	do {
+		if (match($0, /[*]\//)) {
+			$0 = substr($0, RSTART + RLENGTH)
+			return
+		}
+	} while (getline > 0)
+	fatal("unterminated block comment")
+}
+
+function nexttoken() {
+	do {
+		if (match($0, /^[[:space:]]+/)) {
+			$0 = substr($0, RLENGTH + 1)
+		} else if (match($0, /^\/\/.*/)) {
+			$0 = ""
+		} else if (match($0, /^\/[*]/)) {
+			$0 = substr($0, RLENGTH + 1)
+			skipcomment()
+		} else if (match($0, /^[[:alpha:]][[:alnum:]_]*/)) {
+			Token = substr($0, 1, RLENGTH)
+			$0 = substr($0, RLENGTH + 1)
+			return Token
+		# AWK implementations rarely support non-decimal notations
+		# in their implicit string-to-number conversions.
+		} else if (match($0, /^(0|-?[1-9][0-9]*)/)) {
+			Token = substr($0, 1, RLENGTH)
+			$0 = substr($0, RLENGTH + 1)
+			return Token
+		} else if ($0) {
+			Token = substr($0, 1, 1)
+			$0 = substr($0, 2)
+			return Token
+		}
+	} while ($0 || getline > 0)
+	Token = ""
+	return Token
+}
+
+function expect(v) {
+	if (!v)
+		fatal("broken expectations at `" Token "' before `" $0 "'")
+	return v
+}
+
+function accept(what) {
+	if (Token != what)
+		return 0
+	nexttoken()
+	return 1
+}
+
+function identifier(    v) {
+	if (Token !~ /^[[:alpha:]]/)
+		return 0
+	v = Token
+	nexttoken()
+	return v
+}
+
+function number(    v) {
+	if (Token !~ /^(0|-?[1-9])/)
+		return 0
+	v = Token
+	nexttoken()
+	return v
+}
+
+function readnumber(    ident) {
+	ident = identifier()
+	if (!ident)
+		return expect(number())
+	if (!(ident in Consts))
+		fatal("unknown constant: " ident)
+	return Consts[ident]
+}
+
+function defconst(    ident, num) {
+	if (!accept("const"))
+		return 0
+
+	ident = expect(identifier())
+	expect(accept("="))
+	num = readnumber()
+	if (ident in Consts)
+		fatal("constant redefined: " ident)
+
+	Consts[ident] = num
+	codegen_constant(ident, num)
+	return 1
+}
+
+function readtype(    ident) {
+	ident = deftype()
+	if (ident)
+		return ident
+
+	ident = identifier()
+	if (!ident)
+		return 0
+
+	if (!(ident in Types))
+		fatal("unknown type: " ident)
+	return ident
+}
+
+function defenum(    name, ident, value, cg) {
+	delete cg[0]
+
+	name = expect(identifier())
+	expect(accept("{"))
+	while (!accept("}")) {
+		ident = expect(identifier())
+		value = value + 1
+		if (accept("="))
+			value = readnumber() + 0
+		if (!value)
+			fatal("enumeration values cannot be zero")
+		if (value < -128 || value > 127)
+			fatal("enumeration value out of range")
+		expect(accept(","))
+		append(EnumValues, name, SUBSEP ident)
+		if (EnumValues[name, ident]++)
+			fatal("duplicate enum value: " ident)
+		codegen_enum_value(name, ident, value, cg)
+	}
+
+	Types[name] = "enum"
+	codegen_enum(name, cg)
+	return name
+}
+
+function readfield(out,    nonvoid) {
+	nonvoid = !accept("void")
+	if (nonvoid) {
+		out["type"] = expect(readtype())
+		out["name"] = expect(identifier())
+		# TODO: Consider supporting XDR's VLA length limits here.
+		# TODO: Consider supporting XDR's fixed-length syntax for string limits.
+		out["isarray"] = accept("<") && expect(accept(">"))
+	}
+	expect(accept(";"))
+	return nonvoid
+}
+
+function defstruct(    name, d, cg) {
+	delete d[0]
+	delete cg[0]
+
+	name = expect(identifier())
+	expect(accept("{"))
+	while (!accept("}")) {
+		if (readfield(d))
+			codegen_struct_field(d, cg)
+	}
+
+	Types[name] = "struct"
+	codegen_struct(name, cg)
+	return name
+}
+
+function defunion(    name, tag, tagtype, tagvalue, cg, scg, d, a, i, unseen) {
+	delete cg[0]
+	delete scg[0]
+	delete d[0]
+
+	name = expect(identifier())
+	expect(accept("switch"))
+	expect(accept("("))
+	tag["type"] = tagtype = expect(readtype())
+	tag["name"] = expect(identifier())
+	expect(accept(")"))
+
+	if (Types[tagtype] != "enum")
+		fatal("not an enum type: " tagtype)
+	codegen_union_tag(tag, cg)
+
+	split(EnumValues[tagtype], a, SUBSEP)
+	for (i in a)
+		unseen[a[i]]++
+
+	expect(accept("{"))
+	while (!accept("}")) {
+		if (accept("case")) {
+			if (tagvalue)
+				codegen_union_struct(name, tagvalue, cg, scg)
+
+			tagvalue = expect(identifier())
+			expect(accept(":"))
+			if (!unseen[tagvalue]--)
+				fatal("no such value or duplicate case: " tagtype "." tagvalue)
+			codegen_struct_tag(tag, scg)
+		} else if (tagvalue) {
+			if (readfield(d))
+				codegen_struct_field(d, scg)
+		} else {
+			fatal("union fields must fall under a case")
+		}
+	}
+	if (tagvalue)
+		codegen_union_struct(name, tagvalue, cg, scg)
+
+	# What remains non-zero in unseen[2..] is simply not recognized/allowed.
+	Types[name] = "union"
+	codegen_union(name, cg)
+	return name
+}
+
+function deftype() {
+	if (accept("enum"))
+		return defenum()
+	if (accept("struct"))
+		return defstruct()
+	if (accept("union"))
+		return defunion()
+	return 0
+}
+
+{
+	if (PrefixCamel) {
+		PrefixLower = tolower(cameltosnake(PrefixCamel)) "_"
+		PrefixUpper = toupper(cameltosnake(PrefixCamel)) "_"
+	}
+
+	# This is not in a BEGIN clause (even though it consumes all input),
+	# so that the code generator can insert the first FILENAME.
+	codegen_begin()
+
+	nexttoken()
+	while (Token != "") {
+		expect(defconst() || deftype())
+		expect(accept(";"))
+	}
+}