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

CMakeLists.txt

@@ -1,12 +1,12 @@
 project (liberty C)
-cmake_minimum_required (VERSION 2.8.5)
+cmake_minimum_required (VERSION 2.8.12)
 
 # Moar warnings
 if ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUCC)
 	# -Wunused-function is pretty annoying here, as everything is static
-	set (wdisabled "-Wno-unused-function -Wno-implicit-fallthrough")
+	set (wdisabled "-Wno-unused-function")
 	set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -std=c99 -Wall -Wextra ${wdisabled}")
-endif ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUCC)
+endif ()
 
 # Dependencies
 set (CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake)
@@ -16,11 +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})
@@ -32,15 +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})
+	endif ()
-endforeach (extra)
+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

@@ -1,4 +1,4 @@
-Copyright (c) 2014 - 2018, Přemysl Eric Janouch <p@janouch.name>
+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.

README.adoc

@@ -17,6 +17,42 @@ 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 https://git.janouch.name/p/liberty to report any bugs, request features,

cmake/FindNcursesw.cmake

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

cmake/FindUnistring.cmake

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

@@ -1321,7 +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[] =
@@ -1567,13 +1566,12 @@ mpd_client_parse_line (struct mpd_client *self, const char *line)
 	if (!strcmp (line, "list_OK"))
 		strv_append_owned (&self->data, NULL);
 	else if (mpd_client_parse_response (line, &response))
-	{
 		mpd_client_dispatch (self, &response);
-		free (response.current_command);
-		free (response.message_text);
-	}
 	else
 		strv_append (&self->data, line);
+
+	free (response.current_command);
+	free (response.message_text);
 	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.c

@@ -1,7 +1,7 @@
 /*
  * liberty.c: the ultimate C unlibrary
  *
- * Copyright (c) 2014 - 2018, Přemysl Eric Janouch <p@janouch.name>
+ * 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.
@@ -16,8 +16,13 @@
  *
  */
 
-#define _POSIX_C_SOURCE 199309L
+#ifndef _POSIX_C_SOURCE
+#define _POSIX_C_SOURCE 200112L
+#endif
+
+#ifndef _XOPEN_SOURCE
 #define _XOPEN_SOURCE 600
+#endif
 
 #include <stdio.h>
 #include <stddef.h>
@@ -731,6 +736,21 @@ set_blocking (int fd, bool blocking)
 	return prev;
 }
 
+static bool
+xwrite (int fd, const char *data, size_t len, struct error **e)
+{
+	size_t written = 0;
+	while (written < len)
+	{
+		ssize_t res = write (fd, data + written, len - written);
+		if (res >= 0)
+			written += res;
+		else if (errno != EINTR)
+			return error_set (e, "%s", strerror (errno));
+	}
+	return true;
+}
+
 static void
 xclose (int fd)
 {
@@ -1357,6 +1377,7 @@ struct poller_idle
 
 // - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 
+// The heap could definitely be made faster but we'll prefer simplicity
 struct poller_timers
 {
 	struct poller_timer **heap;         ///< Min-heap of timers
@@ -1733,10 +1754,9 @@ poller_run (struct poller *self)
 	self->revents_len = 0;
 }
 
-#elif defined (BSD)
+// Sort of similar to the epoll version.  Let's hope Darwin isn't broken,
-
+// that'd mean reimplementing this in terms of select() just because of Crapple.
-// Mac OS X's kqueue is fatally broken, or so I've been told; leaving it out.
+#elif defined (BSD) || defined (__APPLE__)
-// Otherwise this is sort of similar to the epoll version.
 
 #include <sys/types.h>
 #include <sys/event.h>
@@ -2435,7 +2455,7 @@ write_queue_processed (struct write_queue *self, size_t len)
 }
 
 static bool
-write_queue_is_empty (struct write_queue *self)
+write_queue_is_empty (const struct write_queue *self)
 {
 	return self->head == NULL;
 }
@@ -2530,7 +2550,7 @@ msg_unpacker_make (const void *data, size_t len)
 }
 
 static size_t
-msg_unpacker_get_available (struct msg_unpacker *self)
+msg_unpacker_get_available (const struct msg_unpacker *self)
 {
 	return self->len - self->offset;
 }
@@ -2672,6 +2692,12 @@ strncasecmp_ascii (const char *a, const char *b, size_t n)
 	return 0;
 }
 
+static bool
+iscntrl_ascii (int c)
+{
+	return (c >= 0 && c < 32) || c == 0x7f;
+}
+
 static bool
 isalpha_ascii (int c)
 {
@@ -2700,63 +2726,67 @@ isspace_ascii (int c)
 
 // --- UTF-8 -------------------------------------------------------------------
 
-/// Return a pointer to the next UTF-8 character, or NULL on error
+/// Return the value of the UTF-8 character at `*s` and advance the pointer
-static const char *
+/// to the next one.  Returns -2 if there is only a partial but possibly valid
-utf8_next (const char *s, size_t len, int32_t *codepoint)
+/// character sequence, or -1 on other errors.  Either way, `*s` is untouched.
+static int32_t
+utf8_decode (const char **s, size_t len)
 {
 	// End of string, we go no further
 	if (!len)
-		return NULL;
+		return -1;
 
 	// Find out how long the sequence is (0 for ASCII)
 	unsigned mask = 0x80;
 	unsigned sequence_len = 0;
 
-	const uint8_t *p = (const uint8_t *) s;
+	const uint8_t *p = (const uint8_t *) *s, *end = p + len;
 	while ((*p & mask) == mask)
 	{
 		// Invalid start of sequence
 		if (mask == 0xFE)
-			return NULL;
+			return -1;
 
 		mask |= mask >> 1;
 		sequence_len++;
 	}
 
-	// In the middle of a character or the input is too short
+	// In the middle of a character
-	if (sequence_len == 1 || sequence_len > len)
+	// or an overlong sequence (subset, possibly MUTF-8, not supported)
-		return NULL;
+	if (sequence_len == 1 || *p == 0xC0 || *p == 0xC1)
+		return -1;
 
 	// Check the rest of the sequence
 	uint32_t cp = *p++ & ~mask;
 	while (sequence_len && --sequence_len)
 	{
+		if (p == end)
+			return -2;
 		if ((*p & 0xC0) != 0x80)
-			return NULL;
+			return -1;
 		cp = cp << 6 | (*p++ & 0x3F);
 	}
-	if (codepoint)
+	*s = (const char *) p;
-		*codepoint = cp;
+	return cp;
-	return (const char *) p;
+}
+
+static inline bool
+utf8_validate_cp (int32_t cp)
+{
+	// RFC 3629, CESU-8 not allowed
+	return cp >= 0 && cp <= 0x10FFFF && (cp < 0xD800 || cp > 0xDFFF);
 }
 
 /// Very rough UTF-8 validation, just makes sure codepoints can be iterated
 static bool
 utf8_validate (const char *s, size_t len)
 {
-	const char *next;
+	const char *end = s + len;
-	while (len)
-	{
 	int32_t codepoint;
-		// TODO: better validations
+	while ((codepoint = utf8_decode (&s, end - s)) >= 0
-		if (!(next = utf8_next (s, len, &codepoint))
+		&& utf8_validate_cp (codepoint))
-		 || codepoint > 0x10FFFF)
+		;
-			return false;
+	return s == end;
-
-		len -= next - s;
-		s = next;
-	}
-	return true;
 }
 
 // - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
@@ -2780,12 +2810,12 @@ utf8_iter_next (struct utf8_iter *self, size_t *len)
 		return -1;
 
 	const char *old = self->s;
-	int32_t codepoint;
+	int32_t codepoint = utf8_decode (&self->s, self->len);
-	if (!soft_assert ((self->s = utf8_next (old, self->len, &codepoint))))
+	if (!soft_assert (codepoint >= 0))
 	{
 		// Invalid UTF-8
 		self->len = 0;
-		return -1;
+		return codepoint;
 	}
 
 	size_t advance = self->s - old;
@@ -2917,6 +2947,13 @@ base64_encode (const void *data, size_t len, struct str *output)
 
 // --- Utilities ---------------------------------------------------------------
 
+static void
+cstr_set (char **s, char *new)
+{
+	free (*s);
+	*s = new;
+}
+
 static void
 cstr_split (const char *s, const char *delimiters, bool ignore_empty,
 	struct strv *out)
@@ -2946,10 +2983,10 @@ cstr_strip_in_place (char *s, const char *stripped_chars)
 }
 
 static void
-cstr_transform (char *s, int (*tolower) (int c))
+cstr_transform (char *s, int (*xform) (int c))
 {
 	for (; *s; s++)
-		*s = tolower (*s);
+		*s = xform (*s);
 }
 
 static char *
@@ -2996,6 +3033,7 @@ iconv_xstrdup (iconv_t conv, char *in, size_t in_len, size_t *out_len)
 
 	char *in_ptr = in;
 	if (in_len == (size_t) -1)
+		// XXX: out_len will be one character longer than the string!
 		in_len = strlen (in) + 1;
 
 	while (iconv (conv, (char **) &in_ptr, &in_len,
@@ -3213,7 +3251,7 @@ get_xdg_config_dirs (struct strv *out)
 	str_free (&config_home);
 
 	const char *xdg_config_dirs;
-	if (!(xdg_config_dirs = getenv ("XDG_CONFIG_DIRS")))
+	if (!(xdg_config_dirs = getenv ("XDG_CONFIG_DIRS")) || !*xdg_config_dirs)
 		xdg_config_dirs = "/etc/xdg";
 	cstr_split (xdg_config_dirs, ":", true, out);
 }
@@ -3238,7 +3276,7 @@ get_xdg_data_dirs (struct strv *out)
 	str_free (&data_home);
 
 	const char *xdg_data_dirs;
-	if (!(xdg_data_dirs = getenv ("XDG_DATA_DIRS")))
+	if (!(xdg_data_dirs = getenv ("XDG_DATA_DIRS")) || !*xdg_data_dirs)
 		xdg_data_dirs = "/usr/local/share/:/usr/share/";
 	cstr_split (xdg_data_dirs, ":", true, out);
 }
@@ -3255,16 +3293,8 @@ resolve_relative_data_filename (const char *filename)
 }
 
 static char *
-resolve_relative_runtime_filename (const char *filename)
+resolve_relative_runtime_filename_finish (struct str path)
 {
-	struct str path = str_make ();
-	const char *runtime_dir = getenv ("XDG_RUNTIME_DIR");
-	if (runtime_dir && *runtime_dir == '/')
-		str_append (&path, runtime_dir);
-	else
-		get_xdg_home_dir (&path, "XDG_DATA_HOME", ".local/share");
-	str_append_printf (&path, "/%s/%s", PROGRAM_NAME, filename);
-
 	// Try to create the file's ancestors;
 	// typically the user will want to immediately create a file in there
 	const char *last_slash = strrchr (path.str, '/');
@@ -3277,6 +3307,41 @@ resolve_relative_runtime_filename (const char *filename)
 	return str_steal (&path);
 }
 
+static char *
+resolve_relative_runtime_filename (const char *filename)
+{
+	struct str path = str_make ();
+	const char *runtime_dir = getenv ("XDG_RUNTIME_DIR");
+	if (runtime_dir && *runtime_dir == '/')
+		str_append (&path, runtime_dir);
+	else
+		get_xdg_home_dir (&path, "XDG_DATA_HOME", ".local/share");
+
+	str_append_printf (&path, "/%s/%s", PROGRAM_NAME, filename);
+	return resolve_relative_runtime_filename_finish (path);
+}
+
+/// This differs from resolve_relative_runtime_filename() in that we expect
+/// the filename to be something like a pattern for mkstemp(), so the resulting
+/// path can reside in a system-wide directory with no risk of a conflict.
+/// However, we have to take care about permissions.  Do we even need this?
+static char *
+resolve_relative_runtime_template (const char *template)
+{
+	struct str path = str_make ();
+	const char *runtime_dir = getenv ("XDG_RUNTIME_DIR");
+	const char *tmpdir = getenv ("TMPDIR");
+	if (runtime_dir && *runtime_dir == '/')
+		str_append_printf (&path, "%s/%s", runtime_dir, PROGRAM_NAME);
+	else if (tmpdir && *tmpdir == '/')
+		str_append_printf (&path, "%s/%s.%d", tmpdir, PROGRAM_NAME, geteuid ());
+	else
+		str_append_printf (&path, "/tmp/%s.%d", PROGRAM_NAME, geteuid ());
+
+	str_append_printf (&path, "/%s", template);
+	return resolve_relative_runtime_filename_finish (path);
+}
+
 static char *
 try_expand_tilde (const char *filename)
 {
@@ -3729,7 +3794,7 @@ opt_handler_usage (const struct opt_handler *self, FILE *stream)
 		str_append_printf (&row, "--%s", opt->long_name);
 		if (opt->arg_hint)
 			str_append_printf (&row, (opt->flags & OPT_OPTIONAL_ARG)
-				? " [%s]" : " %s", opt->arg_hint);
+				? "[=%s]" : " %s", opt->arg_hint);
 
 		// TODO: keep the indent if there are multiple lines
 		if (row.len + 2 <= OPT_USAGE_ALIGNMENT_COLUMN)
@@ -4352,11 +4417,13 @@ socket_io_try_write (int socket_fd, struct str *wb)
 //   object  = lws '{' entries endobj
 //   endobj  = lws '}'
 //
-//   string  = lws '"' ('\\' escape / ![\\"] char)* '"'
+//   quoted  = lws '"' (!["\\] char / '\\' escape)* '"'
+//           / lws '`' (![`] char)* '`'
+//   string  = (quoted)+
 //   char    = [\0-\177]  # or any Unicode codepoint in the UTF-8 encoding
 //   escape  = [\\"abfnrtv] / [xX][0-9A-Fa-f][0-9A-Fa-f]? / [0-7][0-7]?[0-7]?
 //
-//   integer = lws '-'? [0-9]+  # whatever strtoll() accepts on your system
+//   integer = lws [-+]? [0-9]+  # whatever strtoll() accepts on your system
 //   null    = lws 'null'
 //   boolean = lws 'yes'  / lws 'YES'  / lws 'no'    / lws 'NO'
 //           / lws 'on'   / lws 'ON'   / lws 'off'   / lws 'OFF'
@@ -4652,8 +4719,10 @@ config_item_write_string (struct str *output, const struct str *s)
 		else if (c == '\t')  str_append (output, "\\t");
 		else if (c == '\\')  str_append (output, "\\\\");
 		else if (c == '"')   str_append (output, "\\\"");
-		else if (c < 32)     str_append_printf (output, "\\x%02x", c);
+		else if (iscntrl_ascii (c))
-		else                 str_append_c      (output, c);
+			str_append_printf (output, "\\x%02x", c);
+		else
+			str_append_c (output, c);
 	}
 	str_append_c (output, '"');
 }
@@ -4977,10 +5046,10 @@ config_tokenizer_escape_sequence
 }
 
 static bool
-config_tokenizer_string
+config_tokenizer_dq_string (struct config_tokenizer *self, struct str *output,
-	(struct config_tokenizer *self, struct str *output, struct error **e)
+	struct error **e)
 {
-	unsigned char c;
+	unsigned char c = config_tokenizer_advance (self);
 	while (self->len)
 	{
 		if ((c = config_tokenizer_advance (self)) == '"')
@@ -4994,6 +5063,44 @@ config_tokenizer_string
 	return false;
 }
 
+static bool
+config_tokenizer_bt_string (struct config_tokenizer *self, struct str *output,
+	struct error **e)
+{
+	unsigned char c = config_tokenizer_advance (self);
+	while (self->len)
+	{
+		if ((c = config_tokenizer_advance (self)) == '`')
+			return true;
+		str_append_c (output, c);
+	}
+	config_tokenizer_error (self, e, "premature end of string");
+	return false;
+}
+
+static bool
+config_tokenizer_string (struct config_tokenizer *self, struct str *output,
+	struct error **e)
+{
+	// Go-like strings, with C/AWK-like automatic concatenation
+	while (self->len)
+	{
+		bool ok = true;
+		if (isspace_ascii (*self->p) && *self->p != '\n')
+			config_tokenizer_advance (self);
+		else if (*self->p == '"')
+			ok = config_tokenizer_dq_string (self, output, e);
+		else if (*self->p == '`')
+			ok = config_tokenizer_bt_string (self, output, e);
+		else
+			break;
+
+		if (!ok)
+			return false;
+	}
+	return true;
+}
+
 static enum config_token
 config_tokenizer_next (struct config_tokenizer *self, struct error **e)
 {
@@ -5018,7 +5125,7 @@ config_tokenizer_next (struct config_tokenizer *self, struct error **e)
 		return CONFIG_T_ABORT;
 
 	case '"':
-		config_tokenizer_advance (self);
+	case '`':
 		str_reset (&self->string);
 		if (!config_tokenizer_string (self, &self->string, e))
 			return CONFIG_T_ABORT;
@@ -5030,18 +5137,21 @@ config_tokenizer_next (struct config_tokenizer *self, struct error **e)
 		return CONFIG_T_STRING;
 	}
 
-	char *end;
+	// Our input doesn't need to be NUL-terminated but we want to use strtoll()
+	char buf[48] = "", *end = buf;
+	size_t buf_len = MIN (sizeof buf - 1, self->len);
+
 	errno = 0;
-	self->integer = strtoll (self->p, &end, 10);
+	self->integer = strtoll (strncpy (buf, self->p, buf_len), &end, 10);
 	if (errno == ERANGE)
 	{
 		config_tokenizer_error (self, e, "integer out of range");
 		return CONFIG_T_ABORT;
 	}
-	if (end != self->p)
+	if (end != buf)
 	{
-		self->len -= end - self->p;
+		self->len -= end - buf;
-		self->p = end;
+		self->p += end - buf;
 		return CONFIG_T_INTEGER;
 	}
 
@@ -5054,7 +5164,7 @@ config_tokenizer_next (struct config_tokenizer *self, struct error **e)
 	str_reset (&self->string);
 	do
 		str_append_c (&self->string, config_tokenizer_advance (self));
-	while (config_tokenizer_is_word_char (*self->p));
+	while (self->len && config_tokenizer_is_word_char (*self->p));
 
 	if (!strcmp (self->string.str, "null"))
 		return CONFIG_T_NULL;

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

siphash.c

@@ -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

@@ -1,7 +1,7 @@
 /*
  * tests/liberty.c
  *
- * Copyright (c) 2015 - 2016, Přemysl Eric Janouch <p@janouch.name>
+ * 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.
@@ -326,10 +326,19 @@ test_str_map (void)
 static void
 test_utf8 (void)
 {
-	const char valid  [] = "2H₂ + O₂ ⇌ 2H₂O, R = 4.7 kΩ, ⌀ 200 mm";
+	const char *full = "\xc5\x99", *partial = full, *empty = full;
-	const char invalid[] = "\xf0\x90\x28\xbc";
+	soft_assert (utf8_decode (&full,    2) == 0x0159);
-	soft_assert ( utf8_validate (valid,   sizeof valid));
+	soft_assert (utf8_decode (&partial, 1) == -2);
-	soft_assert (!utf8_validate (invalid, sizeof invalid));
+	soft_assert (utf8_decode (&empty,   0) == -1);
+
+	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;
@@ -604,6 +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
@@ -622,6 +706,7 @@ main (int argc, char *argv[])
 	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

@@ -77,7 +77,7 @@ test_irc (void)
 static void
 test_http_parser (void)
 {
-	struct str_map parameters = str_map_make (NULL);
+	struct str_map parameters = str_map_make (free);
 	parameters.key_xfrm = tolower_ascii_strxfrm;
 
 	char *type = NULL;
@@ -88,9 +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"));

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(";"))
+	}
+}