liberty-xdg: add desktop entry parser tests

And fix a discovered bug.

CMakeLists.txt

@@ -1,5 +1,5 @@
-project (liberty C)
+cmake_minimum_required (VERSION 3.0...3.27)
-cmake_minimum_required (VERSION 2.8.12)
+project (liberty C CXX)
 
 # Moar warnings
 if ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUCC)
@@ -29,14 +29,14 @@ link_directories (${libssl_LIBRARY_DIRS})
 foreach (extra iconv rt)
 	find_library (extra_lib_${extra} ${extra})
 	if (extra_lib_${extra})
-		list (APPEND common_libraries ${extra})
+		list (APPEND common_libraries ${extra_lib_${extra}})
 	endif ()
 endforeach ()
 
 # Build some unit tests
 include_directories (${PROJECT_SOURCE_DIR})
 enable_testing ()
-set (tests liberty proto)
+set (tests liberty proto xdg)
 
 pkg_check_modules (libpulse libpulse)
 if (libpulse_FOUND)
@@ -72,7 +72,7 @@ add_test (test-cmake-parser
 # Test protocol code generation
 set (lxdrgen_outputs)
 set (lxdrgen_base "${PROJECT_BINARY_DIR}/lxdrgen.lxdr")
-foreach (backend c go mjs)
+foreach (backend c cpp go mjs swift)
 	list (APPEND lxdrgen_outputs ${lxdrgen_base}.${backend})
 	add_custom_command (OUTPUT ${lxdrgen_base}.${backend}
 		COMMAND env LC_ALL=C awk
@@ -91,9 +91,21 @@ 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)
+add_executable (test-lxdrgen-c tests/lxdrgen.c ${lxdrgen_base}.c)
-target_include_directories (test-lxdrgen PUBLIC ${PROJECT_BINARY_DIR})
+target_include_directories (test-lxdrgen-c PUBLIC ${PROJECT_BINARY_DIR})
-add_test (NAME test-lxdrgen-c COMMAND test-lxdrgen)
+add_test (NAME test-lxdrgen-c COMMAND test-lxdrgen-c)
+
+set_source_files_properties (${lxdrgen_base}.cpp
+	PROPERTIES HEADER_FILE_ONLY TRUE)
+if (WIN32)
+	add_executable (test-lxdrgen-cpp tests/lxdrgen.cpp
+		${lxdrgen_base}.cpp tools/lxdrgen-cpp-win32.cpp)
+else ()
+	add_executable (test-lxdrgen-cpp tests/lxdrgen.cpp
+		${lxdrgen_base}.cpp tools/lxdrgen-cpp-posix.cpp)
+endif ()
+target_include_directories (test-lxdrgen-cpp PUBLIC ${PROJECT_BINARY_DIR})
+add_test (NAME test-lxdrgen-cpp COMMAND test-lxdrgen-cpp)
 
 find_program (GO_EXECUTABLE go)
 if (GO_EXECUTABLE)
@@ -108,3 +120,11 @@ if (NODE_EXECUTABLE)
 else ()
 	message (WARNING "Cannot test generated protocol code for Javascript")
 endif ()
+
+find_program (SWIFTC_EXECUTABLE swiftc)
+if (SWIFTC_EXECUTABLE)
+	add_test (test-lxdrgen-swift
+		${SWIFTC_EXECUTABLE} -typecheck ${lxdrgen_base}.swift)
+else ()
+	message (WARNING "Cannot test generated protocol code for Swift")
+endif ()

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2014 - 2022, Přemysl Eric Janouch <p@janouch.name>
+Copyright (c) 2014 - 2024, 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

@@ -44,6 +44,12 @@ lxdrgen.awk::
 lxdrgen-c.awk::
 	LibertyXDR backend that builds on top of the C pseudolibrary.
 
+lxdrgen-cpp.awk::
+lxdrgen-cpp-win32.cpp::
+lxdrgen-cpp-posix.cpp::
+	LibertyXDR backend for C++, primarily targeting Win32 and its wide strings.
+	Link the result together with one of the accompanied source files.
+
 lxdrgen-go.awk::
 	LibertyXDR backend for Go, supporting _encoding/json_ interfaces.  It also
 	produces optimized JSON marshallers (however, note that the _json.Marshaler_
@@ -54,6 +60,9 @@ lxdrgen-mjs.awk::
 	LibertyXDR backend for Javascript, currently for decoding only.
 	It cuts a corner by not using BigInts, on par with `JSON.parse()`.
 
+lxdrgen-swift.awk::
+	LibertyXDR backend for the Swift programming language.
+
 Contributing and Support
 ------------------------
 Use https://git.janouch.name/p/liberty to report any bugs, request features,

cmake/AddThreads.cmake

@@ -9,15 +9,15 @@ find_package (Threads)
 function (add_threads target)
 	if (NOT Threads_FOUND OR NOT CMAKE_USE_PTHREADS_INIT)
 		message (FATAL_ERROR "pthreads not found")
-	endif (NOT Threads_FOUND OR NOT CMAKE_USE_PTHREADS_INIT)
+	endif ()
 
 	if (THREADS_HAVE_PTHREAD_ARG)
 		set_property (TARGET ${target} PROPERTY
 			COMPILE_OPTIONS "-pthread")
 		set_property (TARGET ${target} PROPERTY
 			INTERFACE_COMPILE_OPTIONS "-pthread")
-	endif (THREADS_HAVE_PTHREAD_ARG)
+	endif ()
 	if (CMAKE_THREAD_LIBS_INIT)
 		target_link_libraries (${target} "${CMAKE_THREAD_LIBS_INIT}")
-	endif (CMAKE_THREAD_LIBS_INIT)
+	endif ()
-endfunction (add_threads)
+endfunction ()

cmake/FindNcursesw.cmake

@@ -7,7 +7,7 @@ pkg_check_modules (Ncursesw QUIET ncursesw)
 set (required_vars Ncursesw_LIBRARIES)
 if (NOT Ncursesw_FOUND)
 	find_library (Ncursesw_LIBRARIES NAMES ncursesw)
-	find_path (Ncursesw_INCLUDE_DIRS ncurses.h)
+	find_path (Ncursesw_INCLUDE_DIRS ncurses.h PATH_SUFFIXES ncurses)
 	list (APPEND required_vars Ncursesw_INCLUDE_DIRS)
 endif (NOT Ncursesw_FOUND)
 

cmake/IconUtils.cmake

@@ -0,0 +1,40 @@
+# Public Domain
+
+function (icon_to_png name svg size output_dir output)
+	set (_dimensions "${size}x${size}")
+	set (_png_path "${output_dir}/hicolor/${_dimensions}/apps")
+	set (_png "${_png_path}/${name}.png")
+	set (${output} "${_png}" PARENT_SCOPE)
+
+	set (_find_program_REQUIRE)
+	if (NOT ${CMAKE_VERSION} VERSION_LESS 3.18.0)
+		set (_find_program_REQUIRE REQUIRED)
+	endif ()
+
+	find_program (rsvg_convert_EXECUTABLE rsvg-convert ${_find_program_REQUIRE})
+	add_custom_command (OUTPUT "${_png}"
+		COMMAND ${CMAKE_COMMAND} -E make_directory "${_png_path}"
+		COMMAND ${rsvg_convert_EXECUTABLE} "--output=${_png}"
+			"--width=${size}" "--height=${size}" -- "${svg}"
+		DEPENDS "${svg}"
+		COMMENT "Generating ${name} ${_dimensions} application icon" VERBATIM)
+endfunction ()
+
+# You should include a 256x256 icon--which takes less space as raw PNG.
+function (icon_for_win32 ico pngs pngs_raw)
+	set (_raws)
+	foreach (png ${pngs_raw})
+		list (APPEND _raws "--raw=${png}")
+	endforeach ()
+
+	set (_find_program_REQUIRE)
+	if (NOT ${CMAKE_VERSION} VERSION_LESS 3.18.0)
+		set (_find_program_REQUIRE REQUIRED)
+	endif ()
+
+	find_program (icotool_EXECUTABLE icotool ${_find_program_REQUIRE})
+	add_custom_command (OUTPUT "${ico}"
+		COMMAND ${icotool_EXECUTABLE} -c -o "${ico}" ${_raws} -- ${pngs}
+		DEPENDS ${pngs} ${pngs_raw}
+		COMMENT "Generating Windows program icon" VERBATIM)
+endfunction ()

liberty-proto.c

@@ -1911,14 +1911,14 @@ mpd_client_connect_unix (struct mpd_client *self, const char *address,
 	// Expand tilde if needed
 	char *expanded = resolve_filename (address, xstrdup);
 
-	struct sockaddr_un sun;
+	struct sockaddr_un sau;
-	sun.sun_family = AF_UNIX;
+	sau.sun_family = AF_UNIX;
-	strncpy (sun.sun_path, expanded, sizeof sun.sun_path);
+	strncpy (sau.sun_path, expanded, sizeof sau.sun_path);
-	sun.sun_path[sizeof sun.sun_path - 1] = 0;
+	sau.sun_path[sizeof sau.sun_path - 1] = 0;
 
 	free (expanded);
 
-	if (connect (fd, (struct sockaddr *) &sun, sizeof sun))
+	if (connect (fd, (struct sockaddr *) &sau, sizeof sau))
 	{
 		error_set (e, "%s: %s", "connect", strerror (errno));
 		xclose (fd);

liberty-tui.c

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

liberty-xdg.c

@@ -0,0 +1,772 @@
+/*
+ * liberty-xdg.c: the ultimate C unlibrary: freedesktop.org specifications
+ *
+ * Copyright (c) 2023 - 2024, Přemysl Eric Janouch <p@janouch.name>
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ */
+
+// This files assumes you've already included liberty.c.
+
+#ifdef LIBERTY_XDG_WANT_X11
+#include <X11/Xatom.h>
+#include <X11/Xlib.h>
+#endif
+
+#ifdef LIBERTY_XDG_WANT_ICONS
+#include <png.h>
+#endif
+
+// --- XSettings ---------------------------------------------------------------
+
+#ifdef LIBERTY_XDG_WANT_X11
+
+struct xdg_xsettings_setting
+{
+	enum xdg_xsettings_type
+	{
+		XDG_XSETTINGS_INTEGER,
+		XDG_XSETTINGS_STRING,
+		XDG_XSETTINGS_COLOR,
+	}
+	type;                               ///< What's stored in the union
+	uint32_t serial;                    ///< Serial of the last change
+	union
+	{
+		int32_t integer;
+		struct str string;
+		struct { uint16_t red, green, blue, alpha; } color;
+	};
+};
+
+static void
+xdg_xsettings_setting_destroy (struct xdg_xsettings_setting *self)
+{
+	if (self->type == XDG_XSETTINGS_STRING)
+		str_free (&self->string);
+	free (self);
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+struct xdg_xsettings
+{
+	struct str_map settings;            ///< Name -> xdg_xsettings_setting
+};
+
+static void
+xdg_xsettings_free (struct xdg_xsettings *self)
+{
+	str_map_free (&self->settings);
+}
+
+static struct xdg_xsettings
+xdg_xsettings_make (void)
+{
+	return (struct xdg_xsettings)
+	{
+		.settings =
+			str_map_make ((str_map_free_fn) xdg_xsettings_setting_destroy),
+	};
+}
+
+static void
+xdg_xsettings_update (struct xdg_xsettings *self, Display *dpy)
+{
+	// TODO: We're supposed to lock the server.
+	// TODO: We're supposed to trap X errors.
+	char *selection = xstrdup_printf ("_XSETTINGS_S%d", DefaultScreen (dpy));
+	Window owner
+		= XGetSelectionOwner (dpy, XInternAtom (dpy, selection, True));
+	free (selection);
+	if (!owner)
+		return;
+
+	Atom actual_type = None;
+	int actual_format = 0;
+	unsigned long nitems = 0, bytes_after = 0;
+	unsigned char *buffer = NULL;
+	Atom xsettings = XInternAtom (dpy, "_XSETTINGS_SETTINGS", True);
+	int status = XGetWindowProperty (dpy,
+		owner,
+		xsettings,
+		0L,
+		LONG_MAX,
+		False,
+		xsettings,
+		&actual_type,
+		&actual_format,
+		&nitems,
+		&bytes_after,
+		&buffer);
+	if (status != Success || !buffer)
+		return;
+
+	if (actual_type != xsettings
+	 || actual_format != 8
+	 || nitems < 12)
+		goto fail;
+
+	const struct peeker *peeker = NULL;
+	if (buffer[0] == LSBFirst)
+		peeker = &peeker_le;
+	else if (buffer[0] == MSBFirst)
+		peeker = &peeker_be;
+	else
+		goto fail;
+
+	// We're ignoring the serial for now.
+	uint32_t n_settings = peeker->u32 (buffer + 8);
+	size_t offset = 12;
+	struct str name = str_make ();
+	struct xdg_xsettings_setting *setting = xcalloc (1, sizeof *setting);
+	while (n_settings--)
+	{
+		if (nitems < offset + 4)
+			goto fail_item;
+
+		setting->type = buffer[offset];
+		uint16_t name_len = peeker->u16 (buffer + offset + 2);
+		offset += 4;
+		if (nitems < offset + name_len)
+			goto fail_item;
+
+		str_append_data (&name, buffer + offset, name_len);
+		offset += ((name_len + 3) & ~3);
+		if (nitems < offset + 4)
+			goto fail_item;
+
+		setting->serial = peeker->u32 (buffer + offset);
+		offset += 4;
+		switch (setting->type)
+		{
+		case XDG_XSETTINGS_INTEGER:
+			if (nitems < offset + 4)
+				goto fail_item;
+
+			setting->integer = (int32_t) peeker->u32 (buffer + offset);
+			offset += 4;
+			break;
+		case XDG_XSETTINGS_STRING:
+		{
+			if (nitems < offset + 4)
+				goto fail_item;
+
+			uint32_t value_len = peeker->u32 (buffer + offset);
+			offset += 4;
+			if (nitems < offset + value_len)
+				goto fail_item;
+
+			setting->string = str_make ();
+			str_append_data (&setting->string, buffer + offset, value_len);
+			offset += ((value_len + 3) & ~3);
+			break;
+		}
+		case XDG_XSETTINGS_COLOR:
+			if (nitems < offset + 8)
+				goto fail_item;
+
+			setting->color.red   = peeker->u16 (buffer + offset);
+			setting->color.green = peeker->u16 (buffer + offset + 2);
+			setting->color.blue  = peeker->u16 (buffer + offset + 4);
+			setting->color.alpha = peeker->u16 (buffer + offset + 6);
+			offset += 8;
+			break;
+		default:
+			goto fail_item;
+		}
+
+		// TODO(p): Change detection, by comparing existence and serials.
+		str_map_set (&self->settings, name.str, setting);
+		setting = xcalloc (1, sizeof *setting);
+		str_reset (&name);
+	}
+fail_item:
+	xdg_xsettings_setting_destroy (setting);
+	str_free (&name);
+fail:
+	XFree (buffer);
+}
+
+#endif // LIBERTY_XDG_WANT_X11
+
+// --- Desktop file parser -----------------------------------------------------
+
+// Useful for parsing desktop-entry-spec, icon-theme-spec, trash-spec,
+// mime-apps-spec.  This code is not designed for making changes to the files.
+
+struct desktop_file
+{
+	struct str_map groups;              ///< Group name → Key → Value
+};
+
+static void
+desktop_file_free_group (void *value)
+{
+	str_map_free (value);
+	free (value);
+}
+
+static void
+desktop_file_free (struct desktop_file *self)
+{
+	str_map_free (&self->groups);
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+static void
+desktop_file_parse_line (struct desktop_file *self,
+	char **group_name, const char *line, const char *end)
+{
+	struct str_map *group = NULL;
+	if (*group_name)
+		group = str_map_find (&self->groups, *group_name);
+
+	if (*line == '[')
+	{
+		bool ok = *--end == ']';
+		for (const char *p = ++line; ok && p != end; p++)
+			ok = (unsigned char) *p >= 32 && (unsigned char) *p <= 127
+				&& *p != '[' && *p != ']';
+		if (!ok)
+		{
+			cstr_set (group_name, NULL);
+			print_debug ("invalid desktop file group header");
+			return;
+		}
+
+		cstr_set (group_name, xstrndup (line, end - line));
+		if (str_map_find (&self->groups, *group_name))
+		{
+			print_debug ("duplicate desktop file group: %s", *group_name);
+			return;
+		}
+
+		group = xcalloc (1, sizeof *group);
+		*group = str_map_make (free);
+		str_map_set (&self->groups, *group_name, group);
+		return;
+	}
+	if (!group)
+	{
+		print_debug ("unexpected desktop file entry outside of a group");
+		return;
+	}
+
+	const char *key_end = line;
+	while (key_end != end && (isalnum_ascii (*key_end) || *key_end == '-'))
+		key_end++;
+
+	// We could validate these further, but we just search in them anyway.
+	if (key_end != end && *key_end == '[')
+	{
+		while (++key_end != end && *key_end != ']')
+			;
+		if (key_end != end && *key_end == ']')
+			key_end++;
+	}
+
+	const char *value = key_end;
+	while (value != end && *value == ' ')
+		value++;
+	if (value == end || *value++ != '=')
+	{
+		print_debug ("invalid desktop file entry");
+		return;
+	}
+	while (value != end && *value == ' ')
+		value++;
+
+	char *key = xstrndup (line, key_end - line);
+	if (str_map_find (group, key))
+		print_debug ("duplicate desktop file entry for: %s", key);
+	else
+		str_map_set (group, key, xstrndup (value, end - value));
+	free (key);
+}
+
+static struct desktop_file
+desktop_file_make (const char *data, size_t len)
+{
+	struct desktop_file self = (struct desktop_file)
+		{ .groups = str_map_make (desktop_file_free_group) };
+
+	char *group_name = NULL;
+	const char *p = data, *data_end = p + len;
+	while (p != data_end)
+	{
+		const char *line = p, *line_end = line;
+		while (line_end != data_end && *line_end != '\n')
+			line_end++;
+		if ((p = line_end) != data_end && *p == '\n')
+			p++;
+
+		if (line != line_end && *line != '#')
+			desktop_file_parse_line (&self, &group_name, line, line_end);
+	}
+	free (group_name);
+	return self;
+}
+
+static const char *
+desktop_file_get (struct desktop_file *self, const char *group, const char *key)
+{
+	// TODO(p): Ideally, also implement localised keys.
+	struct str_map *group_map = str_map_find (&self->groups, group);
+	if (!group_map)
+		return NULL;
+
+	return str_map_find (group_map, key);
+}
+
+static struct strv
+desktop_file_unescape (const char *value, bool is_list)
+{
+	struct strv result = strv_make ();
+	struct str s = str_make ();
+
+	// XXX: The unescaping behaviour is underspecified.
+	//   It might make sense to warn about unrecognised escape sequences.
+	bool escape = false;
+	for (const char *p = value; *p; p++)
+	{
+		if (escape)
+		{
+			switch (*p)
+			{
+			break; case 's': str_append_c (&s, ' ');
+			break; case 'n': str_append_c (&s, '\n');
+			break; case 't': str_append_c (&s, '\t');
+			break; case 'r': str_append_c (&s, '\r');
+			break; default:  str_append_c (&s, *p);
+			}
+			escape = false;
+		}
+		else if (*p == '\\' && p[1])
+			escape = true;
+		else if (*p == ';' && is_list)
+		{
+			strv_append_owned (&result, str_steal (&s));
+			s = str_make ();
+		}
+		else
+			str_append_c (&s, *p);
+	}
+
+	if (!is_list || s.len != 0)
+		strv_append_owned (&result, str_steal (&s));
+	else
+		str_free (&s);
+	return result;
+}
+
+static char *
+desktop_file_get_string (struct desktop_file *self,
+	const char *group, const char *key)
+{
+	const char *value = desktop_file_get (self, group, key);
+	if (!value)
+		return NULL;
+
+	struct strv values = desktop_file_unescape (value, false /* is_list */);
+	char *unescaped = strv_steal (&values, 0);
+	strv_free (&values);
+	return unescaped;
+}
+
+static struct strv
+desktop_file_get_stringv (struct desktop_file *self,
+	const char *group, const char *key)
+{
+	const char *value = desktop_file_get (self, group, key);
+	if (!value)
+		return strv_make ();
+
+	return desktop_file_unescape (value, true /* is_list */);
+}
+
+static bool
+desktop_file_get_bool (struct desktop_file *self,
+	const char *group, const char *key)
+{
+	const char *value = desktop_file_get (self, group, key);
+	if (!value)
+		return false;
+
+	// Let's be compatible with pre-1.0 files when it costs us so little.
+	if (!strcmp (value, "true")
+	 || !strcmp (value, "1"))
+		return true;
+	if (!strcmp (value, "false")
+	 || !strcmp (value, "0"))
+		return false;
+
+	print_debug ("invalid desktop file boolean for '%s': %s", key, value);
+	return false;
+}
+
+// Nothing uses the "numeric" type.
+// "icon-theme-spec" uses "integer" and doesn't say what it is.
+static long
+desktop_file_get_integer (struct desktop_file *self,
+	const char *group, const char *key)
+{
+	const char *value = desktop_file_get (self, group, key);
+	if (!value)
+		return 0;
+
+	char *end = NULL;
+	long parsed = (errno = 0, strtol (value, &end, 10));
+	if (errno != 0 || *end)
+		print_debug ("invalid desktop file integer for '%s': %s", key, value);
+	return parsed;
+}
+
+// --- Icon themes -------------------------------------------------------------
+
+// This implements part of the Icon Theme Specification.
+
+#ifdef LIBERTY_XDG_WANT_ICONS
+
+struct icon_theme_icon
+{
+	uint32_t width;                     ///< Width of argb in pixels
+	uint32_t height;                    ///< Height of argb in pixels
+	uint32_t argb[];                    ///< ARGB32 data, unassociated alpha
+};
+
+static void
+icon_theme_open_on_error (png_structp pngp, const char *error)
+{
+	print_debug ("%s: %s", (const char *) png_get_error_ptr (pngp), error);
+	png_longjmp (pngp, 1);
+}
+
+static void
+icon_theme_open_on_warning (png_structp pngp, const char *warning)
+{
+	(void) pngp;
+	(void) warning;
+	// Fuck your "gamma value does not match libpng estimate".
+}
+
+// For simplicity, only support PNG icons, using the most popular library.
+static struct icon_theme_icon *
+icon_theme_open (const char *path)
+{
+	volatile png_bytep buffer = NULL;
+	volatile png_bytepp row_pointers = NULL;
+	struct icon_theme_icon *volatile result = NULL;
+	FILE *fp = fopen (path, "rb");
+	if (!fp)
+	{
+		if (errno != ENOENT)
+			print_debug ("%s: %s", path, strerror (errno));
+		return NULL;
+	}
+
+	// The simplified and high-level APIs aren't powerful enough.
+	png_structp pngp = png_create_read_struct (PNG_LIBPNG_VER_STRING,
+		(png_voidp) path, icon_theme_open_on_error, icon_theme_open_on_warning);
+	png_infop infop = png_create_info_struct (pngp);
+	if (!infop)
+	{
+		print_debug ("%s: %s", path, strerror (errno));
+		goto fail;
+	}
+	if (setjmp (png_jmpbuf (pngp)))
+		goto fail;
+
+	png_init_io (pngp, fp);
+	png_read_info (pngp, infop);
+
+	// Asking for at least 8-bit channels.  This call is a superset of:
+	//  - png_set_palette_to_rgb(),
+	//  - png_set_tRNS_to_alpha(),
+	//  - png_set_expand_gray_1_2_4_to_8().
+	png_set_expand (pngp);
+
+	// Reduce the possibilities further to RGB or RGBA...
+	png_set_gray_to_rgb (pngp);
+
+	// ...and /exactly/ 8-bit channels.
+	// Alternatively, use png_set_expand_16() above to obtain 16-bit channels.
+	png_set_scale_16 (pngp);
+
+	// PNG uses RGBA order, let's change that to ARGB (both in memory order).
+	// This doesn't change a row's `color_type` in png_do_read_filler(),
+	// and the following transformation thus ignores it.
+	png_set_add_alpha (pngp, 0xFFFF, PNG_FILLER_BEFORE);
+	png_set_swap_alpha (pngp);
+
+	(void) png_set_interlace_handling (pngp);
+
+	png_read_update_info (pngp, infop);
+	if (png_get_bit_depth (pngp, infop) != 8
+	 || png_get_channels (pngp, infop) != 4
+	 || png_get_color_type (pngp, infop) != PNG_COLOR_TYPE_RGB_ALPHA)
+		png_error (pngp, "result not A8R8G8B8");
+
+	size_t row_bytes = png_get_rowbytes (pngp, infop);
+	size_t height = png_get_image_height (pngp, infop);
+	buffer = xcalloc (row_bytes, height);
+	row_pointers = xcalloc (height, sizeof buffer);
+	for (size_t y = 0; y < height; y++)
+		row_pointers[y] = buffer + y * row_bytes;
+
+	png_read_image (pngp, row_pointers);
+
+	result = xcalloc (1, sizeof *result + row_bytes * height);
+	result->width = png_get_image_width (pngp, infop);
+	result->height = height;
+
+	uint32_t *dst = (uint32_t *) result->argb, *src = (uint32_t *) buffer;
+	for (size_t pixels = result->width * result->height; pixels--; )
+		*dst++ = ntohl (*src++);
+
+fail:
+	free (buffer);
+	free (row_pointers);
+	png_destroy_read_struct (&pngp, &infop, NULL);
+	fclose (fp);
+	return result;
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+struct icon_theme_find_context
+{
+	struct strv base;                   ///< Base directories
+	struct str_map visited;             ///< Cycle prevention
+
+	ARRAY (struct icon_theme_icon *, icons)
+};
+
+static void
+icon_theme_find__fallback (struct icon_theme_find_context *ctx,
+	const char *name)
+{
+	for (size_t i = 0; i < ctx->base.len; i++)
+	{
+		char *path = xstrdup_printf ("%s/%s.png", ctx->base.vector[i], name);
+		struct icon_theme_icon *icon = icon_theme_open (path);
+		free (path);
+		if (icon)
+		{
+			ARRAY_RESERVE (ctx->icons, 1);
+			ctx->icons[ctx->icons_len++] = icon;
+			return;
+		}
+	}
+}
+
+static struct desktop_file
+icon_theme_find__index (struct icon_theme_find_context *ctx, const char *theme)
+{
+	struct str data = str_make ();
+	for (size_t i = 0; i < ctx->base.len; i++)
+	{
+		struct error *e = NULL;
+		char *path = xstrdup_printf ("%s/%s/index.theme",
+			ctx->base.vector[i], theme);
+		read_file (path, &data, &e);
+		free (path);
+		if (!e)
+			break;
+
+		if (errno != ENOENT)
+			print_debug ("%s", e->message);
+		error_free (e);
+	}
+
+	struct desktop_file index = desktop_file_make (data.str, data.len);
+	str_free (&data);
+	return index;
+}
+
+static void
+icon_theme_find__named (struct icon_theme_find_context *ctx,
+	const char *theme, const char *name)
+{
+	// Either a cycle, or a common ancestor of inherited themes, which is valid.
+	if (str_map_find (&ctx->visited, theme))
+		return;
+
+	str_map_set (&ctx->visited, theme, (void *) (intptr_t) 1);
+	struct desktop_file index = icon_theme_find__index (ctx, theme);
+
+	char *directories =
+		desktop_file_get_string (&index, "Icon Theme", "Directories");
+	if (!directories)
+		goto out;
+
+	// NOTE: The sizes are not deduplicated, and priorities are uncertain.
+	struct strv dirs = strv_make ();
+	cstr_split (directories, ",", true, &dirs);
+	free (directories);
+	for (size_t d = 0; d < dirs.len; d++)
+	{
+		// The hicolor icon theme stuffs everything in Directories.
+		if (desktop_file_get (&index, dirs.vector[d], "Scale")
+		 && desktop_file_get_integer (&index, dirs.vector[d], "Scale") != 1)
+			continue;
+
+		for (size_t i = 0; i < ctx->base.len; i++)
+		{
+			char *path = xstrdup_printf ("%s/%s/%s/%s.png",
+				ctx->base.vector[i], theme, dirs.vector[d], name);
+			struct icon_theme_icon *icon = icon_theme_open (path);
+			free (path);
+			if (icon)
+			{
+				ARRAY_RESERVE (ctx->icons, 1);
+				ctx->icons[ctx->icons_len++] = icon;
+				break;
+			}
+		}
+	}
+	strv_free (&dirs);
+	if (ctx->icons_len)
+		goto out;
+
+	char *inherits =
+		desktop_file_get_string (&index, "Icon Theme", "Inherits");
+	if (inherits)
+	{
+		struct strv parents = strv_make ();
+		cstr_split (inherits, ",", true, &parents);
+		free (inherits);
+		for (size_t i = 0; i < parents.len; i++)
+		{
+			icon_theme_find__named (ctx, parents.vector[i], name);
+			if (ctx->icons_len)
+				break;
+		}
+		strv_free (&parents);
+	}
+
+out:
+	desktop_file_free (&index);
+}
+
+/// Return all base directories appropriate for icon search.
+static struct strv
+icon_theme_get_base_directories (void)
+{
+	struct strv dirs = strv_make ();
+	struct str icons = str_make ();
+	(void) str_append_env_path (&icons, "HOME", false);
+	str_append (&icons, "/.icons");
+	strv_append_owned (&dirs, str_steal (&icons));
+
+	// Note that we use XDG_CONFIG_HOME as well, which might be intended.
+	struct strv xdg = strv_make ();
+	get_xdg_data_dirs (&xdg);
+	for (size_t i = 0; i < xdg.len; i++)
+		strv_append_owned (&dirs, xstrdup_printf ("%s/icons", xdg.vector[i]));
+	strv_free (&xdg);
+
+	strv_append (&dirs, "/usr/share/pixmaps");
+	return dirs;
+}
+
+static int
+icon_theme_find__compare (const void *a, const void *b)
+{
+	const struct icon_theme_icon **ia = (const struct icon_theme_icon **) a;
+	const struct icon_theme_icon **ib = (const struct icon_theme_icon **) b;
+	double pa = (double) (*ia)->width * (*ia)->height;
+	double pb = (double) (*ib)->width * (*ib)->height;
+	return (pa > pb) - (pa < pb);
+}
+
+/// Return all sizes of the named icon.  When the theme name is not NULL,
+/// use it as the preferred theme.  Always consult fallbacks locations.
+/// Ignore icon scales other than 1.
+static struct icon_theme_icon **
+icon_theme_find (const char *theme, const char *name, size_t *len)
+{
+	struct icon_theme_find_context ctx = {};
+	ctx.base = icon_theme_get_base_directories ();
+	ctx.visited = str_map_make (NULL);
+	ARRAY_INIT (ctx.icons);
+
+	if (theme)
+		icon_theme_find__named (&ctx, theme, name);
+	if (!ctx.icons_len)
+		icon_theme_find__named (&ctx, "hicolor", name);
+	if (!ctx.icons_len)
+		icon_theme_find__fallback (&ctx, name);
+
+	strv_free (&ctx.base);
+	str_map_free (&ctx.visited);
+
+	ARRAY_RESERVE (ctx.icons, 1);
+	ctx.icons[ctx.icons_len] = NULL;
+	if (!ctx.icons_len)
+	{
+		free (ctx.icons);
+		return NULL;
+	}
+
+	qsort (ctx.icons,
+		ctx.icons_len, sizeof *ctx.icons, icon_theme_find__compare);
+	*len = ctx.icons_len;
+	return ctx.icons;
+}
+
+static void
+icon_theme_free (struct icon_theme_icon **icons)
+{
+	for (struct icon_theme_icon **p = icons; *p; p++)
+		free (*p);
+	free (icons);
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+#ifdef LIBERTY_XDG_WANT_X11
+
+static void
+icon_theme_set_window_icon (Display *dpy,
+	Window window, const char *theme, const char *name)
+{
+	size_t icons_len = 0;
+	struct icon_theme_icon **icons = icon_theme_find (theme, name, &icons_len);
+	if (!icons)
+		return;
+
+	size_t n = 0;
+	for (size_t i = 0; i < icons_len; i++)
+		n += 2 + icons[i]->width * icons[i]->height;
+
+	unsigned long *data = xcalloc (n, sizeof *data), *p = data;
+	for (size_t i = 0; i < icons_len; i++)
+	{
+		*p++ = icons[i]->width;
+		*p++ = icons[i]->height;
+
+		uint32_t *q = icons[i]->argb;
+		for (size_t k = icons[i]->width * icons[i]->height; k--; )
+			*p++ = *q++;
+	}
+
+	XChangeProperty (dpy, window, XInternAtom (dpy, "_NET_WM_ICON", False),
+		XA_CARDINAL, 32, PropModeReplace, (const unsigned char *) data, n);
+	free (data);
+	icon_theme_free (icons);
+}
+
+#endif // LIBERTY_XDG_WANT_X11
+#endif // LIBERTY_XDG_WANT_ICONS

liberty.c

@@ -1,7 +1,7 @@
 /*
  * liberty.c: the ultimate C unlibrary
  *
- * Copyright (c) 2014 - 2022, Přemysl Eric Janouch <p@janouch.name>
+ * Copyright (c) 2014 - 2024, 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.
@@ -652,6 +652,60 @@ str_pack_u64 (struct str *self, uint64_t x)
 #define str_pack_i32(self, x)  str_pack_u32 ((self), (uint32_t) (x))
 #define str_pack_i64(self, x)  str_pack_u64 ((self), (uint64_t) (x))
 
+// --- Reading binary numbers --------------------------------------------------
+
+// Doing this byte by byte prevents unaligned memory access issues.
+
+static uint64_t
+peek_u64be (const uint8_t *p)
+{
+	return (uint64_t) p[0] << 56 | (uint64_t) p[1] << 48
+		| (uint64_t) p[2] << 40 | (uint64_t) p[3] << 32
+		| (uint64_t) p[4] << 24 | (uint64_t) p[5] << 16 | p[6] << 8 | p[7];
+}
+
+static uint32_t
+peek_u32be (const uint8_t *p)
+{
+	return (uint32_t) p[0] << 24 | (uint32_t) p[1] << 16 | p[2] << 8 | p[3];
+}
+
+static uint16_t
+peek_u16be (const uint8_t *p)
+{
+	return (uint16_t) p[0] << 8 | p[1];
+}
+
+static uint64_t
+peek_u64le (const uint8_t *p)
+{
+	return (uint64_t) p[7] << 56 | (uint64_t) p[6] << 48
+		| (uint64_t) p[5] << 40 | (uint64_t) p[4] << 32
+		| (uint64_t) p[3] << 24 | (uint64_t) p[2] << 16 | p[1] << 8 | p[0];
+}
+
+static uint32_t
+peek_u32le (const uint8_t *p)
+{
+	return (uint32_t) p[3] << 24 | (uint32_t) p[2] << 16 | p[1] << 8 | p[0];
+}
+
+static uint16_t
+peek_u16le (const uint8_t *p)
+{
+	return (uint16_t) p[1] << 8 | p[0];
+}
+
+struct peeker
+{
+	uint64_t (*u64) (const uint8_t *);
+	uint32_t (*u32) (const uint8_t *);
+	uint16_t (*u16) (const uint8_t *);
+};
+
+static const struct peeker peeker_be = {peek_u64be, peek_u32be, peek_u16be};
+static const struct peeker peeker_le = {peek_u64le, peek_u32le, peek_u16le};
+
 // --- Errors ------------------------------------------------------------------
 
 // Error reporting utilities.  Inspired by GError, only much simpler.
@@ -2503,12 +2557,7 @@ msg_reader_get (struct msg_reader *self, size_t *len)
 		return NULL;
 
 	uint8_t *x = (uint8_t *) self->buf.str + self->offset;
-	uint64_t msg_len
+	uint64_t msg_len = peek_u64be (x);
-		= (uint64_t) x[0] << 56 | (uint64_t) x[1] << 48
-		| (uint64_t) x[2] << 40 | (uint64_t) x[3] << 32
-		| (uint64_t) x[4] << 24 | (uint64_t) x[5] << 16
-		| (uint64_t) x[6] << 8  | (uint64_t) x[7];
-
 	if (msg_len < sizeof msg_len)
 	{
 		// The message is shorter than its header
@@ -2573,8 +2622,7 @@ static bool
 msg_unpacker_u16 (struct msg_unpacker *self, uint16_t *value)
 {
 	UNPACKER_INT_BEGIN
-	*value
+	*value = peek_u16be (x);
-		= (uint16_t) x[0] << 8  | (uint16_t) x[1];
 	return true;
 }
 
@@ -2582,9 +2630,7 @@ static bool
 msg_unpacker_u32 (struct msg_unpacker *self, uint32_t *value)
 {
 	UNPACKER_INT_BEGIN
-	*value
+	*value = peek_u32be (x);
-		= (uint32_t) x[0] << 24 | (uint32_t) x[1] << 16
-		| (uint32_t) x[2] << 8  | (uint32_t) x[3];
 	return true;
 }
 
@@ -2592,11 +2638,7 @@ static bool
 msg_unpacker_u64 (struct msg_unpacker *self, uint64_t *value)
 {
 	UNPACKER_INT_BEGIN
-	*value
+	*value = peek_u64be (x);
-		= (uint64_t) x[0] << 56 | (uint64_t) x[1] << 48
-		| (uint64_t) x[2] << 40 | (uint64_t) x[3] << 32
-		| (uint64_t) x[4] << 24 | (uint64_t) x[5] << 16
-		| (uint64_t) x[6] << 8  | (uint64_t) x[7];
 	return true;
 }
 
@@ -3552,6 +3594,8 @@ write_file_safe (const char *filename, const void *data, size_t data_len,
 {
 	// XXX: ideally we would also open the directory, use *at() versions
 	//   of functions and call fsync() on the directory as appropriate
+	// FIXME: this should behave similarly to mkstemp(), just with 0666;
+	//   as it is, this function is not particularly safe
 	char *temp = xstrdup_printf ("%s.new", filename);
 	bool success = write_file (temp, data, data_len, e);
 	if (success && !(success = !rename (temp, filename)))

tests/fuzz.c

@@ -32,13 +32,6 @@
 #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 -------------------------------------------------------------------
 
@@ -211,19 +204,6 @@ test_config_item_parse (const uint8_t *data, size_t size)
 		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
@@ -266,7 +246,6 @@ LLVMFuzzerInitialize (int *argcp, char ***argvp)
 	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;

tests/lxdrgen.c

@@ -53,6 +53,10 @@ test_ser_deser_free (void)
 			u->others.bar = str_make ();
 			for (int i = rand () % 0x30; i > 0; i--)
 				str_append_c (&u->others.bar, 0x30 + i);
+			u->others.baz_len = rand () % 0x30;
+			u->others.baz = xcalloc (1, u->others.baz_len);
+			for (uint32_t i = 0; i < u->others.baz_len; i++)
+				u->others.baz[i] = 0x30 + i;
 			break;
 		case 2:
 			u->tag = PROTO_GEN_ENUM_NOTHING;
@@ -62,6 +66,8 @@ test_ser_deser_free (void)
 		}
 	}
 
+	a.o.tag = PROTO_GEN_ENUM_NOTHING;
+
 	struct str buf = str_make ();
 	hard_assert (proto_gen_struct_serialize (&a, &buf));
 	struct msg_unpacker r = msg_unpacker_make (buf.str, buf.len);
@@ -92,6 +98,9 @@ test_ser_deser_free (void)
 			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));
+			hard_assert (ua->others.baz_len == ub->others.baz_len);
+			hard_assert (!memcmp (ua->others.baz, ub->others.baz,
+				ua->others.baz_len));
 			break;
 		case PROTO_GEN_ENUM_NOTHING:
 			break;
@@ -100,6 +109,8 @@ test_ser_deser_free (void)
 		}
 	}
 
+	hard_assert (a.o.tag == b.o.tag);
+
 	// Emulate partially deserialized data to test disposal of that.
 	for (size_t i = b.u_len - CASES; i < b.u_len; i++)
 	{

tests/lxdrgen.cpp

@@ -0,0 +1,132 @@
+/*
+ * tests/lxdrgen.cpp
+ *
+ * Copyright (c) 2023, 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 "lxdrgen.lxdr.cpp"
+
+#include <cstdlib>
+
+static void
+hard_assert (bool condition, const char *description)
+{
+	if (!condition)
+	{
+		fprintf (stderr, "assertion failed: %s\n", description);
+		abort ();
+	}
+}
+
+#define hard_assert(condition) hard_assert (condition, #condition)
+
+int
+main (int argc, char *argv[])
+{
+	hard_assert (ProtoGen::VERSION == 1);
+
+	enum { CASES = 3 };
+
+	ProtoGen::Struct a = {}, b = {};
+	a.u.resize (CASES + rand () % 100);
+	for (size_t i = 0; i < a.u.size (); i++)
+	{
+		std::unique_ptr<ProtoGen::Union> &u = a.u[i];
+		switch (i % CASES)
+		{
+		case 0:
+		{
+			auto numbers = new ProtoGen::Union_Numbers ();
+			numbers->a = rand () % UINT8_MAX;
+			numbers->b = rand () % UINT16_MAX;
+			numbers->c = rand () % UINT32_MAX;
+			numbers->d = rand () % UINT64_MAX;
+			numbers->e = rand () % UINT8_MAX;
+			numbers->f = rand () % UINT16_MAX;
+			numbers->g = rand () % UINT32_MAX;
+			numbers->h = rand () % UINT64_MAX;
+			u.reset (numbers);
+			break;
+		}
+		case 1:
+		{
+			auto others = new ProtoGen::Union_Others ();
+			others->foo = rand () % 2;
+			for (int i = rand () % 0x30; i > 0; i--)
+				others->bar += 0x30 + i;
+			for (int i = rand () % 0x30; i > 0; i--)
+				others->baz.push_back (0x30 + i);
+			u.reset (others);
+			break;
+		}
+		case 2:
+			u.reset (new ProtoGen::Union_Nothing ());
+			break;
+		default:
+			hard_assert (!"unhandled case");
+		}
+	}
+
+	a.o.reset (new ProtoGen::Onion_Nothing ());
+
+	LibertyXDR::Writer buf;
+	hard_assert (a.serialize (buf));
+	LibertyXDR::Reader r;
+	r.data = buf.data.data ();
+	r.length = buf.data.size ();
+	hard_assert (b.deserialize (r));
+	hard_assert (!r.length);
+
+	hard_assert (a.u.size () == b.u.size ());
+	for (size_t i = 0; i < a.u.size (); i++)
+	{
+		ProtoGen::Union *ua = a.u[i].get ();
+		ProtoGen::Union *ub = b.u[i].get ();
+		hard_assert (ua->tag == ub->tag);
+		switch (ua->tag)
+		{
+		case ProtoGen::Enum::NUMBERS:
+		{
+			auto a = dynamic_cast<ProtoGen::Union_Numbers *> (ua);
+			auto b = dynamic_cast<ProtoGen::Union_Numbers *> (ub);
+			hard_assert (a->a == b->a);
+			hard_assert (a->b == b->b);
+			hard_assert (a->c == b->c);
+			hard_assert (a->d == b->d);
+			hard_assert (a->e == b->e);
+			hard_assert (a->f == b->f);
+			hard_assert (a->g == b->g);
+			hard_assert (a->h == b->h);
+			break;
+		}
+		case ProtoGen::Enum::OTHERS:
+		{
+			auto a = dynamic_cast<ProtoGen::Union_Others *> (ua);
+			auto b = dynamic_cast<ProtoGen::Union_Others *> (ub);
+			hard_assert (a->foo == b->foo);
+			hard_assert (a->bar == b->bar);
+			hard_assert (a->baz == b->baz);
+			break;
+		}
+		case ProtoGen::Enum::NOTHING:
+			break;
+		default:
+			hard_assert (!"unexpected case");
+		}
+	}
+
+	hard_assert (a.o->tag == b.o->tag);
+	return 0;
+}

tests/lxdrgen.lxdr

@@ -17,7 +17,13 @@ struct Struct {
 	case OTHERS:
 		bool foo;
 		string bar;
+		u8 baz<>;
 	case NOTHING:
 		void;
 	} u<>;
+
+	union Onion switch (Enum tag) {
+	case NOTHING:
+		void;
+	} o;
 };

tests/xdg.c

@@ -0,0 +1,67 @@
+/*
+ * tests/xdg.c
+ *
+ * Copyright (c) 2024, 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 "../liberty-xdg.c"
+
+static const char file[] =
+	"# This only tests the happy paths\n"
+	"[Desktop Entry]\n"
+	"Version = 1.0\n"
+	"Name=\\s\\n\\t\\r\\\\\n"
+	"Name[fr]=Nom\n"
+	"Hidden=true\n"
+	"Categories=Utility;TextEditor;\n"
+	"Number=42";
+
+static void
+test_desktop_file (void)
+{
+	struct desktop_file entry = desktop_file_make (file, sizeof file - 1);
+	const char *group = "Desktop Entry";
+
+	char *value = desktop_file_get_string (&entry, group, "Version");
+	hard_assert (!strcmp (value, "1.0"));
+	cstr_set (&value, desktop_file_get_string (&entry, group, "Name"));
+	hard_assert (!strcmp (value, " \n\t\r\\"));
+	free (value);
+
+	hard_assert (desktop_file_get_bool (&entry, group, "Hidden"));
+	struct strv values = desktop_file_get_stringv (&entry, group, "Categories");
+	hard_assert (values.len == 2);
+	hard_assert (!strcmp (values.vector[0], "Utility"));
+	hard_assert (!strcmp (values.vector[1], "TextEditor"));
+	strv_free (&values);
+	hard_assert (desktop_file_get_integer (&entry, group, "Number") == 42);
+
+	desktop_file_free (&entry);
+}
+
+int
+main (int argc, char *argv[])
+{
+	struct test test;
+	test_init (&test, argc, argv);
+
+	test_add_simple (&test, "/desktop-file", NULL, test_desktop_file);
+
+	return test_run (&test);
+}

tools/asciiman.awk

@@ -1,6 +1,6 @@
 # asciiman.awk: simplified AsciiDoc to manual page converter
 #
-# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# Copyright (c) 2022 - 2023, Přemysl Eric Janouch <p@janouch.name>
 # SPDX-License-Identifier: 0BSD
 #
 # This is not intended to produce great output, merely useful output.
@@ -20,21 +20,17 @@ function fatal(message) {
 	exit 1
 }
 
-function haveattribute(name) {
+BEGIN {
-	return name in Attrs || ("asciidoc-" name) in ENVIRON
+	for (name in ENVIRON)
+		if (match(name, /^asciidoc-/))
+			Attrs[substr(name, RSTART + RLENGTH)] = ENVIRON[name]
 }
 
-function getattribute(name) {
+function expand(s,   attrname, v) {
-	if (!(name in Attrs) && ("asciidoc-" name) in ENVIRON)
+	while (match(s, /[{][^{}]+[}]/)) {
-		Attrs[name] = ENVIRON["asciidoc-" name]
+		attrname = substr(s, RSTART + 1, RLENGTH - 2)
-	return Attrs[name]
+		if (attrname in Attrs)
-}
+			v = v substr(s, 1, RSTART - 1) Attrs[attrname]
-
-function expand(s,   attr, v) {
-	while (match(s, /[{][^{}]*[}]/)) {
-		attr = substr(s, RSTART + 1, RLENGTH - 2)
-		if (haveattribute(attr))
-			v = v substr(s, 1, RSTART - 1) getattribute(attr)
 		else
 			v = v substr(s, 1, RSTART + RLENGTH - 1)
 		s = substr(s, RSTART + RLENGTH)
@@ -49,13 +45,20 @@ function escape(s) {
 	return s
 }
 
-function readattribute(line,    attrname, attrvalue) {
+function readattribute(line,    attrname) {
-	if (match(line, /^:[^:]*: /)) {
+	if (match(line, /^:[^:]+:$/)) {
+		Attrs[substr(line, RSTART + 1, RLENGTH - 2)] = ""
+	} else if (match(line, /^:[^:]+!:$/)) {
+		delete Attrs[substr(line, RSTART + 1, RLENGTH - 3)]
+	} else if (match(line, /^:![^:]+:$/)) {
+		delete Attrs[substr(line, RSTART + 2, RLENGTH - 3)]
+	} else if (match(line, /^:[^:]+: /)) {
 		attrname = substr(line, RSTART + 1, RLENGTH - 3)
-		attrvalue = substr(line, RSTART + RLENGTH)
+		Attrs[attrname] = expand(substr(line, RSTART + RLENGTH))
-		Attrs[attrname] = expand(attrvalue)
+	} else {
-		return 1
+		return 0
 	}
+	return 1
 }
 
 NR == 1 {
@@ -80,15 +83,56 @@ NR == 1 {
 	# Requesting tbl(1), even though we currently do not support tables.
 	print "'\\\" t"
 	printf ".TH \"%s\" \"%s\" \"\" \"%s\"",
-		toupper(name), section, getattribute("mansource")
+		toupper(name), section, Attrs["mansource"]
-	if (getattribute("manmanual"))
+	if ("manmanual" in Attrs)
-		printf " \"%s\"", getattribute("manmanual")
+		printf " \"%s\"", Attrs["manmanual"]
 	print ""
 
 	# Hyphenation is indeed rather annoying, in particular with long links.
 	print ".nh"
 }
 
+function readattrlist(line, posattrs, namedattrs,    name, value, n) {
+	if (!match(line, /^\[.*\]$/))
+		return 0
+
+	line = expand(substr(line, RSTART + 1, RLENGTH - 2))
+	while (line) {
+		name = ""
+		if (match(line, /^[[:alnum:]][[:alnum:]-]*/)) {
+			value = substr(line, RSTART, RLENGTH)
+			if (match(substr(line, RSTART + RLENGTH),
+					/^[[:space:]]*=[[:space:]]*/)) {
+				name = value
+				line = substr(line, 1 + length(name) + RLENGTH)
+			}
+		}
+
+		# The quoting syntax actually is awful like this.
+		if (match(line, /^"(\\.|[^"\\])*"/)) {
+			value = substr(line, RSTART + 1, RLENGTH - 2)
+			gsub(/\\"/, "\"", value)
+		} else if (match(line, /^'(\\.|[^'\\])*'/)) {
+			value = substr(line, RSTART + 1, RLENGTH - 2)
+			gsub(/\\'/, "'", value)
+		} else {
+			match(line, /^[^,]*/)
+			value = substr(line, RSTART, RLENGTH)
+			sub(/[[:space:]]*$/, "", value)
+		}
+
+		line = substr(line, RSTART + RLENGTH)
+		sub(/^[[:space:]]*,[[:space:]]*/, "", line)
+		if (!name)
+			posattrs[++n] = value
+		else if (value == "None")
+			delete namedattrs[name]
+		else
+			namedattrs[name] = value
+	}
+	return 1
+}
+
 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.
@@ -151,7 +195,7 @@ function inline(line) {
 }
 
 # Returns 1 iff the left-over $0 should be processed further.
-function process(firstline) {
+function process(firstline,     posattrs, namedattrs) {
 	if (readattribute(firstline))
 		return 0
 	if (getline <= 0) {
@@ -159,6 +203,17 @@ function process(firstline) {
 		return 0
 	}
 
+	# Block attribute list lines.
+	delete posattrs[0]
+	delete namedattrs[0]
+	while (readattrlist(firstline, posattrs, namedattrs)) {
+		firstline = $0
+		if (getline <= 0) {
+			inline(firstline)
+			return 0
+		}
+	}
+
 	# mandoc(1) automatically precedes section headers with blank lines.
 	if (length(firstline) == length($0) && /^-+$/) {
 		print ".SH \"" escape(toupper(expand(firstline))) "\""
@@ -196,7 +251,7 @@ function process(firstline) {
 		return 1
 	}
 
-	# We generally assume these block end with a blank line.
+	# We generally assume these blocks end with a blank line.
 	if (match(firstline, /^[[:space:]]*[*][[:space:]]+/)) {
 		flushspace()
 

tools/lxdrgen-c.awk

@@ -222,7 +222,7 @@ function codegen_struct(name, cg,    ctype, funcname) {
 		delete cg[i]
 }
 
-function codegen_union_tag(d, cg) {
+function codegen_union_tag(name, d, cg) {
 	cg["tagtype"] = d["type"]
 	cg["tagname"] = d["name"]
 	append(cg, "fields", "\t" CodegenCType[d["type"]] " " d["name"] ";\n")
@@ -259,7 +259,7 @@ function codegen_union_struct( \
 		"\t\tbreak;\n")
 }
 
-function codegen_union(name, cg,    f, ctype, funcname) {
+function codegen_union(name, cg, exhaustive,    f, ctype, funcname) {
 	ctype = "union " PrefixLower cameltosnake(name)
 	print ""
 	print ctype " {"
@@ -281,7 +281,7 @@ function codegen_union(name, cg,    f, ctype, funcname) {
 
 		CodegenDispose[name] = "\t" funcname "(&%s);\n"
 	}
-	if (cg["serialize"]) {
+	{
 		funcname = PrefixLower cameltosnake(name) "_serialize"
 		print ""
 		print "static bool\n" \
@@ -299,7 +299,7 @@ function codegen_union(name, cg,    f, ctype, funcname) {
 		CodegenSerialize[name] = "\tif (!" funcname "(&%s, w))\n" \
 			"\t\treturn false;\n"
 	}
-	if (cg["deserialize"]) {
+	{
 		funcname = PrefixLower cameltosnake(name) "_deserialize"
 		print ""
 		print "static bool\n" \

tools/lxdrgen-cpp-posix.cpp

@@ -0,0 +1,67 @@
+// lxdrgen-cpp-posix.cpp: POSIX support code for lxdrgen-cpp.awk.
+//
+// Copyright (c) 2023, Přemysl Eric Janouch <p@janouch.name>
+// SPDX-License-Identifier: 0BSD
+#include <iconv.h>
+
+#include <cstdint>
+#include <string>
+
+// Various BSD derivatives may have a problem here.
+// Linux defines __STDC_ISO_10646__, but also supports "WCHAR_T".
+#ifdef APPLE
+#define ICONV_WCHAR "UTF-32"
+#else
+#define ICONV_WCHAR "WCHAR_T"
+#endif
+
+namespace LibertyXDR {
+
+bool utf8_to_wstring(const uint8_t *utf8, size_t length, std::wstring &wide) {
+	iconv_t conv = iconv_open(ICONV_WCHAR, "UTF-8");
+	if (conv == (iconv_t) -1)
+		return false;
+
+	wchar_t buffer[1024] = {};
+	char *start = (char *) buffer, *out = start, *in = (char *) utf8;
+	size_t available = sizeof buffer;
+	wide.clear();
+	while (iconv(conv, &in, &length, &out, &available) == (size_t) -1) {
+		if (errno != E2BIG) {
+			iconv_close(conv);
+			return false;
+		}
+
+		wide.append(buffer, (out - start) / sizeof *buffer);
+		out = start;
+		available = sizeof buffer;
+	}
+	wide.append(buffer, (out - start) / sizeof *buffer);
+	iconv_close(conv);
+	return true;
+}
+
+bool wstring_to_utf8(const std::wstring &wide, std::string &utf8) {
+	iconv_t conv = iconv_open("UTF-8", ICONV_WCHAR);
+	if (conv == (iconv_t) -1)
+		return false;
+
+	char buffer[1024] = {}, *out = buffer, *in = (char *) wide.data();
+	size_t available = sizeof buffer, length = wide.size() * sizeof wide[0];
+	utf8.clear();
+	while (iconv(conv, &in, &length, &out, &available) == (size_t) -1) {
+		if (errno != E2BIG) {
+			iconv_close(conv);
+			return false;
+		}
+
+		utf8.append(buffer, out - buffer);
+		out = buffer;
+		available = sizeof buffer;
+	}
+	utf8.append(buffer, out - buffer);
+	iconv_close(conv);
+	return true;
+}
+
+} // namespace LibertyXDR

tools/lxdrgen-cpp-win32.cpp

@@ -0,0 +1,47 @@
+// lxdrgen-cpp-win32.cpp: Win32 support code for lxdrgen-cpp.awk.
+//
+// Copyright (c) 2023, Přemysl Eric Janouch <p@janouch.name>
+// SPDX-License-Identifier: 0BSD
+#include <windows.h>
+
+#include <climits>
+#include <cstdint>
+#include <string>
+
+namespace LibertyXDR {
+
+bool utf8_to_wstring(const uint8_t *utf8, size_t length, std::wstring &wide) {
+	wide.clear();
+	if (!length)
+		return true;
+	if (length > INT_MAX)
+		return false;
+
+	int size = MultiByteToWideChar(CP_UTF8, MB_ERR_INVALID_CHARS,
+		(LPCCH) utf8, length, nullptr, 0);
+	if (size <= 0)
+		return false;
+
+	wide.resize(size);
+	return !!MultiByteToWideChar(CP_UTF8, MB_ERR_INVALID_CHARS,
+		(LPCCH) utf8, length, wide.data(), size);
+}
+
+bool wstring_to_utf8(const std::wstring &wide, std::string &utf8) {
+	utf8.clear();
+	if (wide.empty())
+		return true;
+	if (wide.size() > INT_MAX)
+		return false;
+
+	int size = WideCharToMultiByte(CP_UTF8, WC_ERR_INVALID_CHARS,
+		(LPCWCH) wide.data(), wide.size(), nullptr, 0, NULL, NULL);
+	if (size <= 0)
+		return false;
+
+	utf8.resize(size);
+	return !!WideCharToMultiByte(CP_UTF8, WC_ERR_INVALID_CHARS,
+		(LPCWCH) wide.data(), wide.size(), utf8.data(), size, NULL, NULL);
+}
+
+} // namespace LibertyXDR

tools/lxdrgen-cpp.awk

@@ -0,0 +1,350 @@
+# lxdrgen-cpp.awk: C++ backend for lxdrgen.awk.
+#
+# This backend is intended for Windows, it just happens to have a fallback
+# that will probably work on Unices, of which we make use in tests.
+#
+# Copyright (c) 2023, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+
+function define_internal(name, ctype) {
+	Types[name] = "internal"
+	CodegenCType[name] = ctype
+	CodegenSerialize[name] = \
+		"\tw.append(%s);\n"
+	CodegenDeserialize[name] = \
+		"\tif (!r.read(%s))\n" \
+		"\t\treturn false;\n"
+}
+
+function define_int(shortname, ctype) {
+	define_internal(shortname, ctype)
+}
+
+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", "std::wstring")
+	define_internal("bool", "bool")
+
+	CodegenSerialize["string"] = \
+		"\tif (!w.append(%s))\n" \
+		"\t\treturn false;\n"
+
+	print "// Code generated from " FILENAME ". DO NOT EDIT."
+	print ""
+	print "#include <cstdint>"
+	print "#include <memory>"
+	print "#include <string>"
+	print "#include <vector>"
+	print ""
+	print "namespace LibertyXDR {"
+	print ""
+	print "bool utf8_to_wstring("
+	print "\tconst uint8_t *utf8, size_t length, std::wstring &wide);"
+	print "bool wstring_to_utf8("
+	print "\tconst std::wstring &wide, std::string &utf8);"
+	print ""
+	print "struct Reader {"
+	print "\tconst uint8_t *data = {};"
+	print "\tsize_t length = {};"
+	print ""
+	print "\ttemplate<typename T> bool read(T &number) {"
+	print "\t\tif (length < sizeof number)"
+	print "\t\t\treturn false;"
+	print ""
+	print "\t\tnumber = 0;"
+	print "\t\tfor (size_t i = 0; i < sizeof number; i++) {"
+	print "\t\t\tnumber = number << 8 | *data++;"
+	print "\t\t\tlength--;"
+	print "\t\t}"
+	print "\t\treturn true;"
+	print "\t}"
+	print ""
+	print "\tbool read(bool &boolean) {"
+	print "\t\tuint8_t number = 0;"
+	print "\t\tif (!read(number))"
+	print "\t\t\treturn false;"
+	print ""
+	print "\t\tboolean = number != 0;"
+	print "\t\treturn true;"
+	print "\t}"
+	print ""
+	print "\tbool read(std::wstring &string) {"
+	print "\t\tuint32_t size = 0;"
+	print "\t\tif (!read(size) || size > length)"
+	print "\t\t\treturn false;"
+	print "\t\tif (!utf8_to_wstring(data, size, string))"
+	print "\t\t\treturn false;"
+	print ""
+	print "\t\tdata += size;"
+	print "\t\tlength -= size;"
+	print "\t\treturn true;"
+	print "\t}"
+	print ""
+	print "\tbool read(std::vector<uint8_t> &vector) {"
+	print "\t\tuint32_t size = 0;"
+	print "\t\tif (!read(size) || size > length)"
+	print "\t\t\treturn false;"
+	print "\t\tvector.assign(data, data + size);"
+	print ""
+	print "\t\tdata += size;"
+	print "\t\tlength -= size;"
+	print "\t\treturn true;"
+	print "\t}"
+	print "};"
+	print ""
+	print "struct Writer {"
+	print "\tstd::vector<uint8_t> data;"
+	print ""
+	print "\ttemplate<typename T> bool append(T number) {"
+	print "\t\tuint8_t buffer[sizeof number], *p = buffer + sizeof buffer;"
+	print "\t\twhile (p != buffer) {"
+	print "\t\t\t*--p = number;"
+	print "\t\t\tnumber >>= 8;"
+	print "\t\t}"
+	print "\t\tdata.insert(data.end(), buffer, buffer + sizeof buffer);"
+	print "\t\treturn true;"
+	print "\t}"
+	print ""
+	print "\tbool append(int8_t number) {"
+	print "\t\tdata.push_back(number);"
+	print "\t\treturn true;"
+	print "\t}"
+	print ""
+	print "\tbool append(uint8_t number) {"
+	print "\t\tdata.push_back(number);"
+	print "\t\treturn true;"
+	print "\t}"
+	print ""
+	print "\tbool append(bool boolean) {"
+	print "\t\treturn append(uint8_t(boolean));"
+	print "\t}"
+	print ""
+	print "\tbool append(const std::wstring &string) {"
+	print "\t\tif (string.size() > UINT32_MAX)"
+	print "\t\t\treturn false;"
+	print ""
+	print "\t\tstd::string utf8;"
+	print "\t\tif (!wstring_to_utf8(string, utf8))"
+	print "\t\t\treturn false;"
+	print ""
+	print "\t\tappend<uint32_t>(utf8.size());"
+	print "\t\tdata.insert(data.end(), utf8.begin(), utf8.end());"
+	print "\t\treturn true;"
+	print "\t}"
+	print "};"
+	print ""
+	print "} // namespace LibertyXDR"
+	print "namespace " PrefixCamel " {"
+}
+
+END {
+	print ""
+	print "} // namespace " PrefixCamel
+}
+
+function codegen_constant(name, value) {
+	print ""
+	print "enum { " name " = " value " };"
+}
+
+function codegen_enum_value(name, subname, value, cg) {
+	append(cg, "fields", "\t" subname " = " value ",\n")
+}
+
+function codegen_enum(name, cg) {
+	print ""
+	print "enum struct " name " : int8_t {"
+	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] = \
+		"\tw.append(static_cast<int8_t>(%s));\n"
+	CodegenDeserialize[name] = \
+		"\t{\n" \
+		"\t\tint8_t v = 0;\n" \
+		"\t\tif (!r.read(v) || !v)\n" \
+		"\t\t\treturn false;\n" \
+		"\t\t%s = static_cast<" name ">(v);\n" \
+		"\t}\n"
+
+	CodegenCType[name] = name
+	for (i in cg)
+		delete cg[i]
+}
+
+# Some identifiers do not pose a problem in C, but do in our C++.
+function codegen_struct_sanitize(name) {
+	if (name ~ /^(serialize|deserialize)_*$/ ||
+		name ~ /^(catch|class|delete|except|finally|friend|new|operator)_*$/ ||
+		name ~ /^(private|protected|public|template|this|throw|try|virtual)_*$/)
+		return name "_"
+	return name
+}
+
+function codegen_struct_tag(d, cg,    name, f) {
+	name = codegen_struct_sanitize(d["name"])
+	f = "this->" name
+
+	append(cg, "serialize", sprintf(CodegenSerialize[d["type"]], f))
+	# Do not deserialize here, that would be out of order.
+}
+
+function codegen_struct_field(d, cg,    name, f, serialize, deserialize) {
+	name = codegen_struct_sanitize(d["name"])
+	f = "this->" name
+
+	serialize = CodegenSerialize[d["type"]]
+	deserialize = CodegenDeserialize[d["type"]]
+	if (!d["isarray"]) {
+		append(cg, "fields",
+			"\t" CodegenCType[d["type"]] " " name " = {};\n")
+		append(cg, "serialize", sprintf(serialize, f))
+		append(cg, "deserialize", sprintf(deserialize, f))
+		return
+	}
+
+	append(cg, "fields",
+		"\tstd::vector<" CodegenCType[d["type"]] "> " name ";\n")
+
+	# XXX: We should probably pedantically check for overflows.
+	append(cg, "serialize",
+		sprintf(CodegenSerialize["u32"], "uint32_t(" f ".size())") \
+		"\tfor (const auto &it : " f ")\n" \
+		indent(sprintf(serialize, "it")))
+
+	if (d["type"] == "u8") {
+		append(cg, "deserialize",
+			"\tif (!r.read(" f "))\n" \
+			"\t\treturn false;\n")
+	} else if (deserialize) {
+		append(cg, "deserialize",
+			"\t{\n" \
+			"\t\tuint32_t size = 0;\n" \
+			indent(sprintf(CodegenDeserialize["u32"], "size")) \
+			"\t\t" f ".resize(size);\n" \
+			"\t}\n" \
+			"\tfor (auto &it : " f ")\n" \
+			indent(sprintf(deserialize, "it")))
+	}
+}
+
+function codegen_struct(name, cg) {
+	print ""
+	print "struct " name " {"
+	print cg["fields"]
+	print "\tbool serialize(LibertyXDR::Writer &w) const {"
+	print indent(cg["serialize"]) "\t\treturn true;"
+	print "\t}"
+	print ""
+	print "\tbool deserialize([[maybe_unused]] LibertyXDR::Reader &r) {"
+	print indent(cg["deserialize"]) "\t\treturn true;"
+	print "\t}"
+	print "};"
+
+	CodegenSerialize[name] = "\tif (!%s->serialize(w))\n" \
+		"\t\treturn false;\n"
+	CodegenDeserialize[name] = "\tif (!%s->deserialize(r))\n" \
+		"\t\treturn false;\n"
+
+	CodegenCType[name] = name
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_union_tag(name, d, cg,    tagname) {
+	cg["tagtype"] = d["type"]
+	cg["tagname"] = tagname = codegen_struct_sanitize(d["name"])
+
+	print ""
+	print "struct " name " {"
+	print "\t" CodegenCType[d["type"]] " " tagname " = {};"
+	print "\tvirtual ~" name "() = 0;"
+	print "\tvirtual bool serialize(LibertyXDR::Writer &w) const = 0;"
+	print "\tvirtual bool deserialize(LibertyXDR::Reader &r) = 0;"
+	print "};"
+	print ""
+	print name "::~" name "() {}"
+}
+
+function codegen_union_struct(name, casename, cg, scg,     structname) {
+	# And thus not all generated structs are present in Types.
+	structname = name "_" snaketocamel(casename)
+
+	print ""
+	print "struct " structname " : virtual public " name " {"
+	print scg["fields"]
+	print "\t" structname "() {"
+	print "\t\tthis->" cg["tagname"] " = " \
+		CodegenCType[cg["tagtype"]] "::" casename ";"
+	print "\t}"
+	print ""
+	print "\tvirtual bool serialize(LibertyXDR::Writer &w) const {"
+	print indent(scg["serialize"]) "\t\treturn true;"
+	print "\t}"
+	print ""
+	print "\tvirtual bool deserialize([[maybe_unused]] LibertyXDR::Reader &r) {"
+	print indent(scg["deserialize"]) "\t\treturn true;"
+	print "\t}"
+	print "};"
+
+	append(cg, "deserialize",
+		"\tcase " CodegenCType[cg["tagtype"]] "::" casename ":\n" \
+		"\t\treturn new " structname "();\n")
+
+	CodegenSerialize[structname] = "\tif (!%s->serialize(w))\n" \
+		"\t\treturn false;\n"
+	CodegenDeserialize[structname] = "\tif (!%s->deserialize(r))\n" \
+		"\t\treturn false;\n"
+
+	CodegenCType[structname] = structname
+	for (i in scg)
+		delete scg[i]
+}
+
+function codegen_union(name, cg, exhaustive,    ctype) {
+	CodegenSerialize[name] = "\tif (!%s->serialize(w))\n" \
+		"\t\treturn false;\n"
+
+	ctype = "std::unique_ptr<" name ">"
+	if (cg["deserialize"]) {
+		print ""
+		print "static " name " *read" name "(" \
+			CodegenCType[cg["tagtype"]] " " cg["tagname"] ") {"
+		print "\tswitch (" cg["tagname"] ") {"
+		print cg["deserialize"] "\tdefault:"
+		print "\t\treturn nullptr;"
+		print "\t}"
+		print "}"
+		print ""
+		print "static " ctype " read" name "(LibertyXDR::Reader &r) {"
+		print "\tint8_t v = 0;"
+		print "\tif (!r.read(v) || !v)"
+		print "\t\treturn nullptr;"
+		print ""
+		print "\t" ctype " result(read" name "(static_cast<" \
+			CodegenCType[cg["tagtype"]] ">(v)));"
+		print "\tif (!result || !result->deserialize(r))"
+		print "\t\treturn nullptr;"
+		print "\treturn result;"
+		print "}"
+
+		CodegenDeserialize[name] = "\tif (!(%s = read" name "(r)))\n" \
+			"\t\treturn false;\n"
+	}
+
+	CodegenCType[name] = ctype
+	for (i in cg)
+		delete cg[i]
+}

tools/lxdrgen-go.awk

@@ -439,9 +439,9 @@ function codegen_struct(name, cg,    gotype) {
 		delete cg[i]
 }
 
-function codegen_union_tag(d, cg) {
+function codegen_union_tag(name, d, cg) {
 	cg["tagtype"] = d["type"]
-	cg["tagname"] = d["name"]
+	cg["tagname"] = snaketocamel(d["name"])
 	# The tag is implied from the type of struct stored in the interface.
 }
 
@@ -450,8 +450,8 @@ function codegen_union_struct(name, casename, cg, scg,     structname, init) {
 	structname = name snaketocamel(casename)
 	codegen_struct(structname, scg)
 
-	init = CodegenGoType[structname] "{" snaketocamel(cg["tagname"]) \
+	init = CodegenGoType[structname] "{" cg["tagname"] \
-		": " decapitalize(snaketocamel(cg["tagname"])) "}"
+		": " decapitalize(cg["tagname"]) "}"
 	append(cg, "unmarshal",
 		"\tcase " CodegenGoType[cg["tagtype"]] snaketocamel(casename) ":\n" \
 		"\t\ts := " init "\n" \
@@ -467,7 +467,7 @@ function codegen_union_struct(name, casename, cg, scg,     structname, init) {
 		"\t\tu.Interface = &s\n")
 }
 
-function codegen_union(name, cg,    gotype, tagfield, tagvar) {
+function codegen_union(name, cg, exhaustive,    gotype, tagvar) {
 	gotype = PrefixCamel name
 	print "type " gotype " struct {"
 	print "\tInterface any"
@@ -481,18 +481,17 @@ function codegen_union(name, cg,    gotype, tagfield, tagvar) {
 	print "}"
 	print ""
 
-	tagfield = snaketocamel(cg["tagname"])
+	tagvar = decapitalize(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"]] \
+	print "\t\t" cg["tagname"] " " 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 "\tswitch " tagvar " := t." cg["tagname"] "; " tagvar " {"
 	print cg["unmarshal"] "\tdefault:"
 	print "\t\terr = errors.New(`unsupported value: ` + " tagvar ".String())"
 	print "\t}"

tools/lxdrgen-mjs.awk

@@ -183,16 +183,16 @@ function codegen_struct(name, cg) {
 		delete cg[i]
 }
 
-function codegen_union_tag(d, cg) {
+function codegen_union_tag(name, d, cg) {
 	cg["tagtype"] = d["type"]
-	cg["tagname"] = d["name"]
+	cg["tagname"] = decapitalize(snaketocamel(d["name"]))
 }
 
 function codegen_union_struct(name, casename, cg, scg,     structname) {
 	append(scg, "methods",
 		"\n" \
 		"\tconstructor() {\n" \
-		"\t\tthis." decapitalize(snaketocamel(cg["tagname"])) \
+		"\t\tthis." cg["tagname"] \
 			" = " cg["tagtype"] "." snaketocamel(casename) "\n" \
 		"\t}\n")
 
@@ -208,8 +208,8 @@ function codegen_union_struct(name, casename, cg, scg,     structname) {
 		"\t}\n")
 }
 
-function codegen_union(name, cg,    tagvar) {
+function codegen_union(name, cg, exhaustive,    tagvar) {
-	tagvar = decapitalize(snaketocamel(cg["tagname"]))
+	tagvar = cg["tagname"]
 
 	print ""
 	print "export function deserialize" name "(r) {"

tools/lxdrgen-swift.awk

@@ -0,0 +1,277 @@
+# lxdrgen-swift.awk: Swift backend for lxdrgen.awk.
+#
+# Copyright (c) 2023, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+
+function define_internal(name, swifttype) {
+	Types[name] = "internal"
+	CodegenSwiftType[name] = swifttype
+	CodegenDeserialize[name] = "%s.read()"
+}
+
+function define_sint(size,    shortname, swifttype) {
+	shortname = "i" size
+	swifttype = "Int" size
+	define_internal(shortname, swifttype)
+}
+
+function define_uint(size,    shortname, swifttype) {
+	shortname = "u" size
+	swifttype = "UInt" size
+	define_internal(shortname, swifttype)
+}
+
+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("bool", "Bool")
+	define_internal("string", "String")
+
+	print "// Code generated from " FILENAME ". DO NOT EDIT."
+	print "import Foundation"
+	print ""
+	print "public struct " PrefixCamel "Reader {"
+	print "\tpublic var data: Data"
+	print ""
+	print "\tpublic enum ReadError: Error {"
+	print "\t\tcase unexpectedEOF"
+	print "\t\tcase invalidEncoding"
+	print "\t\tcase overflow"
+	print "\t\tcase unexpectedValue"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func read<T: FixedWidthInteger>() throws -> T {"
+	print "\t\tlet size = MemoryLayout<T>.size"
+	print "\t\tguard data.count >= size else {"
+	print "\t\t\tthrow ReadError.unexpectedEOF"
+	print "\t\t}"
+	print "\t\tvar acc: T = 0"
+	print "\t\tdata.prefix(size).forEach { acc = acc << 8 | T($0) }"
+	print "\t\tdata = data.dropFirst(size)"
+	print "\t\treturn acc"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func read() throws -> Bool {"
+	print "\t\ttry read() != UInt8(0)"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func read() throws -> String {"
+	print "\t\tlet size: UInt32 = try self.read()"
+	print "\t\tguard let count = Int(exactly: size) else {"
+	print "\t\t\tthrow ReadError.overflow"
+	print "\t\t}"
+	print "\t\tguard data.count >= count else {"
+	print "\t\t\tthrow ReadError.unexpectedEOF"
+	print "\t\t}"
+	print "\t\tdefer {"
+	print "\t\t\tdata = data.dropFirst(count)"
+	print "\t\t}"
+	print "\t\tif let s = String(data: data.prefix(count), encoding: .utf8) {"
+	print "\t\t\treturn s"
+	print "\t\t} else {"
+	print "\t\t\tthrow ReadError.invalidEncoding"
+	print "\t\t}"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func read<" \
+		"T: RawRepresentable<Int8>>() throws -> T {"
+	print "\t\tguard let value = T(rawValue: try read()) else {"
+	print "\t\t\tthrow ReadError.unexpectedValue"
+	print "\t\t}"
+	print "\t\treturn value"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func read<T>("
+	print "\t\t\t_ read: (inout Self) throws -> T) throws -> [T] {"
+	print "\t\tlet size: UInt32 = try self.read()"
+	print "\t\tguard let count = Int(exactly: size) else {"
+	print "\t\t\tthrow ReadError.overflow"
+	print "\t\t}"
+	print "\t\tvar array = [T]()"
+	print "\t\tarray.reserveCapacity(count)"
+	print "\t\tfor _ in 0..<count {"
+	print "\t\t\tarray.append(try read(&self))"
+	print "\t\t}"
+	print "\t\treturn array"
+	print "\t}"
+	print "}"
+	print ""
+	print "public struct " PrefixCamel "Writer {"
+	print "\tpublic var data = Data()"
+	print ""
+	print "\tpublic mutating func append<T: FixedWidthInteger>(_ number: T) {"
+	print "\t\tvar n = number.byteSwapped"
+	print "\t\tfor _ in 0..<MemoryLayout<T>.size {"
+	print "\t\t\tdata.append(UInt8(truncatingIfNeeded: n))"
+	print "\t\t\tn >>= 8"
+	print "\t\t}"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func append(_ bool: Bool) {"
+	print "\t\tappend(UInt8(bool ? 1 : 0))"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func append(_ string: String) {"
+	print "\t\tlet bytes = string.data(using: .utf8)!"
+	print "\t\tappend(UInt32(bytes.count))"
+	print "\t\tdata.append(bytes)"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func append<T: " \
+		"RawRepresentable<Int8>>(_ value: T) {"
+	print "\t\tappend(value.rawValue)"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func append<T>("
+	print "\t\t\t_ array: Array<T>, _ write: (inout Self, T) -> ()) {"
+	print "\t\tappend(UInt32(array.count))"
+	print "\t\tfor i in 0..<array.count {"
+	print "\t\t\twrite(&self, array[i])"
+	print "\t\t}"
+	print "\t}"
+	print ""
+	print "\tpublic mutating func append<T: " \
+		PrefixCamel "Encodable>(_ value: T) {"
+	print "\t\tvalue.encode(to: &self)"
+	print "\t}"
+	print "}"
+	print ""
+	print "public protocol " PrefixCamel "Encodable { " \
+		"func encode(to: inout " PrefixCamel "Writer) }"
+}
+
+function codegen_constant(name, value) {
+	print ""
+	print "public let " decapitalize(PrefixCamel snaketocamel(name)) " = " value
+}
+
+function codegen_enum_value(name, subname, value, cg) {
+	append(cg, "fields",
+		"\tcase " decapitalize(snaketocamel(subname)) " = " value "\n")
+}
+
+function codegen_enum(name, cg,    swifttype) {
+	swifttype = PrefixCamel name
+	print ""
+	print "public enum " swifttype ": Int8 {"
+	print cg["fields"] "}"
+
+	CodegenSwiftType[name] = swifttype
+	CodegenDeserialize[name] = "%s.read()"
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_struct_field(d, cg,    camel) {
+	camel = decapitalize(snaketocamel(d["name"]))
+	if (!d["isarray"]) {
+		append(cg, "fields",
+			"\tpublic var " camel ": " CodegenSwiftType[d["type"]] "\n")
+		append(cg, "deserialize",
+			"\t\tself." camel " = try " \
+				sprintf(CodegenDeserialize[d["type"]], "from") "\n")
+		append(cg, "serialize",
+			"\t\tto.append(self." camel ")\n")
+		return
+	}
+
+	append(cg, "fields",
+		"\tpublic var " camel ": [" CodegenSwiftType[d["type"]] "]\n")
+	append(cg, "deserialize",
+		"\t\tself." camel " = try from.read() { r in try " \
+			sprintf(CodegenDeserialize[d["type"]], "r") " }\n")
+	append(cg, "serialize",
+		"\t\tto.append(self." camel ") { (w, value) in w.append(value) }\n")
+}
+
+function codegen_struct_tag(d, cg,    camel) {
+	camel = decapitalize(snaketocamel(d["name"]))
+	append(cg, "serialize",
+		"\t\tto.append(self." camel ")\n")
+}
+
+function codegen_struct(name, cg,    swifttype) {
+	swifttype = PrefixCamel name
+	print ""
+	print "public struct " swifttype " {\n" cg["fields"] "}"
+	print ""
+	print "extension " swifttype ": " PrefixCamel "Encodable {"
+	print "\tpublic init(from: inout " PrefixCamel "Reader) throws {"
+	print cg["deserialize"] "\t}"
+	print ""
+	print "\tpublic func encode(to: inout " PrefixCamel "Writer) {"
+	print cg["serialize"] "\t}"
+	print "}"
+
+	CodegenSwiftType[name] = swifttype
+	CodegenDeserialize[name] = "%s.read()"
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_union_tag(name, d, cg) {
+	cg["tagtype"] = d["type"]
+	cg["tagname"] = decapitalize(snaketocamel(d["name"]))
+}
+
+function codegen_union_struct(name, casename, cg, scg,     swifttype) {
+	# And thus not all generated structs are present in Types.
+	swifttype = PrefixCamel name snaketocamel(casename)
+	casename = decapitalize(snaketocamel(casename))
+	print ""
+	print "public struct " swifttype ": " PrefixCamel name " {"
+	print "\tpublic var " cg["tagname"] \
+		": " CodegenSwiftType[cg["tagtype"]] " { ." casename " }"
+	print scg["fields"] "}"
+	print ""
+	print "extension " swifttype ": " PrefixCamel "Encodable {"
+	print "\tfileprivate init(from: inout " PrefixCamel "Reader) throws {"
+	print scg["deserialize"] "\t}"
+	print ""
+	print "\tpublic func encode(to: inout " PrefixCamel "Writer) {"
+	print scg["serialize"] "\t}"
+	print "}"
+
+	append(cg, "cases", "\tcase ." casename ":\n" \
+		"\t\treturn try " swifttype "(from: &from)\n")
+
+	CodegenSwiftType[name] = swifttype
+	CodegenDeserialize[name] = "%s.read()"
+	for (i in scg)
+		delete scg[i]
+}
+
+function codegen_union(name, cg, exhaustive,    swifttype, init) {
+	# Classes don't have automatic member-wise initializers,
+	# thus using structs and protocols.
+	swifttype = PrefixCamel name
+	print ""
+	print "public protocol " swifttype ": " PrefixCamel "Encodable {"
+	print "\tvar " cg["tagname"] ": " CodegenSwiftType[cg["tagtype"]] " { get }"
+	print "}"
+
+	if (!exhaustive)
+		append(cg, "cases", "\tdefault:\n" \
+			"\t\tthrow " PrefixCamel "Reader.ReadError.unexpectedValue\n")
+
+	init = decapitalize(swifttype)
+	print ""
+	print "public func " init \
+		"(from: inout " PrefixCamel "Reader) throws -> " swifttype " {"
+	print "\tlet " cg["tagname"] ": " CodegenSwiftType[cg["tagtype"]] \
+		" = try from.read()"
+	print "\tswitch " cg["tagname"] " {"
+	print cg["cases"] "\t}"
+	print "}"
+
+	CodegenSwiftType[name] = swifttype
+	CodegenDeserialize[name] = init "(from: &%s)"
+	for (i in cg)
+		delete cg[i]
+}

tools/lxdrgen.awk

@@ -1,6 +1,6 @@
 # lxdrgen.awk: an XDR-derived code generator for network protocols.
 #
-# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# Copyright (c) 2022 - 2023, 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 \
@@ -28,9 +28,10 @@ function snaketocamel(s) {
 }
 
 function decapitalize(s) {
-	if (match(s, /[[:upper:]][[:lower:]]/)) {
+	if (match(s, /^[[:upper:]][[:lower:]]/))
 		return tolower(substr(s, 1, 1)) substr(s, 2)
-	}
+	if (match(s, /^[[:upper:]]$/))
+		return tolower(s)
 	return s
 }
 
@@ -216,7 +217,8 @@ function defstruct(    name, d, cg) {
 	return name
 }
 
-function defunion(    name, tag, tagtype, tagvalue, cg, scg, d, a, i, unseen) {
+function defunion(    name, tag, tagtype, tagvalue, cg, scg, d, a, i,
+		unseen, exhaustive) {
 	delete cg[0]
 	delete scg[0]
 	delete d[0]
@@ -230,7 +232,7 @@ function defunion(    name, tag, tagtype, tagvalue, cg, scg, d, a, i, unseen) {
 
 	if (Types[tagtype] != "enum")
 		fatal("not an enum type: " tagtype)
-	codegen_union_tag(tag, cg)
+	codegen_union_tag(name, tag, cg)
 
 	split(EnumValues[tagtype], a, SUBSEP)
 	for (i in a)
@@ -257,9 +259,14 @@ function defunion(    name, tag, tagtype, tagvalue, cg, scg, d, a, i, unseen) {
 	if (tagvalue)
 		codegen_union_struct(name, tagvalue, cg, scg)
 
-	# What remains non-zero in unseen[2..] is simply not recognized/allowed.
+	# Unseen cases are simply not recognized/allowed.
+	exhaustive = 1
+	for (i in unseen)
+		if (i && unseen[i])
+			exhaustive = 0
+
 	Types[name] = "union"
-	codegen_union(name, cg)
+	codegen_union(name, cg, exhaustive)
 	return name
 }