Start X11 and web frontends for xC

For this, we needed a wire protocol.  After surveying available options,
it was decided to implement an XDR-like protocol code generator
in portable AWK.  It now has two backends, per each of:

 - xF, the X11 frontend, is in C, and is meant to be the primary
   user interface in the future.

 - xP, the web frontend, relies on a protocol proxy written in Go,
   and is meant for use on-the-go (no pun intended).

They are very much work-in-progress proofs of concept right now,
and the relay protocol is certain to change.

.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

@@ -3,7 +3,9 @@
 
 # Qt Creator files
 /CMakeLists.txt.user*
-/uirc3.config
-/uirc3.files
-/uirc3.creator*
-/uirc3.includes
+/xK.config
+/xK.files
+/xK.creator*
+/xK.includes
+/xK.cflags
+/xK.cxxflags

CMakeLists.txt

@@ -1,15 +1,21 @@
-cmake_minimum_required (VERSION 3.0)
-project (uirc3 VERSION 1.3.0 LANGUAGES C)
+# Ubuntu 18.04 LTS and OpenBSD 6.4
+cmake_minimum_required (VERSION 3.10)
+project (xK VERSION 1.5.0
+	DESCRIPTION "IRC daemon, bot, TUI client and X11/web frontends" LANGUAGES C)
 
 # Options
 option (WANT_READLINE "Use GNU Readline for the UI (better)" ON)
 option (WANT_LIBEDIT "Use BSD libedit for the UI" OFF)
+option (WANT_XF "Build xF" OFF)
 
 # Moar warnings
+set (CMAKE_C_STANDARD 99)
+set (CMAKE_C_STANDARD_REQUIRED ON)
+set (CMAKE_C_EXTENSIONS OFF)
+
 if ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUCC)
 	# -Wunused-function is pretty annoying here, as everything is static
-	set (wdisabled "-Wno-unused-function")
-	set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -std=c99 -Wall -Wextra ${wdisabled}")
+	set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wextra -Wno-unused-function")
 endif ()
 
 # Version
@@ -57,6 +63,8 @@ if ("${CMAKE_SYSTEM_NAME}" MATCHES "BSD")
 	# Need this for SIGWINCH in FreeBSD and OpenBSD respectively;
 	# our POSIX version macros make it undefined
 	add_definitions (-D__BSD_VISIBLE=1 -D_BSD_SOURCE=1)
+elseif (APPLE)
+	add_definitions (-D_DARWIN_C_SOURCE)
 endif ()
 
 # -lrt is only for glibc < 2.17
@@ -112,10 +120,16 @@ endif ()
 if ((WANT_READLINE AND WANT_LIBEDIT) OR (NOT WANT_READLINE AND NOT WANT_LIBEDIT))
 	message (SEND_ERROR "You have to choose either GNU Readline or libedit")
 elseif (WANT_READLINE)
+	pkg_check_modules (readline readline)
+
 	# OpenBSD's default readline is too old
 	if ("${CMAKE_SYSTEM_NAME}" MATCHES "OpenBSD")
 		include_directories (${OPENBSD_LOCALBASE}/include/ereadline)
 		list (APPEND xC_libraries ereadline)
+	elseif (readline_FOUND)
+		list (APPEND xC_libraries ${readline_LIBRARIES})
+		include_directories (${readline_INCLUDE_DIRS})
+		link_directories (${readline_LIBRARY_DIRS})
 	else ()
 		list (APPEND xC_libraries readline)
 	endif ()
@@ -131,28 +145,55 @@ set (HAVE_EDITLINE "${WANT_LIBEDIT}")
 set (HAVE_LUA      "${WITH_LUA}")
 
 include (GNUInstallDirs)
-configure_file (${PROJECT_SOURCE_DIR}/config.h.in ${PROJECT_BINARY_DIR}/config.h)
+set (project_config ${PROJECT_BINARY_DIR}/config.h)
+configure_file (${PROJECT_SOURCE_DIR}/config.h.in ${project_config})
 include_directories (${PROJECT_SOURCE_DIR} ${PROJECT_BINARY_DIR})
 
 # Generate IRC replies--we need a custom target because of the multiple outputs
 add_custom_command (OUTPUT xD-replies.c xD.msg
-	COMMAND ${PROJECT_SOURCE_DIR}/xD-gen-replies.sh
-		> xD-replies.c < ${PROJECT_SOURCE_DIR}/xD-replies
-	DEPENDS ${PROJECT_SOURCE_DIR}/xD-replies
+	COMMAND env LC_ALL=C awk
+		-f ${PROJECT_SOURCE_DIR}/xD-gen-replies.awk
+		${PROJECT_SOURCE_DIR}/xD-replies > xD-replies.c
+	DEPENDS
+		${PROJECT_SOURCE_DIR}/xD-gen-replies.awk
+		${PROJECT_SOURCE_DIR}/xD-replies
 	COMMENT "Generating files from the list of server numerics")
 add_custom_target (replies DEPENDS ${PROJECT_BINARY_DIR}/xD-replies.c)
 
+add_custom_command (OUTPUT xC-proto.c
+	COMMAND env LC_ALL=C awk
+		-f ${PROJECT_SOURCE_DIR}/xC-gen-proto.awk
+		-f ${PROJECT_SOURCE_DIR}/xC-gen-proto-c.awk
+		${PROJECT_SOURCE_DIR}/xC-proto > xC-proto.c
+	DEPENDS
+		${PROJECT_SOURCE_DIR}/xC-gen-proto.awk
+		${PROJECT_SOURCE_DIR}/xC-gen-proto-c.awk
+		${PROJECT_SOURCE_DIR}/xC-proto
+	COMMENT "Generating xC relay protocol code")
+add_custom_target (xC-proto DEPENDS ${PROJECT_BINARY_DIR}/xC-proto.c)
+
 # Build
 foreach (name xB xC xD)
-	add_executable (${name} ${name}.c ${PROJECT_BINARY_DIR}/config.h)
+	add_executable (${name} ${name}.c ${project_config})
 	target_link_libraries (${name} ${project_libraries})
 	add_threads (${name})
 endforeach ()
 
 add_dependencies (xD replies)
-add_dependencies (xC replies)
+add_dependencies (xC replies xC-proto)
 target_link_libraries (xC ${xC_libraries})
 
+if (WANT_XF)
+	pkg_check_modules (x11 REQUIRED x11 xrender xft fontconfig)
+	include_directories (${x11_INCLUDE_DIRS})
+	link_directories (${x11_LIBRARY_DIRS})
+
+	add_executable (xF xF.c ${project_config})
+	add_dependencies (xF xC-proto)
+	target_link_libraries (xF ${x11_LIBRARIES} ${project_libraries})
+	add_threads (xF)
+endif ()
+
 # Tests
 include (CTest)
 if (BUILD_TESTING)
@@ -184,7 +225,7 @@ add_custom_target (clang-tidy
 # Installation
 install (TARGETS xB xC xD DESTINATION ${CMAKE_INSTALL_BINDIR})
 install (FILES LICENSE DESTINATION ${CMAKE_INSTALL_DOCDIR})
-# XXX: our defaults for XDG_DATA_DIRS expect /usr/local/shore or /usr/share
+# XXX: our defaults for XDG_DATA_DIRS expect /usr/local/share or /usr/share
 install (DIRECTORY plugins/xB/
 	DESTINATION ${CMAKE_INSTALL_DATADIR}/xB/plugins USE_SOURCE_PERMISSIONS)
 install (DIRECTORY plugins/xC/
@@ -192,20 +233,31 @@ install (DIRECTORY plugins/xC/
 
 # Generate documentation from text markup
 find_program (ASCIIDOCTOR_EXECUTABLE asciidoctor)
-if (NOT ASCIIDOCTOR_EXECUTABLE)
-	message (FATAL_ERROR "asciidoctor not found")
+find_program (A2X_EXECUTABLE a2x)
+if (NOT ASCIIDOCTOR_EXECUTABLE AND NOT A2X_EXECUTABLE)
+	message (FATAL_ERROR "Neither asciidoctor nor a2x were found")
 endif ()
 
 foreach (page xB xC xD)
 	set (page_output "${PROJECT_BINARY_DIR}/${page}.1")
 	list (APPEND project_MAN_PAGES "${page_output}")
-	add_custom_command (OUTPUT ${page_output}
-		COMMAND ${ASCIIDOCTOR_EXECUTABLE} -b manpage
-			-a release-version=${project_version}
-			"${PROJECT_SOURCE_DIR}/${page}.adoc"
-			-o "${page_output}"
-		DEPENDS ${page}.adoc
-		COMMENT "Generating man page for ${page}" VERBATIM)
+	if (ASCIIDOCTOR_EXECUTABLE)
+		add_custom_command (OUTPUT ${page_output}
+			COMMAND ${ASCIIDOCTOR_EXECUTABLE} -b manpage
+				-a release-version=${project_version}
+				-o "${page_output}"
+				"${PROJECT_SOURCE_DIR}/${page}.adoc"
+			DEPENDS ${page}.adoc
+			COMMENT "Generating man page for ${page}" VERBATIM)
+	elseif (A2X_EXECUTABLE)
+		add_custom_command (OUTPUT ${page_output}
+			COMMAND ${A2X_EXECUTABLE} --doctype manpage --format manpage
+				-a release-version=${project_version}
+				-D "${PROJECT_BINARY_DIR}"
+				"${PROJECT_SOURCE_DIR}/${page}.adoc"
+			DEPENDS ${page}.adoc
+			COMMENT "Generating man page for ${page}" VERBATIM)
+	endif ()
 endforeach ()
 
 add_custom_target (docs ALL DEPENDS ${project_MAN_PAGES})
@@ -217,7 +269,6 @@ foreach (page ${project_MAN_PAGES})
 endforeach ()
 
 # CPack
-set (CPACK_PACKAGE_DESCRIPTION_SUMMARY "Unreasonable IRC client, daemon and bot")
 set (CPACK_PACKAGE_VERSION "${project_version_safe}")
 set (CPACK_PACKAGE_VENDOR "Premysl Eric Janouch")
 set (CPACK_PACKAGE_CONTACT "Přemysl Eric Janouch <p@janouch.name>")

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2014 - 2021, 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.

NEWS

@@ -1,3 +1,55 @@
+Unreleased
+
+ * xD: implemented WALLOPS, choosing to make it target even non-operators
+
+ * xC: made it show WALLOPS messages, as PRIVMSG for the server buffer
+
+ * xC: all behaviour.* configuration options have been renamed to general.*,
+   with the exception of editor_command/editor, backlog_helper/pager,
+   and backlog_helper_strip_formatting/pager_strip_formatting
+
+ * xC: replaced behaviour.save_on_quit with general.autosave
+
+ * xC: improved pager integration capabilities
+
+ * xC: normalized editline's history behaviour, making it a viable frontend
+
+ * xC: various bugfixes
+
+ * xC: added a relay interface, enabled through the general.relay_bind option
+
+ * Added an experimental X11 frontend for xC called xF
+
+ * Added an experimental web frontend for xC called xP
+
+
+1.5.0 (2021-12-21) "The Show Must Go On"
+
+ * xC: made it possible to pass the cursor position to external editors,
+   in particular VIM and Emacs
+
+ * xC: started quoting text coming from bracketed pastes,
+   to minimize the risk of trying to execute filesystem paths as commands
+
+ * xC: fixed to work with post-2021-08-29 editline
+
+ * xC: extended editline's autocomplete to show all options
+
+ * utm-filter.lua: added Facebook's tracking parameter to the filter
+
+
+1.4.0 (2021-10-06) "Call Me Scruffy Scruffington"
+
+ * xC: made message autosplitting respect text formatting
+
+ * xC: fixed displaying IRC colours above 16
+
+ * xC: offer IRCnet as an IRC network to connect to,
+   rather than the lunatic new Freenode
+
+ * xD: started bumping the soft limit on file descriptors to the hard one
+
+
 1.3.0 (2021-08-07) "New World Order"
 
  * xC: made nick autocompletion offer recent speakers first

README.adoc

@@ -1,73 +1,66 @@
-uirc3
-=====
-:compact-option:
+xK
+==
 
-The unreasonable IRC trinity.  This project consists of an IRC client, daemon,
-and bot.  It's all you're ever going to need for chatting, as long as you can
-make do with minimalist software.
+'xK' (chat kit) is an IRC software suite consisting of a daemon, bot, terminal
+client, and X11/web frontends for the client.  It's all you're ever going to
+need for chatting, so long as you can make do with slightly minimalist software.
 
-All of them have these potentially interesting properties:
-
- - IPv6 support
- - TLS support, including client certificates
- - lean on dependencies (with the exception of 'xC')
- - compact and arguably easy to hack on
- - very permissive license
+They're all lean on dependencies, and offer a maximally permissive licence.
 
 xC
 --
-The IRC client.  It is largely defined by being built on top of GNU Readline
-that has been hacked to death.  Its interface should feel somewhat familiar for
-weechat or irssi users.
+The IRC client, and the core of 'xK'.  It is largely defined by building on top
+of GNU Readline or BSD Editline that have been hacked to death.  Its interface
+should feel somewhat familiar for weechat or irssi users.
 
 image::xC.png[align="center"]
 
-This is the largest application within the project.  It has most of the stuff
-you'd expect of an IRC client, such as being able to set up multiple servers,
-a powerful configuration system, integrated help, text formatting, CTCP queries,
-automatic splitting of overlong messages, autocomplete, logging to file,
-auto-away, command aliases and basic support for Lua scripting.
+It has most features you'd expect of an IRC client, such as being multiserver,
+a powerful configuration system, integrated help, text formatting, automatic
+message splitting, multiline editing, bracketed paste support, word wrapping
+that doesn't break links, autocomplete, logging, CTCP queries, auto-away,
+command aliases, SOCKS proxying, SASL EXTERNAL authentication using TLS client
+certificates, a remote relay interface, or basic support for Lua scripting.
+As a unique bonus, you can launch a full text editor from within.
+
+xF
+--
+The X11 frontend for 'xC', making use of its networked relay interface.
+It's currently in development, and is hidden behind a CMake option.
+
+xP
+--
+The web frontend for 'xC', making use of its networked relay interface.
+It's currently rather elementary, and can be built from within its directory.
 
 xD
 --
-The IRC daemon.  It is designed to be used as a regular user application rather
-than a system-wide daemon.  If all you want is a decent, minimal IRCd for
-testing purposes or a small network of respectful users (or bots), this one will
-do it just fine.
+The IRC daemon.  It is designed for use as a regular user application rather
+than a system-wide daemon, and follows the XDG Base Directory Specification.
+If all you want is a decent, minimal IRCd for testing purposes or a small
+network of respectful users (or bots), this one will do it just fine.
 
-Notable features:
+It autodetects TLS on incoming connections (I'm still wondering why everyone
+doesn't have this), authenticates operators via TLS client certificate
+fingerprints, and supports a number of IRCv3 capabilities.
 
- - TLS autodetection (why doesn't everyone have this?), using secure defaults
- - IRCop authentication via TLS client certificates
- - epoll/kqueue support; this means that it should be able to handle quite
-   a number of concurrent user connections
- - partial IRCv3 support
+What it notably doesn't support is online changes to configuration, any limits
+besides the total number of connections and mode `+l`, or server linking
+(which also means no services).
 
-Not supported:
-
- - server linking (which also means no services); I consider existing protocols
-   for this purpose ugly and tricky to implement correctly; I've also found no
-   use for this feature yet
- - online changes to configuration; the configuration system from 'xC' could
-   be used to implement this feature if needed
- - limits of almost any kind, just connections and mode `+l`
-
-This program has been
-https://git.janouch.name/p/haven/src/branch/master/hid[ported to Go],
-and development continues over there.
+This program has been https://git.janouch.name/p/haven/src/branch/master/hid[
+ported to Go] in a different project, and development continues over there.
 
 xB
 --
-The IRC bot.  It builds upon the concept of my other VitaminA IRC bot.  The main
-characteristic of these two bots is that they run plugins as coprocesses, which
-allows for enhanced reliability and programming language freedom.
+The IRC bot.  While originally intended to be a simple rewrite of my old GNU AWK
+bot in C, it fairly quickly became a playground, and it eventually got me into
+writing the rest of this package.
 
-While originally intended to be a simple rewrite of the original AWK bot in C,
-it fairly quickly became a playground, and it eventually got me into writing
-the rest of the package.
-
-It survives crashes, server disconnects and timeouts, and also has native SOCKS
-support (even though socksify can add that easily to any program).
+Its main characteristic is that it runs plugins as coprocesses, allowing for
+enhanced reliability and programming language freedom.  Moreover, it recovers
+from any crashes, and offers native SOCKS support (even though socksify can add
+that easily to any program).
 
 Packages
 --------
@@ -76,18 +69,17 @@ a package with the latest development version from Archlinux's AUR.
 
 Building
 --------
-Build dependencies: CMake, pkg-config, asciidoctor, awk, liberty (included) +
-Runtime dependencies: openssl +
-Additionally for 'xC': curses, libffi, lua >= 5.3 (optional),
-                       readline >= 6.0 or libedit >= 2013-07-12
+Build-only dependencies:
+ CMake, pkg-config, asciidoctor or asciidoc, awk, liberty (included) +
+Common runtime dependencies: openssl +
+Additionally for 'xC': curses, libffi, +
+ readline >= 6.0 or libedit >= 2013-07-12, lua >= 5.3 (optional) +
+Additionally for 'xF': x11, xft
 
-Avoid libedit if you can, in general it works but at the moment history is
-acting up and I have no clue about fixing it.
-
- $ git clone --recursive https://git.janouch.name/p/uirc3.git
- $ mkdir uirc3/build
- $ cd uirc3/build
- $ cmake .. -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=Debug \
+ $ git clone --recursive https://git.janouch.name/p/xK.git
+ $ mkdir xK/build
+ $ cd xK/build
+ $ cmake .. -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=RelWithDebInfo \
             -DWANT_READLINE=ON -DWANT_LIBEDIT=OFF -DWANT_LUA=ON
  $ make
 
@@ -95,10 +87,10 @@ To install the application, you can do either the usual:
 
  # make install
 
-Or you can try telling CMake to make a package for you.  For Debian it is:
+Or you can try telling CMake to make a package for you:
 
- $ cpack -G DEB
- # dpkg -i uirc3-*.deb
+ $ cpack -G DEB  # also supported: RPM, FreeBSD
+ # dpkg -i xK-*.deb
 
 Usage
 -----
@@ -156,14 +148,18 @@ Beware that you can easily break the program if you're not careful.
 
 How do I make xC look like the screenshot?
 ------------------------------------------
-First of all, you must build it with Lua support.  With the defaults, 'xC'
-doesn't look too fancy because I don't want to depend on Lua or 256-colour
-terminals.  In addition to that, I appear to be one of the few people who use
-black on white terminals.
+With the defaults, 'xC' doesn't look too fancy because I don't want to have
+a hard dependency on either Lua for the bundled script that provides an easily
+adjustable enhanced prompt, or on 256-colour terminals.  Moreover, it's nearly
+impossible to come up with a colour theme that would work well with both
+black-on-white and white-on-black terminals, or anything wild in between.
 
- /set behaviour.date_change_line = "%a %e %b %Y"
- /set behaviour.plugin_autoload += "fancy-prompt.lua"
- /set behaviour.backlog_helper = "LESSSECURE=1 less -R +Gb1d -Ps'Backlog ?ltlines %lt-%lb?L/%L. .?e(END):?pB%pB\\%..'"
+Assuming that your build supports Lua plugins, and that you have a decent,
+properly set-up terminal emulator, it suffices to run:
+
+ /set general.pager = Press Tab here and change +Gb to +Gb1d
+ /set general.date_change_line = "%a %e %b %Y"
+ /set general.plugin_autoload += "fancy-prompt.lua"
  /set attributes.userhost = "\x1b[38;5;109m"
  /set attributes.join = "\x1b[38;5;108m"
  /set attributes.part = "\x1b[38;5;138m"
@@ -184,7 +180,7 @@ configurations accordingly, but I consider it rather messy and unnecessary.
 
 Contributing and Support
 ------------------------
-Use https://git.janouch.name/p/uirc3 to report any bugs, request features,
+Use https://git.janouch.name/p/xK to report any bugs, request features,
 or submit pull requests.  `git send-email` is tolerated.  If you want to discuss
 the project, feel free to join me at ircs://irc.janouch.name, channel #dev.
 

common.c

@@ -1,7 +1,7 @@
 /*
  * common.c: common functionality
  *
- * Copyright (c) 2014 - 2020, 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.
@@ -22,11 +22,11 @@
 #define LIBERTY_WANT_PROTO_IRC
 
 #ifdef WANT_SYSLOG_LOGGING
-	#define print_fatal_data    ((void *) LOG_ERR)
-	#define print_error_data    ((void *) LOG_ERR)
-	#define print_warning_data  ((void *) LOG_WARNING)
-	#define print_status_data   ((void *) LOG_INFO)
-	#define print_debug_data    ((void *) LOG_DEBUG)
+#define print_fatal_data    ((void *) LOG_ERR)
+#define print_error_data    ((void *) LOG_ERR)
+#define print_warning_data  ((void *) LOG_WARNING)
+#define print_status_data   ((void *) LOG_INFO)
+#define print_debug_data    ((void *) LOG_DEBUG)
 #endif // WANT_SYSLOG_LOGGING
 
 #include "liberty/liberty.c"
@@ -48,6 +48,66 @@ init_openssl (void)
 #endif
 }
 
+static char *
+gai_reconstruct_address (struct addrinfo *ai)
+{
+	char host[NI_MAXHOST] = {}, port[NI_MAXSERV] = {};
+	int err = getnameinfo (ai->ai_addr, ai->ai_addrlen,
+		host, sizeof host, port, sizeof port,
+		NI_NUMERICHOST | NI_NUMERICSERV);
+	if (err)
+	{
+		print_debug ("%s: %s", "getnameinfo", gai_strerror (err));
+		return xstrdup ("?");
+	}
+	return format_host_port_pair (host, port);
+}
+
+static bool
+accept_error_is_transient (int err)
+{
+	// OS kernels may return a wide range of unforeseeable errors.
+	// Assuming that they're either transient or caused by
+	// a connection that we've just extracted from the queue.
+	switch (err)
+	{
+	case EBADF:
+	case EINVAL:
+	case ENOTSOCK:
+	case EOPNOTSUPP:
+		return false;
+	default:
+		return true;
+	}
+}
+
+/// Destructively tokenize an address into a host part, and a port part.
+/// The port is only overwritten if that part is found, allowing for defaults.
+static const char *
+tokenize_host_port (char *address, const char **port)
+{
+	// Unwrap IPv6 addresses in format_host_port_pair() format.
+	char *rbracket = strchr (address, ']');
+	if (*address == '[' && rbracket)
+	{
+		if (rbracket[1] == ':')
+		{
+			*port = rbracket + 2;
+			return *rbracket = 0, address + 1;
+		}
+		if (!rbracket[1])
+			return *rbracket = 0, address + 1;
+	}
+
+	char *colon = strchr (address, ':');
+	if (colon)
+	{
+		*port = colon + 1;
+		return *colon = 0, address;
+	}
+	return address;
+}
+
 // --- To be moved to liberty --------------------------------------------------
 
 // FIXME: in xssl_get_error() we rely on error reasons never being NULL (i.e.,
@@ -74,6 +134,15 @@ xerr_describe_error (void)
 	return reason;
 }
 
+static struct str
+str_from_cstr (const char *cstr)
+{
+	struct str self;
+	self.alloc = (self.len = strlen (cstr)) + 1;
+	self.str = memcpy (xmalloc (self.alloc), cstr, self.alloc);
+	return self;
+}
+
 static ssize_t
 strv_find (const struct strv *v, const char *s)
 {
@@ -87,15 +156,15 @@ static time_t
 unixtime_msec (long *msec)
 {
 #ifdef _POSIX_TIMERS
-        struct timespec tp;
-        hard_assert (clock_gettime (CLOCK_REALTIME, &tp) != -1);
-        *msec = tp.tv_nsec / 1000000;
+	struct timespec tp;
+	hard_assert (clock_gettime (CLOCK_REALTIME, &tp) != -1);
+	*msec = tp.tv_nsec / 1000000;
 #else // ! _POSIX_TIMERS
-        struct timeval tp;
-        hard_assert (gettimeofday (&tp, NULL) != -1);
-        *msec = tp.tv_usec / 1000;
+	struct timeval tp;
+	hard_assert (gettimeofday (&tp, NULL) != -1);
+	*msec = tp.tv_usec / 1000;
 #endif // ! _POSIX_TIMERS
-        return tp.tv_sec;
+	return tp.tv_sec;
 }
 
 // --- Logging -----------------------------------------------------------------

plugins/xC/fancy-prompt.lua

@@ -1,7 +1,7 @@
 --
 -- fancy-prompt.lua: the fancy multiline prompt you probably want
 --
--- Copyright (c) 2016, Přemysl Eric Janouch <p@janouch.name>
+-- Copyright (c) 2016 - 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.
@@ -40,7 +40,7 @@ xC.hook_prompt (function (hook)
 		if buffer == current then
 			current_n = i
 		elseif buffer.new_messages_count ~= buffer.new_unimportant_count then
-			if active ~= "" then active = active .. "," end
+			active = active .. ","
 			if buffer.highlighted then
 				active = active .. "!"
 				bg_color = "224"
@@ -48,7 +48,6 @@ xC.hook_prompt (function (hook)
 			active = active .. i
 		end
 	end
-	if active ~= "" then active = "(" .. active .. ")" end
 	local x = current_n .. ":" .. current.name
 	if chan and chan.users_len ~= 0 then
 		local params = ""
@@ -56,25 +55,34 @@ xC.hook_prompt (function (hook)
 			params = params .. " +" .. mode .. " " .. param
 		end
 		local modes = chan.no_param_modes .. params:sub (3)
-		if modes ~= "" then x = x .. "(+" .. modes .. ")" end
+		if modes ~= "" then
+			x = x .. "(+" .. modes .. ")"
+		end
 		x = x .. "{" .. chan.users_len .. "}"
 	end
-	if current.hide_unimportant then x = x .. "<H>" end
-
-	local lines, cols = xC.get_screen_size ()
-	x = x .. " " .. active .. string.rep (" ", cols)
+	if current.hide_unimportant then
+		x = x .. "<H>"
+	end
+	if active ~= "" then
+		x = x .. " (" .. active:sub (2) .. ")"
+	end
 
 	-- Readline 7.0.003 seems to be broken and completely corrupts the prompt.
 	-- However 8.0.004 seems to be fine with these, as is libedit 20191231-3.1.
 	--x = x:gsub("[\128-\255]", "?")
 
-	-- Cut off extra characters and apply formatting, including the hack.
-	-- FIXME: this doesn't count with full-width or zero-width characters.
-	--   We might want to export wcwidth() above term_from_utf8 somehow.
-	local overflow = utf8.offset (x, cols - 1)
-	if overflow then x = x:sub (1, overflow) end
+	-- Align to the terminal's width and apply formatting, including the hack.
+	local lines, cols = xC.get_screen_size ()
+	local trailing, width = " ", xC.measure (x)
+	while cols > 0 and width >= cols do
+		x = x:sub (1, utf8.offset (x, -1) - 1)
+		trailing, width = ">", xC.measure (x)
+	end
+
 	x = "\x01\x1b[0;4;1;38;5;16m\x1b[48;5;" .. bg_color .. "m\x02" ..
-		x ..  "\x01\x1b[0;4;1;7;38;5;" .. bg_color .. "m\x02 \x01\x1b[0;1m\x02"
+		x .. string.rep (" ", cols - width - 1) ..
+		"\x01\x1b[0;4;1;7;38;5;" .. bg_color .. "m\x02" ..
+		trailing .. "\x01\x1b[0;1m\x02"
 
 	local user_prefix = function (chan, user)
 		for i, chan_user in ipairs (chan.users) do

plugins/xC/utm-filter.lua

@@ -1,7 +1,7 @@
 --
--- utm-filter.lua: filter out Google Analytics bullshit from URLs
+-- utm-filter.lua: filter out Google Analytics bullshit etc. from URLs
 --
--- Copyright (c) 2015, Přemysl Eric Janouch <p@janouch.name>
+-- Copyright (c) 2015 - 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.
@@ -19,6 +19,10 @@
 local banned = {
 	gclid        = 1,
 
+	-- Alas, Facebook no longer uses this parameter, see:
+	-- https://news.ycombinator.com/item?id=32117489
+	fbclid       = 1,
+
 	utm_source   = 1,
 	utm_medium   = 1,
 	utm_term     = 1,

xB.adoc

@@ -1,8 +1,8 @@
 xB(1)
 =====
 :doctype: manpage
-:manmanual: uirc3 Manual
-:mansource: uirc3 {release-version}
+:manmanual: xK Manual
+:mansource: xK {release-version}
 
 Name
 ----
@@ -60,14 +60,14 @@ using the IRC protocol.  (Caveat: the standard C library doesn't automatically
 flush FILE streams for pipes on newlines.)  A special *XB* command is introduced
 for RPC, with the following subcommands:
 
-*XB get_config* _key_::
+*XB get_config* __key__::
 	Request the value of the given configuration option.  If no such option
 	exists, the value will be empty.  The response will be delivered in
 	the following format:
 +
-```
+....
 XB :value
-```
+....
 +
 This is particularly useful for retrieving the *prefix* string.
 
@@ -100,5 +100,5 @@ _/usr/share/xB/plugins/_::
 
 Reporting bugs
 --------------
-Use https://git.janouch.name/p/uirc3 to report bugs, request features,
+Use https://git.janouch.name/p/xK to report bugs, request features,
 or submit pull requests.

xC-gen-proto-c.awk

@@ -0,0 +1,325 @@
+# xC-gen-proto-c.awk: C backend for xC-gen-proto.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 "// 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_i32(w, %s);\n"
+	CodegenDeserialize[name] = \
+		"\t{\n" \
+		"\t\tint32_t v = 0;\n" \
+		"\t\tif (!msg_unpacker_i32(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\t" 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\t" 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]
+}

xC-gen-proto-go.awk

@@ -0,0 +1,447 @@
+# xC-gen-proto-go.awk: Go backend for xC-gen-proto.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)
+
+	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) {
+	shortname = "u" size
+	gotype = "uint" size
+	define_internal(shortname, gotype)
+
+	# Both byte and uint8 luckily marshal as base64-encoded JSON strings.
+	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"
+}
+
+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")
+	CodegenSerialize["bool"] = \
+		"\tif %s {\n" \
+		"\t\tdata = append(data, 1)\n" \
+		"\t} else {\n" \
+		"\t\tdata = append(data, 0)\n" \
+		"\t}\n"
+	CodegenDeserialize["bool"] = \
+		"\tif data, ok = protoConsumeBoolFrom(data, &%s); !ok {\n" \
+		"\t\treturn nil, ok\n" \
+		"\t}\n"
+
+	define_internal("string", "string")
+	CodegenSerialize["string"] = \
+		"\tif data, ok = protoAppendStringTo(data, %s); !ok {\n" \
+		"\t\treturn nil, ok\n" \
+		"\t}\n"
+	CodegenDeserialize["string"] = \
+		"\tif data, ok = protoConsumeStringFrom(data, &%s); !ok {\n" \
+		"\t\treturn nil, ok\n" \
+		"\t}\n"
+
+	print "package main"
+	print ""
+	print "import ("
+	print "\t`encoding/binary`"
+	print "\t`encoding/json`"
+	print "\t`errors`"
+	print "\t`math`"
+	print "\t`strconv`"
+	print "\t`unicode/utf8`"
+	print ")"
+	print ""
+
+	print "// protoConsumeBoolFrom 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 protoConsumeBoolFrom(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 ""
+
+	print "// protoAppendStringTo tries to serialize a string value,"
+	print "// appending it to the end of a byte stream."
+	print "func protoAppendStringTo(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 ""
+
+	print "// protoConsumeStringFrom 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 protoConsumeStringFrom(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 ""
+
+	print "// protoUnmarshalEnumJSON converts a JSON fragment to an integer,"
+	print "// ensuring that it's within the expected range of enum values."
+	print "func protoUnmarshalEnumJSON(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.MaxInt32 || n < math.MinInt32 {"
+	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) {
+	gotype = PrefixCamel name
+	print "type " gotype " int"
+	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 ""
+
+	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 json.Marshal(v.String())"
+	print "\t}"
+	print "\treturn json.Marshal(int(v))"
+	print "}"
+	print ""
+
+	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 := protoUnmarshalEnumJSON(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 = binary.BigEndian.AppendUint32(data, uint32(%s))\n"
+	CodegenDeserialize[name] = \
+		"\tif len(data) >= 4 {\n" \
+		"\t\t%s = " gotype "(int32(binary.BigEndian.Uint32(data)))\n" \
+		"\t\tdata = data[4:]\n" \
+		"\t} else {\n" \
+		"\t\treturn nil, false\n" \
+		"\t}\n"
+
+	CodegenGoType[name] = gotype
+	for (i in cg)
+		delete cg[i]
+}
+
+function codegen_struct_field(d, cg,    camel, f, serialize, deserialize) {
+	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) {
+	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["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 ""
+
+	print "func (u *" gotype ") MarshalJSON() ([]byte, error) {"
+	print "\treturn json.Marshal(u.Interface)"
+	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]
+}

xC-gen-proto.awk

@@ -0,0 +1,303 @@
+# xC-gen-proto.awk: an XDR-derived code generator for network protocols.
+#
+# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# SPDX-License-Identifier: 0BSD
+#
+# You may read RFC 4506 for context, however it is only a source of inspiration.
+# Grammar is easy to deduce from the parser.
+#
+# Native types: bool, u{8,16,32,64}, i{8,16,32,64}, string
+#
+# Don't define any new types, unless you hate yourself, then it's okay to do so.
+# Both backends are a pain in the arse, for different reasons.
+#
+# All numbers are encoded in big-endian byte order.
+# Booleans are one byte each.
+# Strings must be valid UTF-8, use u8<> to lift that restriction.
+# String and array lengths are encoded as u32.
+# Enumeration values automatically start at 1, and are encoded as i32.
+# Any struct or union field may be a variable-length array.
+#
+# Message framing is done externally, but also happens to prefix u32 lengths.
+#
+# Usage: env LC_ALL=C awk -v prefix=Relay \
+#  -f xC-gen-proto.awk        < xC-proto \
+#  -f xC-gen-proto-{c,go}.awk > xC-proto.{c,go} | {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
+		} else if (match($0, /^(0[xX][0-9a-fA-F]+|[1-9][0-9]*)/)) {
+			Token = substr($0, 1, RLENGTH)
+			$0 = substr($0, RLENGTH + 1)
+			return Token
+		} else if (/./) {
+			Token = substr($0, 1, 1)
+			$0 = substr($0, 2)
+			return Token
+		}
+	} while (/./ || 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 !~ /^[0-9]/)
+		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()
+		if (!value)
+			fatal("enumeration values cannot be zero")
+		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
+}
+
+BEGIN {
+	PrefixLower = "relay_"
+	PrefixUpper = "RELAY_"
+	PrefixCamel = "Relay"
+
+	print "// Generated by xC-gen-proto.awk. DO NOT MODIFY."
+	codegen_begin()
+
+	nexttoken()
+	while (Token != "") {
+		expect(defconst() || deftype())
+		expect(accept(";"))
+	}
+}

xC-proto

@@ -0,0 +1,120 @@
+// Backwards-compatible protocol version.
+const VERSION = 1;
+
+// From the frontend to the relay.
+struct CommandMessage {
+	u32 command_seq;
+	union CommandData switch (enum Command {
+		HELLO,
+		PING,
+		ACTIVE,
+		BUFFER_COMPLETE,
+		BUFFER_INPUT,
+		BUFFER_ACTIVATE,
+		BUFFER_LOG,
+	} command) {
+	case HELLO:
+		u32 version;
+	case PING:
+		void;
+	case ACTIVE:
+		void;
+	case BUFFER_COMPLETE:
+		string buffer_name;
+		string text;
+		u32 position;
+	case BUFFER_INPUT:
+		string buffer_name;
+		string text;
+	case BUFFER_ACTIVATE:
+		string buffer_name;
+	case BUFFER_LOG:
+		string buffer_name;
+	} data;
+};
+
+// From the relay to the frontend.
+struct EventMessage {
+	u32 event_seq;
+	union EventData switch (enum Event {
+		PING,
+		BUFFER_UPDATE,
+		BUFFER_RENAME,
+		BUFFER_REMOVE,
+		BUFFER_ACTIVATE,
+		BUFFER_LINE,
+		BUFFER_CLEAR,
+		ERROR,
+		RESPONSE,
+	} event) {
+	case PING:
+		void;
+	case BUFFER_UPDATE:
+		string buffer_name;
+	case BUFFER_RENAME:
+		string buffer_name;
+		string new;
+	case BUFFER_REMOVE:
+		string buffer_name;
+	case BUFFER_ACTIVATE:
+		string buffer_name;
+	case BUFFER_LINE:
+		string buffer_name;
+		bool is_unimportant;
+		bool is_highlight;
+		enum Rendition {
+			BARE,
+			INDENT,
+			STATUS,
+			ERROR,
+			JOIN,
+			PART,
+		} rendition;
+		// Unix timestamp in seconds.
+		u64 when;
+		// Broken-up text, with in-band formatting.
+		union ItemData switch (enum Item {
+			TEXT,
+			RESET,
+			FG_COLOR,
+			BG_COLOR,
+			FLIP_BOLD,
+			FLIP_ITALIC,
+			FLIP_UNDERLINE,
+			FLIP_INVERSE,
+			FLIP_CROSSED_OUT,
+			FLIP_MONOSPACE,
+		} kind) {
+		case TEXT:
+			string text;
+		case RESET:
+			void;
+		case FG_COLOR:
+			i16 color;
+		case BG_COLOR:
+			i16 color;
+		case FLIP_BOLD:
+		case FLIP_ITALIC:
+		case FLIP_UNDERLINE:
+		case FLIP_INVERSE:
+		case FLIP_CROSSED_OUT:
+		case FLIP_MONOSPACE:
+			void;
+		} items<>;
+	case BUFFER_CLEAR:
+		string buffer_name;
+	case ERROR:
+		u32 command_seq;
+		string error;
+	case RESPONSE:
+		u32 command_seq;
+		union ResponseData switch (Command command) {
+		case BUFFER_COMPLETE:
+			u32 start;
+			string completions<>;
+		case BUFFER_LOG:
+			// UTF-8, but not guaranteed.
+			u8 log<>;
+		} data;
+	} data;
+};

xC.adoc

@@ -1,8 +1,8 @@
 xC(1)
 =====
 :doctype: manpage
-:manmanual: uirc3 Manual
-:mansource: uirc3 {release-version}
+:manmanual: xK Manual
+:mansource: xK {release-version}
 
 Name
 ----
@@ -25,9 +25,9 @@ Options
 	other formatting marks to ANSI codes retrieved from the *terminfo*(5)
 	database:
 +
-```
+....
 printf '\x02bold\x02\n' | xC -f
-```
+....
 +
 This feature may be used to preview server MOTD files.
 
@@ -62,10 +62,10 @@ their respective function names:
 *M-a*: *goto-activity*::
 	Go to the first following buffer with unseen activity.
 *PageUp*: *display-backlog*::
-	Show the in-memory backlog for this buffer in the backlog helper,
+	Show the in-memory backlog for this buffer using *general.pager*,
 	which is almost certainly the *less*(1) program.
 *M-h*: *display-full-log*::
-	Show the log file for this buffer in the backlog helper.
+	Show the log file for this buffer using *general.pager*.
 *M-H*: *toggle-unimportant*::
 	Hide all join, part and quit messages, as well as all channel mode changes
 	that only relate to user channel modes.  Intended to reduce noise in
@@ -105,7 +105,7 @@ _~/.config/xC/xC.conf_::
 	as the */set* command, to make changes in it.
 
 _~/.local/share/xC/logs/_::
-	When enabled by *behaviour.logging*, log files are stored here.
+	When enabled by *general.logging*, log files are stored here.
 
 _~/.local/share/xC/plugins/_::
 _/usr/local/share/xC/plugins/_::
@@ -114,12 +114,11 @@ _/usr/share/xC/plugins/_::
 
 Bugs
 ----
-The editline (libedit) frontend is more of a proof of concept that mostly seems
-to work but exhibits bugs that are not our fault.
+The editline (libedit) frontend may exhibit some unexpected behaviour.
 
 Reporting bugs
 --------------
-Use https://git.janouch.name/p/uirc3 to report bugs, request features,
+Use https://git.janouch.name/p/xK to report bugs, request features,
 or submit pull requests.
 
 See also

xD-gen-replies.awk

@@ -0,0 +1,29 @@
+#!/usr/bin/awk -f
+BEGIN {
+	# The message catalog is a by-product
+	msg = "xD.msg"
+	print "$quote \"" > msg;
+	print "$set 1" > msg;
+}
+
+/^[0-9]+ *IRC_(ERR|RPL)_[A-Z]+ *".*"$/ {
+	match($0, /".*"/);
+	ids[$1] = $2;
+	texts[$2] = substr($0, RSTART, RLENGTH);
+	print $1 " " texts[$2] > msg
+}
+
+END {
+	printf("enum\n{")
+	for (i in ids) {
+		if (seen_first)
+			printf(",")
+		seen_first = 1
+		printf("\n\t%s = %s", ids[i], i)
+	}
+	print "\n};\n"
+	print "static const char *g_default_replies[] =\n{"
+	for (i in ids)
+		print "\t[" ids[i] "] = " texts[ids[i]] ","
+	print "};"
+}

xD-gen-replies.sh

@@ -1,28 +0,0 @@
-#!/bin/sh
-LC_ALL=C exec awk '
-	BEGIN {
-		# The message catalog is a by-product
-		msg = "xD.msg"
-		print "$quote \"" > msg;
-		print "$set 1" > msg;
-	}
-	/^[0-9]+ *IRC_(ERR|RPL)_[A-Z]+ *".*"$/ {
-		match($0, /".*"/);
-		ids[$1] = $2;
-		texts[$2] = substr($0, RSTART, RLENGTH);
-		print $1 " " texts[$2] > msg
-	}
-	END {
-		printf("enum\n{")
-		for (i in ids) {
-			if (seen_first)
-				printf(",")
-			seen_first = 1
-			printf("\n\t%s = %s", ids[i], i)
-		}
-		print "\n};\n"
-		print "static const char *g_default_replies[] =\n{"
-		for (i in ids)
-			print "\t[" ids[i] "] = " texts[ids[i]] ","
-		print "};"
-	}'

xD.adoc

@@ -1,8 +1,8 @@
 xD(1)
 =====
 :doctype: manpage
-:manmanual: uirc3 Manual
-:mansource: uirc3 {release-version}
+:manmanual: xK Manual
+:mansource: xK {release-version}
 
 Name
 ----
@@ -49,5 +49,5 @@ _/etc/xdg/xD/xD.conf_::
 
 Reporting bugs
 --------------
-Use https://git.janouch.name/p/uirc3 to report bugs, request features,
+Use https://git.janouch.name/p/xK to report bugs, request features,
 or submit pull requests.

xD.c

@@ -1,7 +1,7 @@
 /*
  * xD.c: an IRC daemon
  *
- * Copyright (c) 2014 - 2020, 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.
@@ -22,7 +22,9 @@
 #define WANT_SYSLOG_LOGGING
 #include "common.c"
 #include "xD-replies.c"
+
 #include <nl_types.h>
+#include <sys/resource.h>
 
 enum { PIPE_READ, PIPE_WRITE };
 
@@ -851,8 +853,6 @@ client_send_str (struct client *c, const struct str *s)
 	str_append_data (&c->write_buffer, s->str,
 		MIN (s->len, IRC_MAX_MESSAGE_LENGTH));
 	str_append (&c->write_buffer, "\r\n");
-	// XXX: we might want to move this elsewhere, so that it doesn't get called
-	//   as often; it's going to cause a lot of syscalls with epoll.
 	client_update_poller (c, NULL);
 
 	// Technically we haven't sent it yet but that's a minor detail
@@ -2930,6 +2930,29 @@ irc_handle_links (const struct irc_message *msg, struct client *c)
 	irc_send_reply (c, IRC_RPL_ENDOFLINKS, mask);
 }
 
+static void
+irc_handle_wallops (const struct irc_message *msg, struct client *c)
+{
+	if (msg->params.len < 1)
+		RETURN_WITH_REPLY (c, IRC_ERR_NEEDMOREPARAMS, msg->command);
+	if (!(c->mode & IRC_USER_MODE_OPERATOR))
+		RETURN_WITH_REPLY (c, IRC_ERR_NOPRIVILEGES);
+
+	const char *message = msg->params.vector[0];
+
+	// Our interpretation: anonymize the sender,
+	// and target all users who want to receive these messages
+	struct str_map_iter iter = str_map_iter_make (&c->ctx->users);
+	struct client *target;
+	while ((target = str_map_iter_next (&iter)))
+	{
+		if (target != c && !(target->mode & IRC_USER_MODE_RX_WALLOPS))
+			continue;
+
+		client_send (target, ":%s WALLOPS :%s", c->ctx->server_name, message);
+	}
+}
+
 static void
 irc_handle_kill (const struct irc_message *msg, struct client *c)
 {
@@ -2992,6 +3015,7 @@ irc_register_handlers (struct server_context *ctx)
 		{ "ADMIN",    true,  irc_handle_admin,    0, 0 },
 		{ "STATS",    true,  irc_handle_stats,    0, 0 },
 		{ "LINKS",    true,  irc_handle_links,    0, 0 },
+		{ "WALLOPS",  true,  irc_handle_wallops,  0, 0 },
 
 		{ "MODE",     true,  irc_handle_mode,     0, 0 },
 		{ "PRIVMSG",  true,  irc_handle_privmsg,  0, 0 },
@@ -3069,6 +3093,7 @@ irc_try_read (struct client *c)
 		{
 			buf->str[buf->len += n_read] = '\0';
 			// TODO: discard characters above the 512 character limit
+			// FIXME: we should probably discard the data if closing_link
 			irc_process_buffer (buf, irc_process_message, c);
 			continue;
 		}
@@ -3110,6 +3135,7 @@ irc_try_read_tls (struct client *c)
 		case SSL_ERROR_NONE:
 			buf->str[buf->len += n_read] = '\0';
 			// TODO: discard characters above the 512 character limit
+			// FIXME: we should probably discard the data if closing_link
 			irc_process_buffer (buf, irc_process_message, c);
 			continue;
 		case SSL_ERROR_ZERO_RETURN:
@@ -3395,16 +3421,10 @@ irc_try_fetch_client (struct server_context *ctx, int listen_fd)
 		if (errno == EINTR)
 			return true;
 
-		if (errno == EBADF
-		 || errno == EINVAL
-		 || errno == ENOTSOCK
-		 || errno == EOPNOTSUPP)
+		if (accept_error_is_transient (errno))
+			print_warning ("%s: %s", "accept", strerror (errno));
+		else
 			print_fatal ("%s: %s", "accept", strerror (errno));
-
-		// OS kernels may return a wide range of unforeseeable errors.
-		// Assuming that they're either transient or caused by
-		// a connection that we've just extracted from the queue.
-		print_warning ("%s: %s", "accept", strerror (errno));
 		return true;
 	}
 
@@ -3788,10 +3808,9 @@ irc_lock_pid_file (struct server_context *ctx, struct error **e)
 }
 
 static int
-irc_listen (struct addrinfo *gai_iter)
+irc_listen (struct addrinfo *ai)
 {
-	int fd = socket (gai_iter->ai_family,
-		gai_iter->ai_socktype, gai_iter->ai_protocol);
+	int fd = socket (ai->ai_family, ai->ai_socktype, ai->ai_protocol);
 	if (fd == -1)
 		return -1;
 	set_cloexec (fd);
@@ -3805,21 +3824,13 @@ irc_listen (struct addrinfo *gai_iter)
 #if defined SOL_IPV6 && defined IPV6_V6ONLY
 	// Make NULL always bind to both IPv4 and IPv6, irrespectively of the order
 	// of results; only INADDR6_ANY seems to be affected by this
-	if (gai_iter->ai_family == AF_INET6)
+	if (ai->ai_family == AF_INET6)
 		soft_assert (setsockopt (fd, SOL_IPV6, IPV6_V6ONLY,
 			&yes, sizeof yes) != -1);
 #endif
 
-	char host[NI_MAXHOST], port[NI_MAXSERV];
-	host[0] = port[0] = '\0';
-	int err = getnameinfo (gai_iter->ai_addr, gai_iter->ai_addrlen,
-		host, sizeof host, port, sizeof port,
-		NI_NUMERICHOST | NI_NUMERICSERV);
-	if (err)
-		print_debug ("%s: %s", "getnameinfo", gai_strerror (err));
-
-	char *address = format_host_port_pair (host, port);
-	if (bind (fd, gai_iter->ai_addr, gai_iter->ai_addrlen))
+	char *address = gai_reconstruct_address (ai);
+	if (bind (fd, ai->ai_addr, ai->ai_addrlen))
 		print_error ("bind to %s failed: %s", address, strerror (errno));
 	else if (listen (fd, 16 /* arbitrary number */))
 		print_error ("listen on %s failed: %s", address, strerror (errno));
@@ -3839,12 +3850,12 @@ static void
 irc_listen_resolve (struct server_context *ctx,
 	const char *host, const char *port, struct addrinfo *gai_hints)
 {
-	struct addrinfo *gai_result, *gai_iter;
+	struct addrinfo *gai_result = NULL, *gai_iter = NULL;
 	int err = getaddrinfo (host, port, gai_hints, &gai_result);
 	if (err)
 	{
 		char *address = format_host_port_pair (host, port);
-		print_error ("bind to %s failed: %s: %s",
+		print_error ("binding to %s failed: %s: %s",
 			address, "getaddrinfo", gai_strerror (err));
 		free (address);
 		return;
@@ -3984,6 +3995,21 @@ daemonize (struct server_context *ctx)
 	poller_post_fork (&ctx->poller);
 }
 
+static void
+setup_limits (void)
+{
+	struct rlimit limit;
+	if (getrlimit (RLIMIT_NOFILE, &limit))
+	{
+		print_warning ("%s: %s", "getrlimit", strerror (errno));
+		return;
+	}
+
+	limit.rlim_cur = limit.rlim_max;
+	if (setrlimit (RLIMIT_NOFILE, &limit))
+		print_warning ("%s: %s", "setrlimit", strerror (errno));
+}
+
 int
 main (int argc, char *argv[])
 {
@@ -4030,6 +4056,7 @@ main (int argc, char *argv[])
 
 	print_status (PROGRAM_NAME " " PROGRAM_VERSION " starting");
 	setup_signal_handlers ();
+	setup_limits ();
 	init_openssl ();
 
 	struct server_context ctx;

xF.c

@@ -0,0 +1,172 @@
+/*
+ * xF.c: a toothless IRC client frontend
+ *
+ * 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.
+ *
+ */
+
+#include "config.h"
+#define PROGRAM_NAME "xF"
+
+#include "common.c"
+#include "xC-proto.c"
+
+#include <X11/Xatom.h>
+#include <X11/Xlib.h>
+#include <X11/keysym.h>
+#include <X11/XKBlib.h>
+#include <X11/Xft/Xft.h>
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+static struct
+{
+	bool polling;
+	struct connector connector;
+	int socket;
+}
+g;
+
+static void
+on_connector_connecting (void *user_data, const char *address)
+{
+	(void) user_data;
+	print_status ("connecting to %s...", address);
+}
+
+static void
+on_connector_error (void *user_data, const char *error)
+{
+	(void) user_data;
+	print_status ("connection failed: %s", error);
+}
+
+static void
+on_connector_failure (void *user_data)
+{
+	(void) user_data;
+	exit_fatal ("giving up");
+}
+
+static void
+on_connector_connected (void *user_data, int socket, const char *hostname)
+{
+	(void) user_data;
+	(void) hostname;
+	g.polling = false;
+	g.socket = socket;
+}
+
+static void
+protocol_test (const char *host, const char *port)
+{
+	struct poller poller = {};
+	poller_init (&poller);
+
+	connector_init (&g.connector, &poller);
+	g.connector.on_connecting = on_connector_connecting;
+	g.connector.on_error      = on_connector_error;
+	g.connector.on_connected  = on_connector_connected;
+	g.connector.on_failure    = on_connector_failure;
+
+	connector_add_target (&g.connector, host, port);
+
+	g.polling = true;
+	while (g.polling)
+		poller_run (&poller);
+
+	connector_free (&g.connector);
+
+	struct str s = str_make ();
+	str_pack_u32 (&s, 0);
+	struct relay_command_message m = {};
+	m.data.hello.command = RELAY_COMMAND_HELLO;
+	m.data.hello.version = RELAY_VERSION;
+	if (!relay_command_message_serialize (&m, &s))
+		exit_fatal ("serialization failed");
+
+	uint32_t len = htonl (s.len - sizeof len);
+	memcpy (s.str, &len, sizeof len);
+	if (errno = 0, write (g.socket, s.str, s.len) != (ssize_t) s.len)
+		exit_fatal ("short send or error: %s", strerror (errno));
+
+	char buf[1 << 20] = "";
+	while (errno = 0, read (g.socket, &len, sizeof len) == sizeof len)
+	{
+		len = ntohl (len);
+		if (errno = 0, read (g.socket, buf, MIN (len, sizeof buf)) != len)
+			exit_fatal ("short read or error: %s", strerror (errno));
+
+		struct msg_unpacker r = msg_unpacker_make (buf, len);
+		struct relay_event_message m = {};
+		if (!relay_event_message_deserialize (&m, &r))
+			exit_fatal ("deserialization failed");
+		if (msg_unpacker_get_available (&r))
+			exit_fatal ("trailing data");
+
+		printf ("event: %d\n", m.data.event);
+		relay_event_message_free (&m);
+	}
+	exit_fatal ("short read or error: %s", strerror (errno));
+}
+
+// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+
+int
+main (int argc, char *argv[])
+{
+	static const struct opt opts[] =
+	{
+		{ 'h', "help", NULL, 0, "display this help and exit" },
+		{ 'V', "version", NULL, 0, "output version information and exit" },
+		{ 0, NULL, NULL, 0, NULL }
+	};
+
+	struct opt_handler oh = opt_handler_make (argc, argv, opts,
+		"HOST:PORT", "X11 frontend for xC.");
+
+	int c;
+	while ((c = opt_handler_get (&oh)) != -1)
+	switch (c)
+	{
+	case 'h':
+		opt_handler_usage (&oh, stdout);
+		exit (EXIT_SUCCESS);
+	case 'V':
+		printf (PROGRAM_NAME " " PROGRAM_VERSION "\n");
+		exit (EXIT_SUCCESS);
+	default:
+		print_error ("wrong options");
+		opt_handler_usage (&oh, stderr);
+		exit (EXIT_FAILURE);
+	}
+
+	argc -= optind;
+	argv += optind;
+	if (argc != 1)
+	{
+		opt_handler_usage (&oh, stderr);
+		exit (EXIT_FAILURE);
+	}
+	opt_handler_free (&oh);
+
+	char *address = xstrdup (argv[0]);
+	const char *port = NULL, *host = tokenize_host_port (address, &port);
+	if (!port)
+		exit_fatal ("missing port number/service name");
+
+	// TODO: Actually implement an X11-based user interface.
+	protocol_test (host, port);
+	return 0;
+}

xF.svg

@@ -0,0 +1,36 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg version="1.1" width="48" height="48" viewBox="0 0 48 48"
+   xmlns:xlink="http://www.w3.org/1999/xlink"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:svg="http://www.w3.org/2000/svg">
+
+  <defs>
+    <linearGradient id="background" x1="0" y1="0" x2="1" y2="1">
+      <stop stop-color="#808080" offset="0" />
+      <stop stop-color="#000000" offset="1" />
+    </linearGradient>
+    <!-- librsvg screws up the filter's orientation in a weird way
+         otherwise a larger blur value would look better -->
+    <filter id="shadow" color-interpolation-filters="sRGB">
+      <feOffset dy="0.5" />
+      <feGaussianBlur stdDeviation="0.5" />
+      <feComposite in2="SourceGraphic" operator="in" />
+    </filter>
+    <clipPath id="clip">
+      <rect x="-7" y="-10" width="14" height="20" />
+    </clipPath>
+  </defs>
+
+  <circle cx="24" cy="24" r="20"
+     fill="url(#background)" stroke="#404040" stroke-width="2" />
+
+  <g transform="rotate(-45 24 24)" filter="url(#shadow)">
+    <path d="m 12,25 h 24 v 11 h -5 v -8 h -4.5 v 6 h -5 v -6 h -9.5 z"
+       fill="#ffffff" />
+    <g stroke-width="4" transform="translate(24, 16)" clip-path="url(#clip)"
+       stroke="#ffffff">
+      <line x1="-8" x2="8" y1="-5" y2="5" />
+      <line x1="-8" x2="8" y1="5" y2="-5" />
+    </g>
+  </g>
+</svg>

xP/.gitignore

@@ -0,0 +1,3 @@
+/xP
+/proto.go
+/public/mithril.js

xP/Makefile

@@ -0,0 +1,14 @@
+.POSIX:
+.SUFFIXES:
+
+outputs = xP proto.go public/mithril.js
+all: $(outputs)
+
+xP: xP.go proto.go
+	go build -o $@
+proto.go: ../xC-gen-proto.awk ../xC-gen-proto-go.awk ../xC-proto
+	awk -f ../xC-gen-proto.awk -f ../xC-gen-proto-go.awk ../xC-proto > $@
+public/mithril.js:
+	curl -Lo $@ https://unpkg.com/mithril/mithril.js
+clean:
+	rm -f $(outputs)

xP/go.mod

@@ -0,0 +1,5 @@
+module janouch.name/xK
+
+go 1.18
+
+require golang.org/x/net v0.0.0-20210805182204-aaa1db679c0d

xP/go.sum

@@ -0,0 +1,2 @@
+golang.org/x/net v0.0.0-20210805182204-aaa1db679c0d h1:20cMwl2fHAzkJMEA+8J4JgqBQcQGzbisXo31MIeenXI=
+golang.org/x/net v0.0.0-20210805182204-aaa1db679c0d/go.mod h1:9nx3DQGgdP8bBQD5qxJ1jj9UTztislL4KSBs9R2vV5Y=

xP/public/xP.css

@@ -0,0 +1,109 @@
+body {
+	margin: 0;
+	padding: 0;
+	font-family: sans-serif;
+}
+
+.xP {
+	height: 100vh;
+	display: flex;
+	flex-direction: column;
+	overflow: hidden;
+}
+
+.title, .status {
+	background: #f8f8f8;
+	border-bottom: 1px solid #ccc;
+	padding: .05rem .3rem;
+}
+
+.middle {
+	flex: auto;
+	display: flex;
+	flex-direction: row;
+	overflow: hidden;
+}
+
+.list {
+	overflow-y: auto;
+	border-right: 1px solid #ccc;
+	min-width: 10rem;
+}
+.item {
+	padding: .05rem .3rem;
+	cursor: default;
+}
+.item.active {
+	font-weight: bold;
+}
+
+/* Only Firefox currently supports align-content: safe end, thus this. */
+.buffer-container {
+	flex: auto;
+	display: flex;
+	flex-direction: column;
+	overflow: hidden;
+}
+.filler {
+	flex: auto;
+}
+.buffer {
+	display: grid;
+	grid-template-columns: max-content auto;
+	overflow-y: auto;
+}
+
+.date {
+	padding: .3rem;
+	grid-column: span 2;
+	font-weight: bold;
+}
+.time {
+	padding: .1rem .3rem;
+	background: #f8f8f8;
+	color: #bbb;
+	border-right: 1px solid #ccc;
+}
+.mark {
+	padding-right: .3rem;
+	text-align: center;
+	display: inline-block;
+	min-width: 2rem;
+}
+.mark.error {
+	color: red;
+}
+.mark.join {
+	color: green;
+}
+.mark.part {
+	color: red;
+}
+.content {
+	padding: .1rem .3rem;
+	white-space: pre-wrap;
+}
+.content span.b {
+	font-weight: bold;
+}
+.content span.i {
+	font-style: italic;
+}
+.content span.u {
+	text-decoration: underline;
+}
+.content span.s {
+	text-decoration: line-through;
+}
+.content span.m {
+	font-family: monospace;
+}
+
+.status {
+	border-top: 2px solid #fff;
+}
+
+textarea {
+	padding: .05rem .3rem;
+	font-family: inherit;
+}

xP/public/xP.js

@@ -0,0 +1,188 @@
+// TODO: Probably reset state on disconnect, and indicate to user.
+let socket = new WebSocket(proxy)
+
+let commandSeq = 0
+function send(command) {
+	socket.send(JSON.stringify({commandSeq: ++commandSeq, data: command}))
+}
+
+socket.onopen = function(event) {
+	send({command: 'Hello', version: 1})
+}
+
+let buffers = new Map()
+let bufferCurrent = undefined
+
+socket.onmessage = function(event) {
+	console.log(event.data)
+
+	let e = JSON.parse(event.data).data
+	switch (e.event) {
+	case 'BufferUpdate':
+	{
+		let b = buffers.get(e.bufferName)
+		if (b === undefined) {
+			b = {lines: []}
+			buffers.set(e.bufferName, b)
+		}
+		// TODO: Update any buffer properties.
+		break
+	}
+	case 'BufferRename':
+		buffers.set(e.new, buffers.get(e.bufferName))
+		buffers.delete(e.bufferName)
+		break
+	case 'BufferRemove':
+		buffers.delete(e.bufferName)
+		break
+	case 'BufferActivate':
+		bufferCurrent = e.bufferName
+		// TODO: Somehow scroll to the end of it immediately.
+		// TODO: Focus the textarea.
+		break
+	case 'BufferLine':
+	{
+		let b = buffers.get(e.bufferName)
+		if (b !== undefined)
+			b.lines.push({when: e.when, rendition: e.rendition, items: e.items})
+		break
+	}
+	case 'BufferClear':
+	{
+		let b = buffers.get(e.bufferName)
+		if (b !== undefined)
+			b.lines.length = 0
+		break
+	}
+	}
+
+	m.redraw()
+}
+
+let BufferList = {
+	view: vnode => {
+		let items = []
+		buffers.forEach((b, name) => {
+			let attrs = {
+				onclick: e => {
+					send({command: 'BufferActivate', bufferName: name})
+				},
+			}
+			if (name == bufferCurrent)
+				attrs.class = 'active'
+			items.push(m('.item', attrs, name))
+		})
+		return m('.list', {}, items)
+	},
+}
+
+let Content = {
+	view: vnode => {
+		let line = vnode.children[0]
+		let content = []
+		switch (line.rendition) {
+		case 'Indent': content.push(m('span.mark',       {}, ''));  break
+		case 'Status': content.push(m('span.mark',       {}, '–')); break
+		case 'Error':  content.push(m('span.mark.error', {}, '⚠')); break
+		case 'Join':   content.push(m('span.mark.join',  {}, '→')); break
+		case 'Part':   content.push(m('span.mark.part',  {}, '←')); break
+		}
+
+		let classes = new Set()
+		let flip = c => {
+			if (classes.has(c))
+				classes.delete(c)
+			else
+				classes.add(c)
+		}
+		line.items.forEach(item => {
+			// TODO: Colours.
+			switch (item.kind) {
+			case 'Text':
+				// TODO: Detect and transform links.
+				content.push(m('span', {
+					class: Array.from(classes.keys()).join(' '),
+				}, item.text))
+				break
+			case 'Reset':
+				classes.clear()
+				break
+			case 'FlipBold':       flip('b'); break
+			case 'FlipItalic':     flip('i'); break
+			case 'FlipUnderline':  flip('u'); break
+			case 'FlipInverse':    flip('i'); break
+			case 'FlipCrossedOut': flip('s'); break
+			case 'FlipMonospace':  flip('m'); break
+			}
+		})
+		return m('.content', {}, content)
+	},
+}
+
+let Buffer = {
+	view: vnode => {
+		let lines = []
+		let b = buffers.get(bufferCurrent)
+		if (b === undefined)
+			return
+
+		let lastDateMark = undefined
+		b.lines.forEach(line => {
+			let date = new Date(line.when * 1000)
+			let dateMark = date.toLocaleDateString()
+			if (dateMark !== lastDateMark) {
+				lines.push(m('.date', {}, dateMark))
+				lastDateMark = dateMark
+			}
+
+			lines.push(m('.time', {}, date.toLocaleTimeString()))
+			lines.push(m(Content, {}, line))
+		})
+		return m('.buffer-container', {}, [
+			m('.filler'),
+			m('.buffer', {}, lines),
+		])
+	},
+}
+
+// TODO: This should be remembered across buffer switches,
+// and we'll probably have to intercept /all/ key presses.
+let Input = {
+	view: vnode => {
+		return m('textarea', {
+			rows: 1,
+			onkeydown: e => {
+				// TODO: And perhaps on other actions, too.
+				send({command: 'Active'})
+				if (e.keyCode !== 13)
+					return
+
+				send({
+					command: 'BufferInput',
+					bufferName: bufferCurrent,
+					text: e.currentTarget.value,
+				})
+				e.preventDefault()
+				e.currentTarget.value = ''
+			},
+		})
+	},
+}
+
+let Main = {
+	view: vnode => {
+		return m('.xP', {}, [
+			m('.title', {}, "xP"),
+			m('.middle', {}, [m(BufferList), m(Buffer)]),
+			m('.status', {}, bufferCurrent),
+			m(Input),
+		])
+	},
+}
+
+// TODO: Buffer names should work as routes.
+window.addEventListener('load', () => {
+	m.route(document.body, '/', {
+		'/': Main,
+	})
+})

xP/xP.example.json

@@ -0,0 +1,2 @@
+{
+}

xP/xP.go

@@ -0,0 +1,186 @@
+package main
+
+import (
+	"context"
+	"encoding/binary"
+	"encoding/json"
+	"fmt"
+	"html/template"
+	"io"
+	"log"
+	"net"
+	"net/http"
+	"os"
+	"time"
+
+	"golang.org/x/net/websocket"
+)
+
+var (
+	addressBind    string
+	addressConnect string
+)
+
+func clientToRelay(
+	ctx context.Context, ws *websocket.Conn, conn net.Conn) bool {
+	var j string
+	if err := websocket.Message.Receive(ws, &j); err != nil {
+		log.Println("Command receive failed: " + err.Error())
+		return false
+	}
+
+	log.Printf("?> %s\n", j)
+
+	var m RelayCommandMessage
+	if err := json.Unmarshal([]byte(j), &m); err != nil {
+		log.Println("Command unmarshalling failed: " + err.Error())
+		return false
+	}
+
+	b, ok := m.AppendTo(make([]byte, 4))
+	if !ok {
+		log.Println("Command serialization failed")
+		return false
+	}
+	binary.BigEndian.PutUint32(b[:4], uint32(len(b)-4))
+	if _, err := conn.Write(b); err != nil {
+		log.Println("Command send failed: " + err.Error())
+		return false
+	}
+
+	log.Printf("-> %v\n", b)
+	return true
+}
+
+func relayToClient(
+	ctx context.Context, ws *websocket.Conn, conn net.Conn) bool {
+	var length uint32
+	if err := binary.Read(conn, binary.BigEndian, &length); err != nil {
+		log.Println("Event receive failed: " + err.Error())
+		return false
+	}
+	b := make([]byte, length)
+	if _, err := io.ReadFull(conn, b); err != nil {
+		log.Println("Event receive failed: " + err.Error())
+		return false
+	}
+
+	log.Printf("<? %v\n", b)
+
+	var m RelayEventMessage
+	if after, ok := m.ConsumeFrom(b); !ok {
+		log.Println("Event deserialization failed")
+		return false
+	} else if len(after) != 0 {
+		log.Println("Event deserialization failed: trailing data")
+		return false
+	}
+
+	j, err := json.Marshal(&m)
+	if err != nil {
+		log.Println("Event marshalling failed: " + err.Error())
+		return false
+	}
+	if err := websocket.Message.Send(ws, string(j)); err != nil {
+		log.Println("Event send failed: " + err.Error())
+		return false
+	}
+
+	log.Printf("<- %s\n", j)
+	return true
+}
+
+func errorToClient(ws *websocket.Conn, err error) bool {
+	j, err := json.Marshal(&RelayEventMessage{
+		EventSeq: 0,
+		Data: RelayEventData{
+			Interface: RelayEventDataError{
+				Event:      RelayEventError,
+				CommandSeq: 0,
+				Error:      err.Error(),
+			},
+		},
+	})
+	if err != nil {
+		log.Println("Event marshalling failed: " + err.Error())
+		return false
+	}
+	if err := websocket.Message.Send(ws, string(j)); err != nil {
+		log.Println("Event send failed: " + err.Error())
+		return false
+	}
+	return true
+}
+
+func handleWebSocket(ws *websocket.Conn) {
+	conn, err := net.Dial("tcp", addressConnect)
+	if err != nil {
+		errorToClient(ws, err)
+		return
+	}
+
+	// We don't need to intervene, so it's just two separate pipes so far.
+	ctx, cancel := context.WithCancel(ws.Request().Context())
+	go func() {
+		for clientToRelay(ctx, ws, conn) {
+		}
+		cancel()
+	}()
+	go func() {
+		for relayToClient(ctx, ws, conn) {
+		}
+		cancel()
+	}()
+	<-ctx.Done()
+}
+
+var staticHandler = http.FileServer(http.Dir("."))
+
+var page = template.Must(template.New("/").Parse(`<!DOCTYPE html>
+<html>
+<head>
+	<title>xP</title>
+	<meta charset="utf-8" />
+	<link rel="stylesheet" href="xP.css" />
+</head>
+<body>
+	<script src="mithril.js">
+	</script>
+	<script>
+	let proxy = '{{ . }}'
+	</script>
+	<script src="xP.js">
+	</script>
+</body>
+</html>`))
+
+func handleDefault(w http.ResponseWriter, r *http.Request) {
+	if r.URL.Path != "/" {
+		staticHandler.ServeHTTP(w, r)
+		return
+	}
+
+	wsURI := fmt.Sprintf("ws://%s/ws", r.Host)
+	if err := page.Execute(w, wsURI); err != nil {
+		log.Println("Template execution failed: " + err.Error())
+	}
+}
+
+func main() {
+	if len(os.Args) != 3 {
+		log.Fatalf("usage: %s BIND CONNECT\n", os.Args[0])
+	}
+
+	addressBind, addressConnect = os.Args[1], os.Args[2]
+
+	http.Handle("/ws", websocket.Handler(handleWebSocket))
+	http.Handle("/", http.HandlerFunc(handleDefault))
+
+	s := &http.Server{
+		Addr:           addressBind,
+		ReadTimeout:    60 * time.Second,
+		WriteTimeout:   60 * time.Second,
+		MaxHeaderBytes: 32 << 10,
+	}
+	log.Fatal(s.ListenAndServe())
+}