Update .gitignore

Since poll() is implemented in terms of kqueue() there,
it doesn't seem like this could have improved anything.

Besides man 3 ev, libevent code, and [1],
I haven't managed to find much relevant information.

[1] https://daniel.haxx.se/blog/2016/10/11/poll-on-mac-10-12-is-broken/

.clang-format

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

.gitignore

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

CMakeLists.txt

@@ -4,9 +4,9 @@ cmake_minimum_required (VERSION 2.8.5)
 # Moar warnings
 if ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUCC)
 	# -Wunused-function is pretty annoying here, as everything is static
-	set (wdisabled "-Wno-unused-function -Wno-implicit-fallthrough")
+	set (wdisabled "-Wno-unused-function")
 	set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -std=c99 -Wall -Wextra ${wdisabled}")
-endif ("${CMAKE_C_COMPILER_ID}" MATCHES "GNU" OR CMAKE_COMPILER_IS_GNUCC)
+endif ()
 
 # Dependencies
 set (CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake)
@@ -16,11 +16,9 @@ find_package (PkgConfig REQUIRED)
 pkg_check_modules (libssl REQUIRED libssl libcrypto)
 
 if ("${CMAKE_SYSTEM_NAME}" MATCHES "BSD")
-	include_directories (/usr/local/include)
-	link_directories (/usr/local/lib)
 	# Our POSIX version macros make these undefined
 	add_definitions (-D__BSD_VISIBLE=1 -D_BSD_SOURCE=1)
-endif ("${CMAKE_SYSTEM_NAME}" MATCHES "BSD")
+endif ()
 
 set (common_libraries ${libssl_LIBRARIES})
 include_directories (${libssl_INCLUDE_DIRS})
@@ -32,8 +30,8 @@ foreach (extra iconv rt)
 	find_library (extra_lib_${extra} ${extra})
 	if (extra_lib_${extra})
 		list (APPEND common_libraries ${extra})
-	endif (extra_lib_${extra})
+	endif ()
-endforeach (extra)
+endforeach ()
 
 # Build some unit tests
 include_directories (${PROJECT_SOURCE_DIR})
@@ -43,4 +41,4 @@ foreach (name liberty proto)
 	add_threads (test-${name})
 	target_link_libraries (test-${name} ${common_libraries})
 	add_test (NAME test-${name} COMMAND test-${name})
-endforeach (name)
+endforeach ()

liberty-proto.c

@@ -148,7 +148,7 @@ irc_free_message (struct irc_message *msg)
 
 static void
 irc_process_buffer (struct str *buf,
-	void (*callback)(const struct irc_message *, const char *, void *),
+	void (*callback) (const struct irc_message *, const char *, void *),
 	void *user_data)
 {
 	char *start = buf->str, *end = start + buf->len;

liberty.c

@@ -1749,10 +1749,9 @@ poller_run (struct poller *self)
 	self->revents_len = 0;
 }
 
-#elif defined (BSD)
+// Sort of similar to the epoll version.  Let's hope Darwin isn't broken,
-
+// that'd mean reimplementing this in terms of select() just because of Crapple.
-// Mac OS X's kqueue is fatally broken, or so I've been told; leaving it out.
+#elif defined (BSD) || defined (__APPLE__)
-// Otherwise this is sort of similar to the epoll version.
 
 #include <sys/types.h>
 #include <sys/event.h>
@@ -2748,7 +2747,8 @@ utf8_decode (const char **s, size_t len)
 	}
 
 	// In the middle of a character
-	if (sequence_len == 1)
+	// or an overlong sequence (subset, possibly MUTF-8, not supported)
+	if (sequence_len == 1 || *p == 0xC0 || *p == 0xC1)
 		return -1;
 
 	// Check the rest of the sequence
@@ -2765,6 +2765,13 @@ utf8_decode (const char **s, size_t len)
 	return cp;
 }
 
+static inline bool
+utf8_validate_cp (int32_t cp)
+{
+	// RFC 3629, CESU-8 not allowed
+	return cp >= 0 && cp <= 0x10FFFF && (cp < 0xD800 || cp > 0xDFFF);
+}
+
 /// Very rough UTF-8 validation, just makes sure codepoints can be iterated
 static bool
 utf8_validate (const char *s, size_t len)
@@ -2772,7 +2779,7 @@ utf8_validate (const char *s, size_t len)
 	const char *end = s + len;
 	int32_t codepoint;
 	while ((codepoint = utf8_decode (&s, end - s)) >= 0
-		&& codepoint <= 0x10FFFF /* TODO: better validations */)
+		&& utf8_validate_cp (codepoint))
 		;
 	return s == end;
 }
@@ -3239,7 +3246,7 @@ get_xdg_config_dirs (struct strv *out)
 	str_free (&config_home);
 
 	const char *xdg_config_dirs;
-	if (!(xdg_config_dirs = getenv ("XDG_CONFIG_DIRS")))
+	if (!(xdg_config_dirs = getenv ("XDG_CONFIG_DIRS")) || !*xdg_config_dirs)
 		xdg_config_dirs = "/etc/xdg";
 	cstr_split (xdg_config_dirs, ":", true, out);
 }
@@ -3264,7 +3271,7 @@ get_xdg_data_dirs (struct strv *out)
 	str_free (&data_home);
 
 	const char *xdg_data_dirs;
-	if (!(xdg_data_dirs = getenv ("XDG_DATA_DIRS")))
+	if (!(xdg_data_dirs = getenv ("XDG_DATA_DIRS")) || !*xdg_data_dirs)
 		xdg_data_dirs = "/usr/local/share/:/usr/share/";
 	cstr_split (xdg_data_dirs, ":", true, out);
 }
@@ -4409,7 +4416,7 @@ socket_io_try_write (int socket_fd, struct str *wb)
 //   char    = [\0-\177]  # or any Unicode codepoint in the UTF-8 encoding
 //   escape  = [\\"abfnrtv] / [xX][0-9A-Fa-f][0-9A-Fa-f]? / [0-7][0-7]?[0-7]?
 //
-//   integer = lws '-'? [0-9]+  # whatever strtoll() accepts on your system
+//   integer = lws [-+]? [0-9]+  # whatever strtoll() accepts on your system
 //   null    = lws 'null'
 //   boolean = lws 'yes'  / lws 'YES'  / lws 'no'    / lws 'NO'
 //           / lws 'on'   / lws 'ON'   / lws 'off'   / lws 'OFF'
@@ -4705,8 +4712,10 @@ config_item_write_string (struct str *output, const struct str *s)
 		else if (c == '\t')  str_append (output, "\\t");
 		else if (c == '\\')  str_append (output, "\\\\");
 		else if (c == '"')   str_append (output, "\\\"");
-		else if (c < 32)     str_append_printf (output, "\\x%02x", c);
+		else if (iscntrl_ascii (c))
-		else                 str_append_c      (output, c);
+			str_append_printf (output, "\\x%02x", c);
+		else
+			str_append_c (output, c);
 	}
 	str_append_c (output, '"');
 }

siphash.c

@@ -82,4 +82,3 @@ siphash (const unsigned char key[16], const unsigned char *m, size_t len)
 
 	return v0 ^ v1 ^ v2 ^ v3;
 }
-

tests/liberty.c

@@ -331,10 +331,14 @@ test_utf8 (void)
 	soft_assert (utf8_decode (&partial, 1) == -2);
 	soft_assert (utf8_decode (&empty,   0) == -1);
 
-	const char valid  [] = "2H₂ + O₂ ⇌ 2H₂O, R = 4.7 kΩ, ⌀ 200 mm";
+	const char valid_1[] = "2H₂ + O₂ ⇌ 2H₂O, R = 4.7 kΩ, ⌀ 200 mm";
-	const char invalid[] = "\xf0\x90\x28\xbc";
+	const char valid_2[] = "\xf0\x93\x82\xb9";
-	soft_assert ( utf8_validate (valid,   sizeof valid));
+	const char invalid_1[] = "\xf0\x90\x28\xbc";
-	soft_assert (!utf8_validate (invalid, sizeof invalid));
+	const char invalid_2[] = "\xc0\x80";
+	soft_assert ( utf8_validate (valid_1,   sizeof valid_1));
+	soft_assert ( utf8_validate (valid_2,   sizeof valid_2));
+	soft_assert (!utf8_validate (invalid_1, sizeof invalid_1));
+	soft_assert (!utf8_validate (invalid_2, sizeof invalid_2));
 
 	struct utf8_iter iter = utf8_iter_make ("fóọ");
 	size_t ch_len;