[httpc] 解决几个ipv6下面的请求规范的问题 (#2149)

* 按照RFC 3986的规范处理ipv6 host的格式化问题,ipv6地址需要用 []括起来
参考: https://www.rfc-editor.org/rfc/rfc3986#section-3.2.2

* 解决ipv6格式不对的问题

* 只有域名的情况下才 set_ext_host_name

lualib/http/httpc.lua

@@ -33,11 +33,11 @@ local function parse_host(host)
 
 	if host:find(":", colon1 + 1, true) then
 		-- two or more colons: ipv6
-		local ipv6, port = host:match("^%[(.-)%]:(%d+)$")
+		local ipv6, port = host:match("^%[(.-)%]:?(%d*)$")
 		if ipv6 then
-			return ipv6, tonumber(port), "ipv6"
+			return ipv6, port ~= "" and tonumber(port) or nil, "ipv6"
 		end
-		return host, nil, "ipv6"
+		error(string.format("Invalid host: bare IPv6 address '%s', use '[%s]' instead", host, host))
 	end
 
 	-- single colon: host:port
@@ -57,9 +57,14 @@ local function parse_url(host)
 		protocol = "http"
 		hostname = host
 	end
-	local htype
+	local hostheader = hostname
+	local htype, port
 	hostname, port, htype = parse_host(hostname)
-	return protocol, hostname, port or default_port[protocol] or error("Invalid protocol: " .. protocol), htype
+	port = port or default_port[protocol]
+	if not port then
+		error("Invalid protocol: " .. protocol)
+	end
+	return protocol, hostname, port, htype, hostheader
 end
 
 local SSLCTX_CLIENT = nil
@@ -91,7 +96,7 @@ local function gen_interface(protocol, fd, hostname)
 end
 
 local function connect(host, timeout)
-	local protocol, hostname, port, htype = parse_url(host)
+	local protocol, hostname, port, htype, hostheader = parse_url(host)
 	local hostaddr = hostname
 	if htype == "hostname" then
 		if async_dns then
@@ -116,9 +121,9 @@ local function connect(host, timeout)
 		end)
 	end
 	if interface.init then
-		interface.init(host)
+		interface.init(htype == "hostname" and hostname or nil)
 	end
-	return fd, interface, host
+	return fd, interface, hostheader
 end
 
 local function close_interface(interface, fd)

lualib/http/tlshelper.lua

@@ -7,7 +7,9 @@ function tlshelper.init_requestfunc(fd, tls_ctx)
     local readfunc = socket.readfunc(fd)
     local writefunc = socket.writefunc(fd)
     return function (hostname)
-        tls_ctx:set_ext_host_name(hostname)
+        if hostname then
+            tls_ctx:set_ext_host_name(hostname)
+        end
         local ds1 = tls_ctx:handshake()
         writefunc(ds1)
         while not tls_ctx:finished() do

test/testhttp.lua

@@ -99,6 +99,25 @@ local function test_connect_formats()
 		end
 	end
 
+	-- verify that invalid input is properly rejected
+	local function try_error(desc, host, pattern)
+		local ok, err = pcall(httpc.get, host, "/", {})
+		total = total + 1
+		if ok then
+			print(string.format("  FAIL: %-35s expected error but succeeded", desc))
+			return false
+		end
+		err = tostring(err)
+		if err:find(pattern) then
+			passed = passed + 1
+			print(string.format("  PASS: %-35s rejected as expected", desc))
+			return true
+		else
+			print(string.format("  FAIL: %-35s wrong error=%s", desc, err))
+			return false
+		end
+	end
+
 	-- hostname variants
 	try_get("bare hostname",            "baidu.com")
 	try_get("hostname:port",            "baidu.com:80")
@@ -127,16 +146,34 @@ local function test_connect_formats()
 		print("  SKIP: https tests (no ltls module)")
 	end
 
-	-- ipv6 variants (parsing verification, connection may fail without ipv6 network)
+	-- bare ipv6 should be rejected (RFC 3986: brackets required)
-	try_parse("bare ipv6 ::1",            "::1")
+	try_error("bare ipv6 ::1 (reject)",    "::1",         "bare IPv6")
+	try_error("bare full ipv6 (reject)",   "2001:db8::1", "bare IPv6")
+
+	-- bracketed ipv6 (parsing verification, connection may fail without ipv6 network)
+	try_parse("[ipv6] no port",            "[::1]")
 	try_parse("[ipv6]:port",              "[::1]:8080")
-	try_parse("bare full ipv6",           "2001:db8::1")
+	try_parse("[full ipv6] no port",      "[2001:db8::1]")
 	try_parse("[full ipv6]:port",         "[2001:db8::1]:80")
+	try_parse("http://[ipv6]",            "http://[::1]")
 	try_parse("http://[ipv6]:port",       "http://[::1]:80")
 	if pcall(require, "ltls.c") then
 		try_parse("https://[ipv6]:port",   "https://[::1]:443")
 	end
 
+	-- ipv6 real connection test via AAAA DNS resolution
+	local ok_aaaa, ipv6 = pcall(dns.resolve, "ipv6.google.com", true)
+	if ok_aaaa and ipv6 then
+		local ipv6_host = string.format("[%s]", ipv6)
+		print(string.format("  resolved ipv6.google.com AAAA -> %s", ipv6))
+		try_parse("[ipv6] from AAAA",          ipv6_host)
+		try_parse("[ipv6]:80 from AAAA",       ipv6_host .. ":80")
+		try_parse("http://[ipv6] from AAAA",    "http://" .. ipv6_host)
+		try_parse("http://[ipv6]:80 from AAAA", "http://" .. ipv6_host .. ":80")
+	else
+		print("  SKIP: ipv6 AAAA tests (dns resolve failed for ipv6.google.com)")
+	end
+
 	-- k8s internal domain variants (parsing verification, connection will fail outside k8s)
 	-- domains ending with digits must not be mistaken for ipv4
 	try_parse("k8s svc short",            "my-svc:8080")