mod_cloud_notify_encrypted: Ensure body substring remains valid UTF-8 The `body:sub()` call risks splitting the string in the middle of a multi-byte UTF-8 sequence. This should have been caught by util.stanza validation, but that would have caused some havoc, at the very least causing the notification to not be sent. There have been no reports of this happening. Likely because this module isn't widely deployed among users with languages that use many longer UTF-8 sequences. The util.encodings.utf8.valid() function is O(n) where only the last sequence really needs to be checked, but it's in C and expected to be fast.

author: Kim Alvefur <zash@zash.se>
date: Sun, 22 Aug 2021 13:22:59 +0200
parent: 4650:44af84178cea
child: 5054:62480053c87b
--- a/mod_cloud_notify_encrypted/mod_cloud_notify_encrypted.lua	Sun Aug 22 10:46:31 2021 +0100
+++ b/mod_cloud_notify_encrypted/mod_cloud_notify_encrypted.lua	Sun Aug 22 13:22:59 2021 +0200
@@ -1,5 +1,6 @@
 local array = require "util.array";
 local base64 = require "util.encodings".base64;
+local valid_utf8 = require "util.encodings".utf8.valid;
 local ciphers = require "openssl.cipher";
 local jid = require "util.jid";
 local json = require "util.json";
@@ -78,6 +79,9 @@
 		body = original_stanza:get_child_text("body");
 		if body and #body > 255 then
 			body = body:sub(1, 255);
+			if not valid_utf8(body) then
+				body = body:gsub("[\194-\244][\128-\191]*$", "");
+			end
 		end
 	end