Time  Nick      Message
12:30 kilbith   https://www.phoronix.com/scan.php?page=news_item&px=Zstd-1.5.1-Released
12:32 MTDiscord <Sublayer plank> so it's smaller, faster and all this in a maintenance release? very cool
12:58 MTDiscord <luatic> What's the matter with the claim that "string.char" is "unsafe" introduced by the PNG writer PR: https://github.com/minetest/minetest/blob/master/doc/lua_api.txt#L4703 ? How is this substantiated?
13:01 MTDiscord <luatic> Is it because no "locale" can be guaranteed? Why doesn't builtin just call os.setlocale then to settle this? (arguably this might break mods which rely on a specific locale, but introducing insecurity concerning the reliability of string.char and string.byte seems unacceptable to me)
13:06 MTDiscord <luatic> The Lua 5.1 implementation of string.char at least doesn't show anything I consider locale-specific (grep for str_char in the minetest dir to have a look)
13:08 MTDiscord <luatic> The LuaJIT codebase is similar, just grep for string_char there. Just a cast to char.
13:54 sfan5     @luatic https://github.com/minetest/minetest/pull/11485#discussion_r676166902
13:54 sfan5     "Why doesn't builtin just call os.setlocale then" <- gettext also uses the set locale, I think this should answer your question
13:55 sfan5     tochigi: trying with gdb/lldb and checking the stacktraces would probably help determine where it's hanging
16:01 MTDiscord <luatic> sfan5: wouldn't it be possible to just set the locale for Lua though?
16:01 sfan5     no
16:01 MTDiscord <luatic> or to override string.char and string.byte to provide platform-independent behavior?
16:02 sfan5     if you say the implementations are safe why would we need that
16:02 MTDiscord <luatic> well, according to hecktest's comment I was wrong
17:53 tochigi   sfan5: i am too lazy
17:54 sfan5     ¯\_(ツ)_/¯