String buffer overflow on ping for long Forge modlists

[chloeruka]

itzg/mc-monitor#21 has some of the context for the error.

I have a Forge modded server running Enigmatica 6 on MC 1.16.5, but it breaks on ping as FML2 server handshakes can be unpredictably long since they also ship the entire modlist for the client compatibility check. The exact error I'm speaking of originates from here:
https://github.com/Raqbit/mc-pinger/blob/9baeff22a28b666d7f564d760331c9a5df7ed61d/encoding/types.go#L135-L137

I see there's a draft to replace some of this code with a new code-gen library described in #9. I have a workaround so I'm not that bothered by this if you don't want to tackle Forge compatibility, but if you and @itzg did have an appetite to resolve this and want a hand I can try to find time to contribute if you can indicate what needs doing. It may also be worth considering if you can join forces somehow with https://github.com/xrjr/mcutils as that ping implementation does work for me in this narrow use-case (see itzg/mc-monitor#22).

Reproducing

I have some logs and a hex dump for my initial reproduction of the issue.

If I get the time this weekend I'll try to pop a more reproducable test case into the discussion.

[raqbit]

Thank you for the report! Funnily enough I was recently looking into FML handshakes myself but I had not realized they actually increase the length that a protocol string can be (Following wiki.vg). (Though it doesn't surprise me as they are sending the mod list as binary data in JSON, using a custom format optimized for UTF-16 codepoint count)

I indeed started some work on redoing this code based on a separate library with code-gen in #9 some time ago but that still requires some more work, so I think simply updating the check & adding a test for these big query responses should be the way to go for now. It'd be great if you could get a reproducible test case, a simple (redacted) json dump of the query response would probably suffice.