Null-check item signs tag in TagCache

[auvq]

Fixes #6261.

TagCache.loadPacket NPEs if the server doesn't send the signs item tag. Added a null check like the block branch right above.

[Novampr]

Approving on the basis that this works, but to note that later we'll probably use ViaVersion packet payload in order to properly determine the server version and what to emulate.

[onebeastchris]

Have you tested this? This shouldn't ever be null as it's part of standard tags - if it is, I'm curious if the server then follows up by a proper tags packet

[auvq]

This one specifically I haven't, the stacktrace showed it was reachable. Agreed signs shouldn't normally be null on vanilla, I assume viaversion-wrapped or modded setups where the first tags packet doesn't include it.

I'm almost certain this fixes it

[onebeastchris]

I agree that it'll fix the NPE, but if it doesn't fix the actual sign detection, that should be accounted for

Signs were, iirc, always sent by Via; did that change?

[auvq]

Good question on Via - I don't know if something changed there. If Via has been consistently sending signs, then null here isn't the "normal pre-1.13" case I was assuming.

Without a proper alternate signal (the ViaVersion payload work mentioned earlier), I don't have a clean way to detect what emulatePost1_13Logic should actually be when signs is missing. I can either preserve prior state and stop overriding, log a warning and keep current behavior, or close this and defer to the proper version detection work.

My main goal is to stop the crash, but I wouldn't mind working on it further.

Whichever works for you.

[auvq]

Actually, I looked into it. Via started sending empty tags packets recently (march-ish) as a placeholder during config stage that gets filled in later. That would explain the null case here.

If that's right, probably makes sense to only update emulatePost1_13Logic when signs is actually present, so a later real tags packet handles detection naturally. Or close this and defer, up to you.

[onebeastchris]

Are you able to test whether the signs tag is sent / followed up by via? If that really is followed up, lgtm

[auvq]

Tested it on vanilla Paper 1.21.11 with the PR applied. Signs tag is sent in the initial UPDATE_TAGS, no NPE, signs work normally on Bedrock. The null branch doesn't fire in the vanilla flow since signs is always present, so for vanilla it's just defensive.

Don't have a Via setup to verify the follow-up behavior I mentioned earlier, so can't confirm that side. But the fix itself is confirmed safe on vanilla at minimum.