Substitute MT_NULL with MT_RAY in Lua

As of !1973 (merged), P_SpawnMobj no longer replaces MT_NULL with MT_RAY, and instead aborts spawning MT_NULL mobj types. However, some Lua scripts, including [Silver the Hedgehog], expect MT_NULL to be spawnable.
This merge request replaces MT_NULL with MT_RAY in Lua's mobj-spawning functions, with a // TODO: 2.3: comment for making it show a Lua error instead for v2.3 (because usually, trying to spawn MT_NULL means that something has gone wrong, that you don't want to spawn anything and so don't need to call a spawning function, or that you should be using MT_RAY instead).
(SOC actions trying to spawn MT_NULL will still not spawn anything, though.)

TL;DR: Reverted Lua mobj-spawning to v2.2.13's behaviour. Not worth mentioning in v2.2.14's changelog.

mentioned in merge request !1973 (merged)

mentioned in merge request !2267 (merged)

Maybe raise a Lua warning here instead? MT_NULL really shouldn't be spawned, so it's important that we raise awareness to the Lua developer about that.

Is it worth it for semantics' sake only? The behaviour (automatically substituting MT_NULL with MT_RAY) is the same as in v2.2.13, and updating the script to use MT_RAY doesn't cause any in-game difference. That would cause harmless-but-annoying warnings to pop up for end users, "requiring" a mod to be updated just for semantics.

(For reference, this CONS_Debug line is taken verbatim from pre-​!1973 (merged) P_SpawnMobj.)

It's not really semantics, though. From a comment you added right below this:

TODO: 2.3: Use the below NOSPAWNNULL define instead. P_SpawnMobj used to say "if MT_NULL, use MT_RAY instead", so the above define maintains Lua script compatibility until v2.3

That, to me, sounds like MT_NULL is now obsolete and shouldn't be used. We already warn in cases where scripts use P_TeleportMove in favor of P_*Origin, so why not do that here too? If we don't warn, script writers won't be able to prepare for a breaking change we've planned for in 2.3, which can lead to unexpected breakage to some script writers once 2.3 is released.

I agree. Trying to spawn an invalid object is very distinctly a script error, and it should be treated as such.

I reported the Silver crash because Lua causing crashes is always a flaw. It ended up just being a genuine regression, but if it was due to bad Lua, then that should've also been changed to error.

changed this line in version 2 of the diff

@ Hanicef - [...] We already warn in cases where scripts use P_TeleportMove in favor of P_*Origin, so why not do that here too? [...]

Funnily enough, I considered bringing that up in my previous comment. There's a difference between changing P_TeleportMove to P_MoveOrigin and changing it to P_SnapOrigin (i.e. it's not just semantic), and SRB2 can't automatically tell what the Lua scripter originally intended.

I'd be happy to add a Lua warning here too, but I'd like to hear STJr's opinion first - and with LJ Sonic giving a thumbs-up and Sal commenting, I've heard their opinion now.

added Changed behavior Lua labels

added 1 commit

• bdeec562 - Substitute MT_NULL with MT_RAY in Lua

Compare with previous version

added Regression label

changed milestone to %2.2.14

added 294 commits

• bdeec562...d4053057 - 293 commits from branch STJr:next
• 7f68a727 - Merge branch SRB2:next into substitute-null-with-ray

Compare with previous version

merged

mentioned in commit add091ba