seal repro

documenting a bug i've experienced using seal compile on a darklua-bundled project

this doesn't affect me that much, but i do have a wip project i'm working on slowly that is affected by this

issue

• i require something (it might not be std-specific, i haven't looked into it that far) in two different files
• when using seal run or seal build/init.luau it works fine, with both fs a and b being correct
• however, running the executable directly (via seal scripts/build.luau and repro.exe) results in a being nil while b still exists
• it should output fs a: table: <blablabla> and fs b: table: <blablabla>
• instead, it outputs fs a: nil and fs b: table: <blablabla>

versions

• seal 0.0.6 and 0.0.7, both with the same results
• darklua 0.17.3
• rokit 1.2.0

probably unrelated stuff

• os: windows 10 home 22H2 (19045.6466, i think?)
• i've tried with both powershell 7 and command prompt, both with the same results

build/init.luau, if you're feeling lazy like me

added this for convenience

local a = { cache = {} :: any }
do
	do
		local function __modImpl()
			local b = require("@std/fs")

			print(`fs b: {b}`)

			return nil
		end
		function a.a(): typeof(__modImpl())
			local b = a.cache.a
			if not b then
				b = { c = __modImpl() }
				a.cache.a = b
			end
			return b.c
		end
	end
end
local b = require("@std/fs")

print(`fs a: {b}`)
a.a()