fit bitmask overflow initial dirty value

[tanhauhau]

Fixes #4263

the initial dirty value needs to be depend on whether bitmask context overflow

[Conduitry]

Well GitHub apparently also attributed this one to me 🙁

[Conduitry]

I've reopened the issue because, disconcertingly, the tests pass if you run them with npm test but bitmask-overflow-if fails with PUBLISH=true npm test which tests against the actual bundle that would be published. 😱 I have no idea.

[Conduitry]

The compiler has different output in this test depending on whether it was build with PUBLISH=true. Without it, it generates current_block_type_index$ = select_block_type$(ctx, [-1]); - and in publish mode, it generates current_block_type_index$ = select_block_type$(ctx, -1);.

In the production/PUBLISH build, the method is compiled from

get_initial_dirty_bit() {	const _this = this;	// TODO: context-overflow make it less gross	const val = x`-1`;	return {	...val,	elements: [val],	get type() {	return _this.renderer.context_overflow ? 'ArrayExpression' : 'UnaryExpression';	},	}; } 

to

get_initial_dirty_bit() {	const _this = this;	// TODO: context-overflow make it less gross	const val = x `-1`;	return Object.assign({}, val, { elements: [val], get type() {	return _this.renderer.context_overflow ? 'ArrayExpression' : 'UnaryExpression';	} }); } 

i.e., the spread is compiled away.

According to some tests I just did, apparently Object.assign() will aggressively evaluate the getter, while the spread lets it remain lazy. Holy shit.

[tanhauhau]

oh my.. we should really look into fixing those lazy code...
anyway, i have a patch for it temporarily #4508