chunk is already sorted here. Could we just use chunks.indices().collect<Vec<_>>?

The share chunks ordering is not stable because we using unstale ordering at https://github.com/rolldown/rolldown/blob/main/crates/rolldown/src/stages/generate_stage/code_splitting.rs#L141

If we fix it look like it could be get expected stable ordering for assgin name

The share chunks ordering is not stable because we using unstale ordering at https://github.com/rolldown/rolldown/blob/main/crates/rolldown/src/stages/generate_stage/code_splitting.rs#L141

Shared chunks are sorted by their first module's execution order. It seems stable.

rolldown/crates/rolldown/src/stages/generate_stage/code_splitting.rs

Lines 169 to 216 in cc41c94

	chunks
	.iter_mut()
	.sorted_by(|a, b| {
	let a_should_be_first = Ordering::Less;
	let b_should_be_first = Ordering::Greater;
	match (&a.kind, &b.kind) {
	(
	ChunkKind::EntryPoint { module: a_module_id, .. },
	ChunkKind::EntryPoint { module: b_module_id, .. },
	) => self.link_output.module_table.modules[*a_module_id]
	.exec_order()
	.cmp(&self.link_output.module_table.modules[*b_module_id].exec_order()),
	(ChunkKind::EntryPoint { module: a_module_id, .. }, ChunkKind::Common) => {
	let a_module_exec_order =
	self.link_output.module_table.modules[*a_module_id].exec_order();
	let b_chunk_first_module_exec_order =
	self.link_output.module_table.modules[b.modules[0]].exec_order();
	if a_module_exec_order == b_chunk_first_module_exec_order {
	a_should_be_first
	} else {
	a_module_exec_order.cmp(&b_chunk_first_module_exec_order)
	}
	}
	(ChunkKind::Common, ChunkKind::EntryPoint { module: b_module_id, .. }) => {
	let b_module_exec_order =
	self.link_output.module_table.modules[*b_module_id].exec_order();
	let a_chunk_first_module_exec_order =
	self.link_output.module_table.modules[a.modules[0]].exec_order();
	if a_chunk_first_module_exec_order == b_module_exec_order {
	b_should_be_first
	} else {
	a_chunk_first_module_exec_order.cmp(&b_module_exec_order)
	}
	}
	(ChunkKind::Common, ChunkKind::Common) => {
	let a_chunk_first_module_exec_order =
	self.link_output.module_table.modules[a.modules[0]].exec_order();
	let b_chunk_first_module_exec_order =
	self.link_output.module_table.modules[b.modules[0]].exec_order();
	a_chunk_first_module_exec_order.cmp(&b_chunk_first_module_exec_order)
	}
	}
	})
	.enumerate()
	.for_each(|(i, chunk)| {
	chunk.exec_order = i.try_into().expect("Too many chunks, u32 overflowed.");
	});

chunk is already sorted here. Could we just use chunks.indices().collect<Vec<_>>?

Oh. I missing the your meanings. Here the order is not correct, because the share chunk execute order could be at first. But the entry chunks should be at first for assign name.

Probably should this logic.

If we fix it look like it could be get expected stable ordering for assgin name

Here I have a try to changed it, but looke like the sorted_modules hasn't dynamic import ids.