关于 Python 递归时修改全局变量失败的问题

def dfs_maze(path, target):
global paths
maze = [[" ","#","#"," "," "],[" "," "," ","#"," "],["#"," "," ","#"," "],[" "," "," ","#","#"],[" "," "," "," "," "]]

if path[-1] == target:
paths.append(path)
print(path)
return
for each in [[0,1], [0,-1], [1, 0], [-1, 0]]:
node = [path[-1][0] + each[0], path[-1][1] + each[1]]
if -1 not in node and 5 not in node and node not in path and maze[node[0]][node[1]] != '#':
path.append(node)
dfs_maze(path, target)
path.pop(-1)

def main():
dfs_maze([[0,0]], [4, 4])

for each in paths:
print(each)
paths = []
main()

就是一个深搜解决迷宫最短路径的函数，path 是一个二维数组，储存已经走过的每一个点，target 是终点。在边界条件满足时，打印出来的路径是没问题的。但在深搜结束后，打印出来的 paths 是一堆[[0,0]]的二维数组，这是为何?
不知为啥没有代码排版，sorry...