这个概念很简单。将任何连续的代码行（即没有 if、goto 或函数调用或标签）想象为图的一个节点。每个 goto 或函数调用都会创建从当前节点到 goto 标签所在节点或其正在调用的函数的定向链接。请记住，函数本身可以是一个图，而不是一个简单的节点，因为它内部可能有 if 或其他函数调用。每个函数调用还会创建从函数的叶节点（函数返回的位置）到包含函数调用后代码的节点的定向链接。（这可以创建很多从函数图传出的链接，因为可以在代码的许多部分调用该函数）

同样，如果您有一个 if，则您有两个来自当前节点的方向链接到 if 语句的 if 部分和 else 部分（除非您检测到 if(0) 或if(1) 就像你说的，在这种情况下只有一个链接到正确的位置）

图表的根是 main 的入口点。现在，要找到死代码，您必须做的就是从根位置简单地遍历图形（例如使用 DFS 或 BFS），并最终查看哪些节点未被访问。这向您显示了死代码，即代码中无论您的程序采取什么方向，都不会到达这些位置的位置。

如果您想自己实现，可以采用递归方法（类似于解析代码，但更简单）。例如，如果您看到 if，您可以说：

typedef char *line;
FlowGraph *get_flow_graph(line *code)
{
    FlowGraph *current_node = malloc(sizeof *current_node);
    current_node->flow_to = malloc(some_maximum * sizeof *current_node->flow_to);
    current_node->flow_to_count = 0;
    ...
    if (is_if_statement(code[0]))
    {
        FlowGraph *if_part = get_flow_graph(code + 1);
        FlowGraph *else_part = get_flow_graph(code + find_matching_else(code));
        current_node->flow_to[current_node->flow_to_count++] = if_part;
        current_node->flow_to[current_node->flow_to_count++] = else_part;
    }
    else
    ...
}

I can explain to you the concept of control flow graphs, but I am not familiar with the library itself.

The concept is simple. Imagine any sequential lines of code (that is without if, goto or function call or labels) as one node of a graph. Every goto or function call creates a directional link from the current node to the node where the goto label is or the function it is calling. Remember that a function itself could be a graph and not a simple node, because it may have ifs or other function calls inside. Each function call also creates a directional link from leaf nodes of the function (where the function returns) to the node containing the codes right after the function call. (That can create a lot of links outgoing from the function graph because the function can be called in many parts of the code)

Likewise, if you have an if, you have two direction links from the current node to both the if part and the else part of the if statement (unless you detect if(0) or if(1) like you said in which case there is only one link to the proper location)

The root of your graph is the entry point of main. Now what you must do to find dead code is to simply traverse the graph from the root position (using DFS or BFS for example) and in the end see which nodes were NOT visited. This shows you the dead codes, that is places in the code that no matter what direction your program takes, it won't reach those locations.

If you want to implement this yourself, you can take a recursive approach (similar to parsing the code but simpler). For example if you see an if you say:

typedef char *line;
FlowGraph *get_flow_graph(line *code)
{
    FlowGraph *current_node = malloc(sizeof *current_node);
    current_node->flow_to = malloc(some_maximum * sizeof *current_node->flow_to);
    current_node->flow_to_count = 0;
    ...
    if (is_if_statement(code[0]))
    {
        FlowGraph *if_part = get_flow_graph(code + 1);
        FlowGraph *else_part = get_flow_graph(code + find_matching_else(code));
        current_node->flow_to[current_node->flow_to_count++] = if_part;
        current_node->flow_to[current_node->flow_to_count++] = else_part;
    }
    else
    ...
}

回复收藏 0 原文