转换GDB调用栈到流程图
来源:互联网 发布:学数据库安全 编辑:程序博客网 时间:2024/05/16 06:42
如果你想在GDB调试时把调用堆栈保存下来归档,那下面这个脚本就方便你了。原理是将调用堆栈的函数抽取出来,再完成调用关系就可以了。
首先你要安装dot (Mac OS下安装Graphviz), 如果你想转为文本格式,就可安装Perl的Graph::Easy包(命令行:sudo perl -MCPAN -e 'install Graph::Easy', Ubuntu下直接安装libgraph-easy-perl)。
然后按需要执行脚本就可以了, 假定如下的调用栈:
#0 WebCore::FrameLoader::FunctionA (this=0x2a7d91f8) at /FrameLoader.cpp
#1 0x4efd2514 in WebCore::FrameLoader::FunctionB (this=0x2a7d91f8) at /FrameLoader.cpp:553
#2 0x4efd1918 in FunctionC ()at /mainFile.cpp:100
1. 转为图片
python convertStackToDot.pystack.txt|dot -Tpng>output.png
默认由上到下排列,如果你想改变,可以通过在脚本的参数中增加dot设定调整,比如:
python convertStackToDot.py stack.txt 'rankdir=LR;'|dot -Tpng>output.png
就会变成横排:
2. 转为文本
有时你希望转为文本,还好有Graph Easy包,不然你就要使用asciio自己画了。
python convertStackToDot.py stack.txt ‘rankdir=LR;’|graph-easy -as_ascii>output.txt
效果如下:
+-----------+ +------------------------+ +------------------------+
| FunctionC | --> | FrameLoader::FunctionB | --> | FrameLoader::FunctionA |
+-----------+ +------------------------+ +------------------------+
3. 多个调用栈的解析
如果有多个调用栈一起组成完整的流程,只要在各个调用栈中间加入空格就可以了。比如下面的堆栈:
#0 WebCore::FrameLoader::FunctionF (this=0x2a7d90f8,at /FrameLoader.cpp
#1 WebCore::FrameLoader::FunctionA (this=0x2a7d90f8,at /FrameLoader.cpp
#2 0x4ffd2514 in WebCore::FrameLoader::FunctionB (this=0x2a7d90f8, at /FrameLoader.cpp:553
#3 0x4ffd1918 in FunctionC (this=0x2a7d90f8 at /mainFile.cpp:100
#0 WebCore::FrameLoader::FunctionE (this=0x2a7d90f8,at /FrameLoader.cpp
#1 0x4ffd2514 in WebCore::FrameLoader::FunctionB (this=0x2a7d90f8, at /FrameLoader.cpp:553
#2 0x4ffd1918 in FunctionC (this=0x2a7d90f8 at /mainFile.cpp:100
输出结果如下, 线段上的数字代码的堆栈序号:
转载请注明出处: http://blog.csdn.net/horkychen
再次修改了一下,对于重复的调用,可以使用-d选项选择是否全部显示出来,同时可以允许指定一个正则表达式将某部分节点高亮显示,如允许显示重复调用路径的效果:
python convertStackToDot.py -d -l "A|PolicyChecker" stack.txt|dot -Tpng>output.png
脚本如下:
#!/usr/bin/python#coding=utf-8#python convertStackToDot.py stack.txt|dot -Tpng>output.png#To generate ascii flow chart, graph_easy should be installed:# sudo apt-get install libgraph-easy-perl#The use below command line:#python convertStackToDot.py stack.txt|graph-easy -as_ascii>output.txtimport sysimport reimport argparseimport osfunction_name_str = r"[a-zA-Z][a-zA-Z0-9]*::[~a-zA-Z0-9\_\-<: >=]*\(|[a-zA-Z0-9_\-<>]* \("class_name_str = r"::[a-zA-Z][a-zA-Z0-9]*::"known_namespaces = []libCName="/system/lib/libc.so"Colors=["#000000","#ff0000","#00aa00","#0000ff","#800080","#daa520","#ff00b4","#d2691e","#00bfff", "#D0A020","#006000","#305050","#101870"]Styles=["solid","dashed","dotted"]MAX_COLOR_COUNT = len(Colors)MAX_LINE_WIDTH_COUNT = 4MAX_STYLE_COUNT = len(Styles)FirstNodeAttr = ' style="rounded,filled" fillcolor=\"#00bb0050\"'HighlightAttr = ' style="rounded,filled" fillcolor=\"yellow\"'blockNum = 0 nodeNo = 0nodeList={}nodeOrderList={}firstNodeList={}nodeAttr={}outputText = ''callingStack = ''newBlock=TruewillCommit=False #For filtering purposeblockBackTrace = ''blockNodeList={}def getTextOfBlockNodeList(lastNodeName,lastNodeLabel): global firstNodeList strBlockNodeText = '' for key in blockNodeList.keys(): if not nodeList.has_key(key): name = blockNodeList[key] strBlockNodeText = strBlockNodeText + name + nodeAttr[name]+'\n' nodeList[key] = name #Replace the attribute of the last node if len(lastNodeName)>0 and not firstNodeList.has_key(lastNodeName): oldStr = lastNodeName+'[label="'+lastNodeLabel+'" shape=box ];'; newStr = lastNodeName+'[label="'+lastNodeLabel+'" shape=box '+FirstNodeAttr+' ];' strBlockNodeText = strBlockNodeText.replace(oldStr,newStr,1) firstNodeList[lastNodeName] = True return strBlockNodeTextdef submitAndResetForNewBlock(args,lastNodeName,lastNodeLabel): global blockBackTrace,newBlock,callingStack global blockNodeList,willCommit,outputText newBlock = True if willCommit and len(blockBackTrace)>0: callingStack = blockBackTrace + '\n' + callingStack blockNodeText = getTextOfBlockNodeList(lastNodeName,lastNodeLabel) outputText = outputText+blockNodeText blockNodeList = {} blockBackTrace = '' willCommit = (len(args.filter)==0)def getClassName(text): m = re.search(class_name_str,text) if m: className=text[0:m.end()-2] elif not text[:text.find('::')] in known_namespaces: className = text[:text.find('::')] else: className = text return classNamedef getNodeName(text,nodeNo,args): global willCommit,blockNodeList,newBlock processText = text if len(args.ignore)>0 and re.search(args.ignore,text): return '' if args.onlyClass: processText = getClassName(text) if nodeList.has_key(processText): nodeName = nodeList[processText] elif blockNodeList.has_key(processText): nodeName = blockNodeList[processText] else: nodeName = 'Node'+str(nodeNo) blockNodeList[processText]=nodeName extraAttr = '' try: if len(args.highlight)>0 and re.search(args.highlight,processText): extraAttr = HighlightAttr except: extraAttr = '' nodeAttr[nodeName] = '[label="'+processText+'" shape=box '+extraAttr+'];' if len(args.filter)>0 and re.search(args.filter,text): willCommit = True return nodeNamedef createNewRelation(nodeName,lastNodeName,blockNum): global blockBackTrace tempKey = "%s_%s"%(nodeName,lastNodeName) if args.duplicate or not nodeOrderList.has_key(tempKey): lineColor = Colors[(blockNum-1)%MAX_COLOR_COUNT] linePenWidth = str((int((blockNum-1)/MAX_COLOR_COUNT)%MAX_LINE_WIDTH_COUNT)+1) lineStyle = Styles[((blockNum-1)/(MAX_COLOR_COUNT*MAX_LINE_WIDTH_COUNT))%MAX_STYLE_COUNT] if nodeOrderList.has_key(tempKey): linePenWidth = '1' lineColor = lineColor+'50' #Set alpha value blockBackTrace = nodeName+'->'+lastNodeName+'[label='+str(blockNum)+\ ',color=\"'+lineColor+'\"'+\ ',style=\"'+lineStyle+'\"'+\ ',penwidth='+linePenWidth+']\n'+ \ blockBackTrace nodeOrderList[tempKey] = Truedef combineOutputText(): global outputText,callingStack if len(callingStack)>0: outputText = outputText+callingStack+"\n}" return outputText else: return ''def initialize(args): global outputText,callingStack outputText = "digraph backtrace{ \nnode [style=rounded fontname=\"Helvetica Bold\"];\n" + args.extraDotOptions +"\n"def convertToDot(file,args): global willCommit,outputText,newBlock,blockNum,nodeNo global outputText,callingStack,blockBackTrace lastNodeName = '' lastNodeLabel = '' willCommit = (len(args.filter)==0) #To specify the initial value according to the filter. f = open(file, 'r') for line in f: line = line.strip() if(len(line)==0) or line.startswith("#0 ") or line.startswith("#00 "): if not newBlock: #Start with new block here. submitAndResetForNewBlock(args, lastNodeName, lastNodeLabel) if(len(line.strip())==0): continue if not line.startswith("#"): continue text = "" m = re.search(function_name_str, line) if m: nodeNo = nodeNo+1 text=m.group(0).strip() text = text[:-1] text = text.strip() elif line.find(libCName)>0: nodeNo = nodeNo+1 text='FunctionInLibC' if(len(text)==0): continue #Get the existing node or create new one. Anyway, just ask for the name. nodeName = getNodeName(text,nodeNo,args) #To throw it away if no valid name was returned according to your arguments. if(len(nodeName)==0): continue if newBlock: newBlock = False blockNum = blockNum + 1 else: createNewRelation(nodeName,lastNodeName,blockNum) lastNodeName = nodeName lastNodeLabel = text if args.onlyClass: lastNodeLabel = getClassName(text) if len(blockBackTrace)>0: #Wow, one block was built successfully, sumit it. submitAndResetForNewBlock(args, lastNodeName, lastNodeLabel) f.close()if __name__=="__main__": parser = argparse.ArgumentParser() parser.add_argument('file', type=str, help='The text file which contains GDB call stacks.') parser.add_argument('-e','--extraDotOptions', default='', help='Extra graph options. For example: rankdir=LR; That means to show functions in horizontal.') parser.add_argument('-l','--highlight', default='', help='Regular Expression Pattern. Nodes are highlighted whose name match the pattern.') parser.add_argument('-f','--filter', default='', help='Regular Expression Pattern. The calling stack are shown only if which include the matched nodes.') parser.add_argument('-d','--duplicate', action='store_true', default=False, help='Leave duplicated callings.') parser.add_argument('-i','--ignore', default='', help='To hide some nodes, try this.') parser.add_argument('-c','--onlyClass', action='store_true', default=False, help='To simplify the output with less nodes, only Class node will be listed.') if len(sys.argv)<=1: parser.print_help() print " Any comment, please feel free to contact horky.chen@gmail.com." quit() args = parser.parse_args() if args.file is None: quit() initialize(args) if os.path.isfile(args.file): convertToDot(args.file,args) else: filenames = os.listdir(args.file) for filename in filenames: convertToDot(os.path.join(args.file,filename),args) resultDotString = combineOutputText() if len(resultDotString)>0: print(resultDotString)
参考:
1. GDB扩展之Command File - 提高调试效率
2. [WebKit]C++类的数据结构及在反汇编上的应用
3. 使用LLDB脚本简化打印复杂数据的操作
- 转换GDB调用栈到流程图
- gdb 调用栈
- gdb 调用栈 (call stack)
- gdb查看函数调用栈
- 算数类型转换流程图。
- 异步调用流程图
- ioctl调用流程图
- 用gdb获得正确的调用栈
- 异步系统接口调用流程图
- OkHttp拆解之调用流程图
- audioTrack调用的简单流程图
- gdb调用笔记
- 流程图一:Setting中用到的的流程图
- Openssl command到API转换--password和key、iv转化以及openssl进行gdb调试
- GDB attach到进程
- 一起talk GDB吧(第四回:GDB调用栈调试)
- gdb调试(四)函数调用栈—stack
- gdb调试(四)函数调用栈之--frame
- 我要学习Androiod应用开发
- 待修改,树的深度搜索和广度搜索
- UVa 11800 Determine the Shape
- 大规模并发
- Makefile 中:= ?= += =的区别
- 转换GDB调用栈到流程图
- 安装vim-full出错,提示unable to locate package vim-full
- windows无法完成格式化U盘与U盘修复解决方法
- 随手整理3
- HL7协议报文的解析脚本
- chapter05—Alien Security(POJ 1130)
- 大型电子商务网站架构之-前端优化
- 解析SVD
- Makefile 语法分析