介绍一款调试Python的神器

MAN

前段时间在Pi上用Python和gtk搞了个应用，到最后测试阶段一直被闪退折磨。闪退是我自己起的名词，比如你玩手机的时候，玩得正高兴的时候，游戏莫名其妙的闪退了，你郁闷不，开发者其实更郁闷，因为闪退嘛，死都不知道怎么死的。
经过一周的追踪，用尽各种手段，终于让我发现一款神奇。先来说说事情的起因。
先省略几个小时的log。。。

1. 
   
2. batching times: 1
   
3. thread batching starts...
   
4. 
   
5. set,  ['?RD.EC', '9006.38']
   
6. thread batching stopped.
   
7. 
   
8. ('READDATA', ('COND', None, None), [('?RD.EC', None)])
   
9. get ?RD.EC
   
10. in {'?RD.EC': '9006.38'}
    
11. batching times: 1
    
12. thread batching starts...
    
13. 
    
14. set,  ['OHM:5']
    
15. thread batching stopped.
    
16. 
    
17. *** glibc detected *** python: double free or corruption (fasttop): 0x01b73a28 ***
    
18. 已放弃
    
19. pi@raspberrypi ~/calibration/cali04a $

复制代码

看到了吧，重复了几个小时的代码，突然间闪退了，就给你这么点答案，释放空指针，靠，这前不着村后不着地的python，你让我去给你查指针释放去？经人介绍，用pydb(Extended Python Debugger)来跑，无功而返，对于这种Segment fault，根本跟没用差不多嘛，pydb也就能调调正常的程序。

1. ('READDATA', ('COND', None, None), [('?RD.EC', None)])
   
2. get ?RD.EC
   
3. in {'?RD.EC': '225.04'}
   
4. batching times: 1
   
5. thread batching starts...
   
6. 
   
7. set,  ['OHM:1000']
   
8. thread batching stopped.
   
9. 
   
10. batching times: 1
    
11. thread batching starts...
    
12. 
    
13. set,  ['?RD.EC', '450.25']
    
14. thread batching stopped.
    
15. 
    
16. ('READDATA', ('COND', None, None), [('?RD.EC', None)])
    
17. get ?RD.EC
    
18. in {'?RD.EC': '450.25'}
    
19. *** glibc detected *** /usr/bin/python: double free or corruption (!prev)batching times: 1
    
20. : 0x00c12800 ***
    
21. 已放弃
    
22. pi@raspberrypi ~/calibration/src_cali04a $

复制代码

好吧，逼急了上gdb吧，好歹知道是哪里的错误了。这里要吐槽下gtk用到的gdk，满篇的断言，真不知该夸你还是该扁你！

1. ('READDATA', ('COND', None, None), [('?RD.EC', None)])
   
2. get ?RD.EC
   
3. in {'?RD.EC': '22.50'}
   
4. batching times: 1
   
5. thread batching starts...
   
6. 
   
7. set,  ['OHM:10000']
   
8. thread batching stopped.
   
9. 
   
10. batching times: 1
    
11. thread batching starts...
    
12. 
    
13. set,  ['?RD.EC', '45.01']
    
14. thread batching stopped.
    
15. 
    
16. ('READDATA', ('COND', None, None), [('?RD.EC', None)])
    
17. get ?RD.EC
    
18. in {'?RD.EC': '45.01'}
    
19. batching times: 1
    
20. thread batching starts...
    
21. 
    
22. **
    
23. Gdk:ERROR:/build/gtk+2.0-x7uJoT/gtk+2.0-2.24.10/gdk/gdkregion-generic.c:1206:miUnionO: assertion failed: (pReg->numRects<=pReg->size)
    
24. bt
    
25. 
    
26. 
    
27. 
    
28. 
    
29. 
    
30. [New Thread 0x47eff470 (LWP 15726)]
    
31. [Thread 0x48eff470 (LWP 15725) exited]
    
32. [New Thread 0x48eff470 (LWP 15727)]
    
33. [Thread 0x47eff470 (LWP 15726) exited]
    
34. [New Thread 0x47eff470 (LWP 15728)]
    
35. [Thread 0x47eff470 (LWP 15728) exited]
    
36. [New Thread 0x467bf470 (LWP 15729)]
    
37. [Thread 0x48eff470 (LWP 15727) exited]
    
38. [New Thread 0x48eff470 (LWP 15730)]
    
39. 
    
40. Program received signal SIGABRT, Aborted.
    
41. 0x40384bfc in raise () from /lib/arm-linux-gnueabihf/libc.so.6
    
42. (gdb) bt
    
43. #0  0x40384bfc in raise () from /lib/arm-linux-gnueabihf/libc.so.6
    
44. #1  0x4038897c in abort () from /lib/arm-linux-gnueabihf/libc.so.6
    
45. #2  0x4046bd2c in ?? () from /lib/arm-linux-gnueabihf/libc.so.6
    
46. #3  0x4046bd2c in ?? () from /lib/arm-linux-gnueabihf/libc.so.6
    
47. Backtrace stopped: previous frame identical to this frame (corrupt stack?)
    
48. (gdb)

复制代码

上面报gdk error的地方已经被无数函数调用过，我用bt命令回溯stack，貌似也看不到啥有用的信息，谁让咱用的是多线程呢，好吧，接下来该大杀器出场了。

1. set,  ['?RD.EC', '225.06']
   
2. thread batching stopped.
   
3. 
   
4. ('READDATA', ('COND', None, None), [('?RD.EC', None)])
   
5. get ?RD.EC
   
6. in {'?RD.EC': '225.06', '?NTC.T': 'None', 'ID': 'None'}
   
7. batching times: 1
   
8. thread batching starts...
   
9. 
   
10. set,  ['OHM:1000']
    
11. thread batching stopped.
    
12. 
    
13. **
    
14. Gdk:ERROR:/build/gtk+2.0-x7uJoT/gtk+2.0-2.24.10/gdk/gdkregion-generic.c:1110:miUnionNonO: assertion failed: (y1 < y2)
    
15. Fatal Python error: Segmentation fault
    
16. 
    
17. Thread 0x474a6470:
    
18. 
    
19. Thread 0x486b9470:
    
20.   File "/usr/lib/python2.7/threading.py", line 244 in wait
    
21.   File "/usr/lib/python2.7/threading.py", line 404 in wait
    
22.   File "/usr/lib/python2.7/threading.py", line 500 in start
    
23.   File "d:/workspace/calibration/cali_test.py", line 43 in timer_event
    
24.   File "/usr/lib/python2.7/threading.py", line 760 in run
    
25.   File "/usr/lib/python2.7/threading.py", line 552 in __bootstrap_inner
    
26.   File "/usr/lib/python2.7/threading.py", line 525 in __bootstrap
    
27. 
    
28. Thread 0x4666f470:
    
29.   File "d:/workspace/calibration/basinterp.py", line 476 in run
    
30.   File "d:/workspace/calibration/cali_test.py", line 388 in plying
    
31.   File "/usr/lib/python2.7/threading.py", line 505 in run
    
32.   File "/usr/lib/python2.7/threading.py", line 552 in __bootstrap_inner
    
33.   File "/usr/lib/python2.7/threading.py", line 525 in __bootstrap
    
34. 
    
35. Thread 0x45de1470:
    
36.   File "/usr/lib/python2.7/threading.py", line 244 in wait
    
37.   File "/usr/lib/python2.7/threading.py", line 668 in join
    
38.   File "d:/workspace/calibration/cali_test.py", line 255 in receiving
    
39.   File "/usr/lib/python2.7/threading.py", line 505 in run
    
40.   File "/usr/lib/python2.7/threading.py", line 552 in __bootstrap_inner
    
41.   File "/usr/lib/python2.7/threading.py", line 525 in __bootstrap
    
42. 
    
43. Current thread 0x400d7210:
    
44.   File "d:/workspace/calibration/cali_test.py", line 127 in on_TextViewOfLog_size_allocate
    
45.   File "d:/workspace/calibration/cali_test.py", line 786 in main
    
46.   File "d:/workspace/calibration/cali_test.py", line 791 in <module>
    
47. 段错误
    
48. pi@raspberrypi ~/calibration/cali04a $
    

复制代码

看到这里爽伐？每个线程的stack都被完整的打印出来，找问题不跟玩似的，呵呵。问题就是出在最后这个on_TextViewOfLog_size_allocate，究其原因，我没有把gtk的窗口设成固定大小，又因为我有一个Frame在不停地变换长短不一的文字，所以整个窗体一会大一点一会小一点，gdk经过几个小时的折磨，终于不小心被自己的断言给干掉了gdkregion-generic.c:1110:miUnionNonO: assertion failed: (y1 < y2)

好了，就这样。对了，忘了说神器叫啥了，faulthandler，截至发帖版本是2.2，地址在：https://pypi.python.org/pypi/faulthandler，用法如下：
在你的python主程序启动前，加上如下代码，然后就等着兔子上门吧：

1. import faulthandler
   
2. faulthandler.enable()

复制代码

有什么问题欢迎加入QQ群讨论，号码：204148284（开源硬件）

ayablue

就是他了  找了好久