OLD | NEW |
---|---|
(Empty) | |
1 #!/usr/bin/env python | |
M-A Ruel
2012/07/19 15:21:15
Remove the shebang on files not executable.
Dai Mikurube (NOT FULLTIME)
2012/07/20 04:32:39
Done.
| |
2 | |
3 import bisect | |
4 import os | |
5 import re | |
6 import sys | |
7 | |
8 class ParsingException(Exception): | |
9 def __init__(self, value): | |
M-A Ruel
2012/07/19 15:21:15
The __init__ is not strictly necessary. You can us
Dai Mikurube (NOT FULLTIME)
2012/07/20 04:32:39
Thanks. Done it.
| |
10 self.value = value | |
11 def __str__(self): | |
12 return repr(self.value) | |
13 | |
14 | |
15 class ProcedureBoundary(object): | |
16 def __init__(self, start, end, name): | |
17 self.start = start | |
18 self.end = end | |
19 self.name = name | |
20 | |
21 | |
22 class ProcedureBoundaryTable(object): | |
23 def __init__(self): | |
24 self.sorted_value_list = [] | |
25 self.dictionary = {} | |
26 self.sorted = True | |
27 | |
28 def append(self, entry): | |
29 if self.sorted_value_list: | |
30 if self.sorted_value_list[-1] > entry.start: | |
31 self.sorted = False | |
32 elif self.sorted_value_list[-1] == entry.start: | |
33 return | |
34 self.sorted_value_list.append(entry.start) | |
35 self.dictionary[entry.start] = entry | |
36 | |
37 def find_procedure(self, address): | |
38 if not self.sorted: | |
39 self.sorted_value_list.sort() | |
40 self.sorted = True | |
41 found_index = bisect.bisect_left(self.sorted_value_list, address) | |
42 found_start_address = self.sorted_value_list[found_index - 1] | |
43 return self.dictionary[found_start_address] | |
44 | |
45 | |
46 _ARGUMENT_TYPE_PATTERN = re.compile('\([^()]*\)(\s*const)?') | |
47 _TEMPLATE_ARGUMENT_PATTERN = re.compile('<[^<>]*>') | |
48 _LEADING_TYPE_PATTERN = re.compile('^.*\s+(\w+::)') | |
49 | |
50 def _get_short_function_name(function): | |
51 while True: | |
52 function, number = _ARGUMENT_TYPE_PATTERN.subn('', function) | |
M-A Ruel
2012/07/19 15:21:15
A constant is not strictly necessary when used onl
Dai Mikurube (NOT FULLTIME)
2012/07/20 04:32:39
As far as I remember, only one regexp is kept in m
| |
53 if number == 0: | |
M-A Ruel
2012/07/19 15:21:15
if not number:
unless you want to differentiate wi
Dai Mikurube (NOT FULLTIME)
2012/07/20 04:32:39
Done.
| |
54 break | |
55 while True: | |
56 function, number = _TEMPLATE_ARGUMENT_PATTERN.subn('', function) | |
57 if number == 0: | |
58 break | |
59 function = _LEADING_TYPE_PATTERN.sub('\g<1>', function) | |
60 return function | |
61 | |
62 | |
63 def get_procedure_boundaries_from_nm_bsd(FILE, mangled=False): | |
64 """Gets procedure boundaries from a result of nm -n --format bsd. | |
65 | |
66 Args: | |
67 FILE: A file object containing a result of nm. It must be sorted and | |
68 in BSD-style. (Use "[eu-]nm -n --format bsd") | |
69 | |
70 Returns: | |
71 A result ProcedureBoundaryTable object. | |
72 """ | |
73 symbol_table = ProcedureBoundaryTable() | |
74 | |
75 last_start = 0; | |
76 routine = ''; | |
M-A Ruel
2012/07/19 15:21:15
; ?
Dai Mikurube (NOT FULLTIME)
2012/07/20 04:32:39
Wow, removed them. (I was translating this code f
| |
77 | |
78 for line in FILE: | |
79 symbol_info = line.rstrip().split(None, 2) | |
80 if len(symbol_info) == 3: | |
81 if len(symbol_info[0]) == 1: | |
82 symbol_info = line.split(None, 1) | |
83 (sym_type, this_routine) = symbol_info | |
84 sym_value = '' | |
85 else: | |
86 (sym_value, sym_type, this_routine) = symbol_info | |
87 elif len(symbol_info) == 2: | |
88 if len(symbol_info[0]) == 1: | |
89 (sym_type, this_routine) = symbol_info | |
90 sym_value = '' | |
91 elif len(symbol_info[0]) == 8 or len(symbol_info[0]) == 16: | |
92 (sym_value, this_routine) = symbol_info | |
93 sym_type = ' ' | |
94 else: | |
95 raise ParsingException('Invalid output 1 from (eu-)nm.') | |
96 else: | |
97 raise ParsingException('Invalid output 2 from (eu-)nm.') | |
98 | |
99 if sym_value == '': | |
100 continue | |
101 | |
102 start_val = int(sym_value, 16) | |
103 | |
104 # It's possible for two symbols to share the same address, if | |
105 # one is a zero-length variable (like __start_google_malloc) or | |
106 # one symbol is a weak alias to another (like __libc_malloc). | |
107 # In such cases, we want to ignore all values except for the | |
108 # actual symbol, which in nm-speak has type "T". The logic | |
109 # below does this, though it's a bit tricky: what happens when | |
110 # we have a series of lines with the same address, is the first | |
111 # one gets queued up to be processed. However, it won't | |
112 # *actually* be processed until later, when we read a line with | |
113 # a different address. That means that as long as we're reading | |
114 # lines with the same address, we have a chance to replace that | |
115 # item in the queue, which we do whenever we see a 'T' entry -- | |
116 # that is, a line with type 'T'. If we never see a 'T' entry, | |
117 # we'll just go ahead and process the first entry (which never | |
118 # got touched in the queue), and ignore the others. | |
119 if start_val == last_start and (sym_type == 't' or sym_type == 'T'): | |
120 # We are the 'T' symbol at this address, replace previous symbol. | |
121 routine = this_routine | |
122 continue | |
123 elif start_val == last_start: | |
124 # We're not the 'T' symbol at this address, so ignore us. | |
125 continue | |
126 | |
127 # Tag this routine with the starting address in case the image | |
128 # has multiple occurrences of this routine. We use a syntax | |
129 # that resembles template paramters that are automatically | |
130 # stripped out by ShortFunctionName() | |
131 this_routine += "<%016x>" % start_val; | |
132 | |
133 if not mangled: | |
134 routine = _get_short_function_name(routine) | |
135 symbol_table.append(ProcedureBoundary(last_start, start_val, routine)) | |
136 | |
137 last_start = start_val | |
138 routine = this_routine | |
139 | |
140 if not mangled: | |
141 routine = _get_short_function_name(routine) | |
142 symbol_table.append(ProcedureBoundary(last_start, last_start, routine)) | |
143 return symbol_table | |
OLD | NEW |