line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
package GCC::TranslationUnit; |
2
|
1
|
|
|
1
|
|
43055
|
use strict; |
|
1
|
|
|
|
|
2
|
|
|
1
|
|
|
|
|
49
|
|
3
|
|
|
|
|
|
|
|
4
|
|
|
|
|
|
|
our $VERSION = "1.00"; |
5
|
|
|
|
|
|
|
|
6
|
1
|
|
|
1
|
|
675
|
use GCC::Tree; |
|
1
|
|
|
|
|
3
|
|
|
1
|
|
|
|
|
297
|
|
7
|
|
|
|
|
|
|
|
8
|
|
|
|
|
|
|
package GCC::TranslationUnit::Parser; |
9
|
|
|
|
|
|
|
|
10
|
|
|
|
|
|
|
# This class parses the GCC translation unit, as dumped by the |
11
|
|
|
|
|
|
|
# -fdump-translation-unit flag in gcc-3.2.2, and sticks it into a Perl |
12
|
|
|
|
|
|
|
# datastructure intended to somewhat mirror the tree structure |
13
|
|
|
|
|
|
|
# documented in gcc/doc/c-tree.texi, from that same version of gcc. |
14
|
|
|
|
|
|
|
|
15
|
|
|
|
|
|
|
# C++ overloaded operator names, as per cp/dump.c in gcc-3.2.2 |
16
|
|
|
|
|
|
|
our %ops = ( |
17
|
|
|
|
|
|
|
'new' => "new", |
18
|
|
|
|
|
|
|
vecnew => "new[]", |
19
|
|
|
|
|
|
|
'delete' => "delete", |
20
|
|
|
|
|
|
|
vecdelete => "delete[]", |
21
|
|
|
|
|
|
|
'pos' => "+", |
22
|
|
|
|
|
|
|
neg => "-", |
23
|
|
|
|
|
|
|
addr => "&", |
24
|
|
|
|
|
|
|
deref => "*", |
25
|
|
|
|
|
|
|
'not' => "~", |
26
|
|
|
|
|
|
|
lnot => "!", |
27
|
|
|
|
|
|
|
preinc => "++", |
28
|
|
|
|
|
|
|
predec => "--", |
29
|
|
|
|
|
|
|
plus => "+", |
30
|
|
|
|
|
|
|
plusassign => "+=", |
31
|
|
|
|
|
|
|
minus => "-", |
32
|
|
|
|
|
|
|
minusassign => "-=", |
33
|
|
|
|
|
|
|
mult => "*", |
34
|
|
|
|
|
|
|
multassign => "*=", |
35
|
|
|
|
|
|
|
div => "/", |
36
|
|
|
|
|
|
|
divassign => "/=", |
37
|
|
|
|
|
|
|
mod => "%", |
38
|
|
|
|
|
|
|
modassign => "%=", |
39
|
|
|
|
|
|
|
'and' => "&", |
40
|
|
|
|
|
|
|
andassign => "&=", |
41
|
|
|
|
|
|
|
'or' => "|", |
42
|
|
|
|
|
|
|
orassign => "|=", |
43
|
|
|
|
|
|
|
'xor' => "^", |
44
|
|
|
|
|
|
|
xorassign => "^=", |
45
|
|
|
|
|
|
|
lshift => "<<", |
46
|
|
|
|
|
|
|
lshiftassign => "<<=", |
47
|
|
|
|
|
|
|
rshift => ">>", |
48
|
|
|
|
|
|
|
rshiftassign => ">>=", |
49
|
|
|
|
|
|
|
'eq' => "==", |
50
|
|
|
|
|
|
|
'ne' => "!=", |
51
|
|
|
|
|
|
|
'lt' => "<", |
52
|
|
|
|
|
|
|
'gt' => ">", |
53
|
|
|
|
|
|
|
'le' => "<=", |
54
|
|
|
|
|
|
|
'ge' => ">=", |
55
|
|
|
|
|
|
|
land => "&&", |
56
|
|
|
|
|
|
|
lor => "||", |
57
|
|
|
|
|
|
|
compound => ",", |
58
|
|
|
|
|
|
|
memref => "->*", |
59
|
|
|
|
|
|
|
'ref' => "->", |
60
|
|
|
|
|
|
|
subs => "[]", |
61
|
|
|
|
|
|
|
postinc => "++", |
62
|
|
|
|
|
|
|
postdec => "--", |
63
|
|
|
|
|
|
|
call => "()", |
64
|
|
|
|
|
|
|
assign => "=" |
65
|
|
|
|
|
|
|
); |
66
|
|
|
|
|
|
|
|
67
|
|
|
|
|
|
|
# op names for use in regex |
68
|
|
|
|
|
|
|
my $opnames = join '|', keys(%ops); |
69
|
|
|
|
|
|
|
|
70
|
|
|
|
|
|
|
# I'm using the standard -fdump-translation-unit format. Anyone is welcome |
71
|
|
|
|
|
|
|
# to provide an XML parser for the various XML format patches to GCC. |
72
|
|
|
|
|
|
|
|
73
|
1
|
|
|
1
|
|
1750
|
use IO::File; |
|
1
|
|
|
|
|
16745
|
|
|
1
|
|
|
|
|
7724
|
|
74
|
|
|
|
|
|
|
|
75
|
|
|
|
|
|
|
# My initial parser was regarded as 'unreadable' and 'uncommented' by |
76
|
|
|
|
|
|
|
# some people. That was unfortunate. Here, have a comment: |
77
|
|
|
|
|
|
|
|
78
|
|
|
|
|
|
|
# Usage: $tu = GCC::TranslationUnit::Parser->parsefile("file.c.tu") |
79
|
|
|
|
|
|
|
# |
80
|
|
|
|
|
|
|
# Better yet, save yourself some memory and do: |
81
|
|
|
|
|
|
|
# $node = GCC::TranslationUnit::Parser->parsefile("file.c.tu")->root; |
82
|
|
|
|
|
|
|
# |
83
|
|
|
|
|
|
|
# Remember, $tu is an N-element array, where N is the number of nodes GCC |
84
|
|
|
|
|
|
|
# dumped. Only keep the array if you plan to iterate through every element |
85
|
|
|
|
|
|
|
# in the GCC tree; otherwise, free the memory used by the array, and let |
86
|
|
|
|
|
|
|
# the Perl reference counter free the node branches if you delete them. |
87
|
|
|
|
|
|
|
sub parsefile { |
88
|
0
|
|
|
0
|
|
|
my $class = shift; |
89
|
0
|
|
|
|
|
|
my $file = shift; |
90
|
|
|
|
|
|
|
|
91
|
0
|
|
|
|
|
|
my $fh = new IO::File $file; |
92
|
0
|
0
|
|
|
|
|
return undef unless defined $fh; # Not my problem if I can't open your file |
93
|
|
|
|
|
|
|
|
94
|
0
|
|
|
|
|
|
my $self = bless [], $class; |
95
|
0
|
|
|
|
|
|
my $dump = ""; # the text of a single dumped node |
96
|
0
|
|
|
|
|
|
my $index = 0; # numerical index of the "current" node |
97
|
|
|
|
|
|
|
|
98
|
0
|
|
|
|
|
|
my $line; |
99
|
0
|
|
|
|
|
|
while($line = $fh->getline) { |
100
|
0
|
0
|
|
|
|
|
if($line =~ /^\@(\d+)/) { |
101
|
|
|
|
|
|
|
# The first line of a node should look like: |
102
|
|
|
|
|
|
|
# "@123 some_node ..." |
103
|
|
|
|
|
|
|
# Every other line is requried to be indented |
104
|
|
|
|
|
|
|
# |
105
|
|
|
|
|
|
|
# When we find that "first" line, or EOF, parse the previous node |
106
|
0
|
|
|
|
|
|
$self->parsenode($dump, $self->node($index)); |
107
|
0
|
|
|
|
|
|
$self->node($index)->{position} = $index; |
108
|
0
|
|
|
|
|
|
$dump = $line; |
109
|
0
|
|
|
|
|
|
$index = $1; |
110
|
|
|
|
|
|
|
} else { |
111
|
0
|
|
|
|
|
|
$dump .= $line; |
112
|
|
|
|
|
|
|
} |
113
|
|
|
|
|
|
|
} |
114
|
0
|
|
|
|
|
|
$self->parsenode($dump, $self->node($index)); |
115
|
|
|
|
|
|
|
|
116
|
0
|
|
|
|
|
|
return $self; |
117
|
|
|
|
|
|
|
} |
118
|
|
|
|
|
|
|
|
119
|
|
|
|
|
|
|
# Returns the root node, from GCC's perspective. |
120
|
|
|
|
|
|
|
# Usage: $tu->root |
121
|
0
|
|
|
0
|
|
|
sub root { shift->[1] } |
122
|
|
|
|
|
|
|
|
123
|
|
|
|
|
|
|
# Take the complete dumped text in $dump from a single GCC node and |
124
|
|
|
|
|
|
|
# stuff it into $node. |
125
|
|
|
|
|
|
|
# |
126
|
|
|
|
|
|
|
# Usage: $tu->parsenode($dump, $tu->node($index)); |
127
|
|
|
|
|
|
|
sub parsenode { |
128
|
0
|
|
|
0
|
|
|
my($self, $dump, $node) = @_; |
129
|
0
|
0
|
|
|
|
|
return unless $dump; |
130
|
|
|
|
|
|
|
# print "".("-" x 70) . "\n"; |
131
|
|
|
|
|
|
|
# print $dump; |
132
|
|
|
|
|
|
|
|
133
|
|
|
|
|
|
|
# Note that this regex leaves a space before the first key. |
134
|
|
|
|
|
|
|
# |
135
|
|
|
|
|
|
|
# That's because the key matching rule is /\s\w.{3}:\s/ |
136
|
0
|
0
|
|
|
|
|
unless($dump =~ s/^\@(\d+)\s+(\w+)(?=\s)//) { |
137
|
0
|
|
|
|
|
|
warn "Unknown node format:\n$dump"; |
138
|
0
|
|
|
|
|
|
return; |
139
|
|
|
|
|
|
|
} |
140
|
|
|
|
|
|
|
|
141
|
0
|
|
|
|
|
|
my $index = $1; |
142
|
0
|
|
|
|
|
|
my $type = $2; |
143
|
|
|
|
|
|
|
# print "tu[$1] = $type\n"; |
144
|
|
|
|
|
|
|
|
145
|
0
|
|
|
|
|
|
bless $node, "GCC::Node::$type"; |
146
|
0
|
|
|
|
|
|
$node->{INDEX} = $index; |
147
|
|
|
|
|
|
|
|
148
|
|
|
|
|
|
|
# First, some examples as to what's possible as a node element |
149
|
|
|
|
|
|
|
# |
150
|
|
|
|
|
|
|
# dump_index() writes: |
151
|
|
|
|
|
|
|
# node: @123 |
152
|
|
|
|
|
|
|
# |
153
|
|
|
|
|
|
|
# dump_pointer() writes: |
154
|
|
|
|
|
|
|
# node: 2468ace0 |
155
|
|
|
|
|
|
|
# |
156
|
|
|
|
|
|
|
# dump_int() writes: |
157
|
|
|
|
|
|
|
# node: 1234567 |
158
|
|
|
|
|
|
|
# |
159
|
|
|
|
|
|
|
# dump_string() writes: |
160
|
|
|
|
|
|
|
# string |
161
|
|
|
|
|
|
|
# |
162
|
|
|
|
|
|
|
# dump_string_field() writes: |
163
|
|
|
|
|
|
|
# node: string |
164
|
|
|
|
|
|
|
|
165
|
|
|
|
|
|
|
|
166
|
|
|
|
|
|
|
# The string nodes can seriously disrupt us, since they aren't quoted. |
167
|
|
|
|
|
|
|
# They are VERY inconvenient, so we must eliminate them first. |
168
|
|
|
|
|
|
|
# |
169
|
|
|
|
|
|
|
# Anything that looks like an identifier_node string must be squashed ASAP |
170
|
0
|
0
|
|
|
|
|
if($dump =~ s/\s+strg:\s(.*)\slngt:\s(\d+)//s) { |
171
|
|
|
|
|
|
|
# identifier_node and string_cst come here, at least |
172
|
0
|
|
|
|
|
|
my($string, $length) = ($1, $2); |
173
|
|
|
|
|
|
|
|
174
|
|
|
|
|
|
|
# string_cst's lngt includes the NUL character, which fprintf() |
175
|
|
|
|
|
|
|
# doesn't print, obviously. Make sure to factor that in... |
176
|
0
|
0
|
|
|
|
|
$length-- if $type eq 'string_cst'; |
177
|
|
|
|
|
|
|
|
178
|
0
|
|
|
|
|
|
$node->{'string'} = substr($string, 0, $length); |
179
|
0
|
|
|
|
|
|
$node->{'length'} = $length; |
180
|
|
|
|
|
|
|
# print "string: '$node->{string}'\n"; |
181
|
|
|
|
|
|
|
} |
182
|
|
|
|
|
|
|
|
183
|
|
|
|
|
|
|
# The srcp key is BAD. It contains a colon as part of the value, which |
184
|
|
|
|
|
|
|
# could ruin the key parser regex. It's gotta go. Not to mention filenames |
185
|
|
|
|
|
|
|
# with spaces... |
186
|
|
|
|
|
|
|
# |
187
|
|
|
|
|
|
|
# Example: |
188
|
|
|
|
|
|
|
# srcp: file.c:123 |
189
|
0
|
0
|
|
|
|
|
$node->{'source'} = $1 if $dump =~ s/\ssrcp:\s(.*?:\d+)(?=\s)//; |
190
|
|
|
|
|
|
|
# print "source: '$node->{source}'\n" if $node->{source}; |
191
|
|
|
|
|
|
|
|
192
|
|
|
|
|
|
|
# The remaining nodes are pretty regular and easy to parse. |
193
|
|
|
|
|
|
|
# However, the flags which crop up everywhere disrupt my ability |
194
|
|
|
|
|
|
|
# to determine the end of a value corresponding to a key. for instance: |
195
|
|
|
|
|
|
|
# |
196
|
|
|
|
|
|
|
# key : value protected |
197
|
|
|
|
|
|
|
# |
198
|
|
|
|
|
|
|
# Depending on what the key represents, the protected flag may or |
199
|
|
|
|
|
|
|
# may not be a part of the value. In order to remove the ambiguity, |
200
|
|
|
|
|
|
|
# we have to manually parse out any keys whose values aren't captured |
201
|
|
|
|
|
|
|
# with a trivial /\s(\w.{3}):\s(\S)\s/ match. |
202
|
|
|
|
|
|
|
|
203
|
|
|
|
|
|
|
# First violator: |
204
|
|
|
|
|
|
|
# "qual: %c%c%c " |
205
|
0
|
0
|
|
|
|
|
if($dump =~ s/\squal:\s(.{3})\s//) { |
206
|
0
|
|
|
|
|
|
my $qual = $1; |
207
|
0
|
0
|
|
|
|
|
$node->{'const'} = 1 if $qual =~ /c/; |
208
|
0
|
0
|
|
|
|
|
$node->{'volatile'} = 1 if $qual =~ /v/; |
209
|
0
|
0
|
|
|
|
|
$node->{'restrict'} = 1 if $qual =~ /r/; |
210
|
|
|
|
|
|
|
} |
211
|
|
|
|
|
|
|
|
212
|
|
|
|
|
|
|
# next violator: |
213
|
|
|
|
|
|
|
# base: @1234 virtual public |
214
|
|
|
|
|
|
|
# |
215
|
0
|
|
|
|
|
|
while($dump =~ s/\sbase:\s\@(\d+)\s+(.*?)\s*(public|private|protected)//) { |
216
|
|
|
|
|
|
|
# base is the only key which can appear multiple times in the same |
217
|
|
|
|
|
|
|
# node, since it's spit out from an array. We need to put it back. |
218
|
|
|
|
|
|
|
# |
219
|
|
|
|
|
|
|
# The other vector nodes have a conveniently unique number, like 'op 0' |
220
|
0
|
|
|
|
|
|
my $classid = $1; |
221
|
0
|
0
|
|
|
|
|
my $virtual = $2 ? 1 : 0; # yes, yes, !!$2. vim don't like it, though |
222
|
0
|
|
|
|
|
|
my $access = $3; |
223
|
0
|
|
|
|
|
|
my $base = { |
224
|
|
|
|
|
|
|
class => $self->node($classid), |
225
|
|
|
|
|
|
|
virtual => $virtual, |
226
|
|
|
|
|
|
|
access => $access |
227
|
|
|
|
|
|
|
}; |
228
|
|
|
|
|
|
|
# print "base: $virtual $access $classid\n"; |
229
|
0
|
|
|
|
|
|
push @{ $node->{'base'} }, $base; |
|
0
|
|
|
|
|
|
|
230
|
|
|
|
|
|
|
} |
231
|
|
|
|
|
|
|
|
232
|
|
|
|
|
|
|
# Some GCC developer forgot to read the -fdump spec before dipping his |
233
|
|
|
|
|
|
|
# fingers into gcc/cp/dump.c. |
234
|
|
|
|
|
|
|
# |
235
|
|
|
|
|
|
|
# Remember: 4 character max per key! |
236
|
0
|
0
|
|
|
|
|
$node->{raises} = $self->node($1) if $dump =~ s/\sraises: \@(\d+)//; |
237
|
|
|
|
|
|
|
|
238
|
|
|
|
|
|
|
# At this point, we assume all the remaining key/value pairs match |
239
|
|
|
|
|
|
|
# the following regex. |
240
|
0
|
|
|
|
|
|
while($dump =~ s/\s(\w.{0,3}?)\s*:\s(\S+)//) { |
241
|
0
|
|
|
|
|
|
my($key, $value) = ($1, $2); |
242
|
0
|
0
|
|
|
|
|
$value = $self->node($1) if $value =~ /^\@(\d+)/; |
243
|
|
|
|
|
|
|
# print "'$key': '$value'\n"; |
244
|
|
|
|
|
|
|
|
245
|
|
|
|
|
|
|
# If the key looks like it came out of a tree_vec or operand list, |
246
|
|
|
|
|
|
|
# stick it back into an array, to save us the trouble of doing |
247
|
|
|
|
|
|
|
# bounds-checking hash fetching voodoo. |
248
|
|
|
|
|
|
|
# |
249
|
|
|
|
|
|
|
# Although, those tree_vec nodes skip elements when indexed in decimal. |
250
|
0
|
0
|
|
|
|
|
if($key =~ /^\d+$/) { |
|
|
0
|
|
|
|
|
|
251
|
0
|
|
|
|
|
|
$node->{vector}[$key] = $value; |
252
|
|
|
|
|
|
|
} elsif($key =~ /^op (\d+)$/) { |
253
|
0
|
|
|
|
|
|
$node->{operand}[$1] = $value; |
254
|
|
|
|
|
|
|
} else { |
255
|
0
|
|
|
|
|
|
$node->{$key} = $value; |
256
|
|
|
|
|
|
|
} |
257
|
|
|
|
|
|
|
} |
258
|
|
|
|
|
|
|
|
259
|
|
|
|
|
|
|
# Before we can consider the rest of the data as a sequence of flags, |
260
|
|
|
|
|
|
|
# I need to remove a few special-case flags which can be thought of as |
261
|
|
|
|
|
|
|
# being "intentionally sequential" for whatever reason. |
262
|
|
|
|
|
|
|
|
263
|
|
|
|
|
|
|
# operator is fun. The next flag is GCC's operator "name", for which there |
264
|
|
|
|
|
|
|
# is a mapping to the C operator, above, in the declaration of %ops. |
265
|
0
|
0
|
|
|
|
|
if($dump =~ s/\soperator\s+($opnames)\b//o) { |
266
|
0
|
|
|
|
|
|
$node->{operator} = $1; |
267
|
|
|
|
|
|
|
# print "operator $ops{$node->{operator}}\n"; |
268
|
|
|
|
|
|
|
} |
269
|
|
|
|
|
|
|
|
270
|
|
|
|
|
|
|
# Some flags have spaces in them! Parse it as one string. |
271
|
0
|
|
|
|
|
|
while($dump =~ s/\s(global init|global fini|pseudo tmpl)(?=\s)//) { |
272
|
|
|
|
|
|
|
# Honestly, I'd rather s/foo bar/foo_bar/g instead |
273
|
|
|
|
|
|
|
# print "TRUE $1\n"; |
274
|
0
|
|
|
|
|
|
$node->{$1} = 1; |
275
|
|
|
|
|
|
|
} |
276
|
|
|
|
|
|
|
|
277
|
|
|
|
|
|
|
# For sanity's sake, lets save the access |
278
|
0
|
0
|
|
|
|
|
$node->{access} = $1 if $dump =~ /\b(public|private|protected)\b/; |
279
|
|
|
|
|
|
|
|
280
|
|
|
|
|
|
|
# All that should remain is flags |
281
|
0
|
|
|
|
|
|
while($dump =~ s/(\w+)//) { |
282
|
|
|
|
|
|
|
# print "TRUE $1\n"; |
283
|
0
|
|
|
|
|
|
$node->{$1} = 1; |
284
|
|
|
|
|
|
|
} |
285
|
|
|
|
|
|
|
|
286
|
|
|
|
|
|
|
# For debugging purposes, check for extra characters. |
287
|
|
|
|
|
|
|
# If this warning occurs, it's a bug in the parser. Let me know about it. |
288
|
|
|
|
|
|
|
# First, make sure you didn't run out of diskspace when writing the |
289
|
|
|
|
|
|
|
# -fdump file. That would truncate the file and cause this warning, and |
290
|
|
|
|
|
|
|
# that's not my fault. |
291
|
|
|
|
|
|
|
# |
292
|
|
|
|
|
|
|
# Please include details with any error reports: the version of GCC you |
293
|
|
|
|
|
|
|
# used, your operating system (both kernel and distribution), and the |
294
|
|
|
|
|
|
|
# source of the file which generated the error, if possible. |
295
|
0
|
0
|
|
|
|
|
if($dump =~ /\S/) { |
296
|
0
|
|
|
|
|
|
$dump =~ s/\s+/ /g; |
297
|
0
|
|
|
|
|
|
warn "Unparsed data: $dump\nFrom: $_[1] "; |
298
|
|
|
|
|
|
|
} |
299
|
|
|
|
|
|
|
|
300
|
0
|
|
|
|
|
|
return $node; |
301
|
|
|
|
|
|
|
} |
302
|
|
|
|
|
|
|
|
303
|
|
|
|
|
|
|
# Since the dump format includes forward references, we need to pre-initialize |
304
|
|
|
|
|
|
|
# nodes we haven't parsed yet before we assign them to the nodes which |
305
|
|
|
|
|
|
|
# reference them. Therefore, you must ALWAYS get a node using this function |
306
|
|
|
|
|
|
|
# until the parsefile() routine returns a complete translation unit! |
307
|
|
|
|
|
|
|
# |
308
|
|
|
|
|
|
|
# Usage: $tu->node($id) |
309
|
|
|
|
|
|
|
sub node { |
310
|
0
|
|
|
0
|
|
|
my($self, $id) = @_; |
311
|
0
|
0
|
|
|
|
|
unless($self->[$id]) { |
312
|
|
|
|
|
|
|
# Don't un-block this unless; I like adding debug stuff here |
313
|
0
|
|
|
|
|
|
$self->[$id] = {}; |
314
|
|
|
|
|
|
|
} |
315
|
0
|
|
|
|
|
|
return $self->[$id]; |
316
|
|
|
|
|
|
|
} |
317
|
|
|
|
|
|
|
|
318
|
|
|
|
|
|
|
# vim:set shiftwidth=4 softtabstop=4: |
319
|
|
|
|
|
|
|
1; |
320
|
|
|
|
|
|
|
__END__ |