1 // Copyright 2011 Google Inc. All Rights Reserved.
2 //
3 // Licensed under the Apache License, Version 2.0 (the "License");
4 // you may not use this file except in compliance with the License.
5 // You may obtain a copy of the License at
6 //
7 // http://www.apache.org/licenses/LICENSE-2.0
8 //
9 // Unless required by applicable law or agreed to in writing, software
10 // distributed under the License is distributed on an "AS IS" BASIS,
11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 // See the License for the specific language governing permissions and
13 // limitations under the License.
14
15 #include "depfile_parser.h"
16 #include "util.h"
17
18 #include <algorithm>
19
20 using namespace std;
21
DepfileParser(DepfileParserOptions options)22 DepfileParser::DepfileParser(DepfileParserOptions options)
23 : options_(options)
24 {
25 }
26
27 // A note on backslashes in Makefiles, from reading the docs:
28 // Backslash-newline is the line continuation character.
29 // Backslash-# escapes a # (otherwise meaningful as a comment start).
30 // Backslash-% escapes a % (otherwise meaningful as a special).
31 // Finally, quoting the GNU manual, "Backslashes that are not in danger
32 // of quoting ‘%’ characters go unmolested."
33 // How do you end a line with a backslash? The netbsd Make docs suggest
34 // reading the result of a shell command echoing a backslash!
35 //
36 // Rather than implement all of above, we follow what GCC/Clang produces:
37 // Backslashes escape a space or hash sign.
38 // When a space is preceded by 2N+1 backslashes, it is represents N backslashes
39 // followed by space.
40 // When a space is preceded by 2N backslashes, it represents 2N backslashes at
41 // the end of a filename.
42 // A hash sign is escaped by a single backslash. All other backslashes remain
43 // unchanged.
44 //
45 // If anyone actually has depfiles that rely on the more complicated
46 // behavior we can adjust this.
Parse(string* content, string* err)47 bool DepfileParser::Parse(string* content, string* err) {
48 // in: current parser input point.
49 // end: end of input.
50 // parsing_targets: whether we are parsing targets or dependencies.
51 char* in = &(*content)[0];
52 char* end = in + content->size();
53 bool have_target = false;
54 bool parsing_targets = true;
55 bool poisoned_input = false;
56 bool is_empty = true;
57 while (in < end) {
58 bool have_newline = false;
59 // out: current output point (typically same as in, but can fall behind
60 // as we de-escape backslashes).
61 char* out = in;
62 // filename: start of the current parsed filename.
63 char* filename = out;
64 for (;;) {
65 // start: beginning of the current parsed span.
66 const char* start = in;
67 char* yymarker = NULL;
68 /*!re2c
69 re2c:define:YYCTYPE = "unsigned char";
70 re2c:define:YYCURSOR = in;
71 re2c:define:YYLIMIT = end;
72 re2c:define:YYMARKER = yymarker;
73
74 re2c:yyfill:enable = 0;
75
76 re2c:indent:top = 2;
77 re2c:indent:string = " ";
78
79 nul = "\000";
80 newline = '\r'?'\n';
81
82 '\\\\'* '\\ ' {
83 // 2N+1 backslashes plus space -> N backslashes plus space.
84 int len = (int)(in - start);
85 int n = len / 2 - 1;
86 if (out < start)
87 memset(out, '\\', n);
88 out += n;
89 *out++ = ' ';
90 continue;
91 }
92 '\\\\'+ ' ' {
93 // 2N backslashes plus space -> 2N backslashes, end of filename.
94 int len = (int)(in - start);
95 if (out < start)
96 memset(out, '\\', len - 1);
97 out += len - 1;
98 break;
99 }
100 '\\'+ '#' {
101 // De-escape hash sign, but preserve other leading backslashes.
102 int len = (int)(in - start);
103 if (len > 2 && out < start)
104 memset(out, '\\', len - 2);
105 out += len - 2;
106 *out++ = '#';
107 continue;
108 }
109 '\\'+ ':' [\x00\x20\r\n\t] {
110 // Backslash followed by : and whitespace.
111 // It is therefore normal text and not an escaped colon
112 int len = (int)(in - start - 1);
113 // Need to shift it over if we're overwriting backslashes.
114 if (out < start)
115 memmove(out, start, len);
116 out += len;
117 if (*(in - 1) == '\n')
118 have_newline = true;
119 break;
120 }
121 '\\'+ ':' {
122 // De-escape colon sign, but preserve other leading backslashes.
123 // Regular expression uses lookahead to make sure that no whitespace
124 // nor EOF follows. In that case it'd be the : at the end of a target
125 int len = (int)(in - start);
126 if (len > 2 && out < start)
127 memset(out, '\\', len - 2);
128 out += len - 2;
129 *out++ = ':';
130 continue;
131 }
132 '$$' {
133 // De-escape dollar character.
134 *out++ = '$';
135 continue;
136 }
137 '\\'+ [^\000\r\n] | [a-zA-Z0-9+,/_:.~()}{%=@\x5B\x5D!\x80-\xFF-]+ {
138 // Got a span of plain text.
139 int len = (int)(in - start);
140 // Need to shift it over if we're overwriting backslashes.
141 if (out < start)
142 memmove(out, start, len);
143 out += len;
144 continue;
145 }
146 nul {
147 break;
148 }
149 '\\' newline {
150 // A line continuation ends the current file name.
151 break;
152 }
153 newline {
154 // A newline ends the current file name and the current rule.
155 have_newline = true;
156 break;
157 }
158 [^] {
159 // For any other character (e.g. whitespace), swallow it here,
160 // allowing the outer logic to loop around again.
161 break;
162 }
163 */
164 }
165
166 int len = (int)(out - filename);
167 const bool is_dependency = !parsing_targets;
168 if (len > 0 && filename[len - 1] == ':') {
169 len--; // Strip off trailing colon, if any.
170 parsing_targets = false;
171 have_target = true;
172 }
173
174 if (len > 0) {
175 is_empty = false;
176 StringPiece piece = StringPiece(filename, len);
177 // If we've seen this as an input before, skip it.
178 std::vector<StringPiece>::iterator pos = std::find(ins_.begin(), ins_.end(), piece);
179 if (pos == ins_.end()) {
180 if (is_dependency) {
181 if (poisoned_input) {
182 *err = "inputs may not also have inputs";
183 return false;
184 }
185 // New input.
186 ins_.push_back(piece);
187 } else {
188 // Check for a new output.
189 if (std::find(outs_.begin(), outs_.end(), piece) == outs_.end())
190 outs_.push_back(piece);
191 }
192 } else if (!is_dependency) {
193 // We've passed an input on the left side; reject new inputs.
194 poisoned_input = true;
195 }
196 }
197
198 if (have_newline) {
199 // A newline ends a rule so the next filename will be a new target.
200 parsing_targets = true;
201 poisoned_input = false;
202 }
203 }
204 if (!have_target && !is_empty) {
205 *err = "expected ':' in depfile";
206 return false;
207 }
208 return true;
209 }
210