1 // Protocol Buffers - Google's data interchange format
2 // Copyright 2008 Google Inc.  All rights reserved.
3 // https://developers.google.com/protocol-buffers/
4 //
5 // Redistribution and use in source and binary forms, with or without
6 // modification, are permitted provided that the following conditions are
7 // met:
8 //
9 //     * Redistributions of source code must retain the above copyright
10 // notice, this list of conditions and the following disclaimer.
11 //     * Redistributions in binary form must reproduce the above
12 // copyright notice, this list of conditions and the following disclaimer
13 // in the documentation and/or other materials provided with the
14 // distribution.
15 //     * Neither the name of Google Inc. nor the names of its
16 // contributors may be used to endorse or promote products derived from
17 // this software without specific prior written permission.
18 //
19 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 
31 // Author: kenton@google.com (Kenton Varda)
32 //  Based on original Protocol Buffers design by
33 //  Sanjay Ghemawat, Jeff Dean, and others.
34 
35 #ifdef _MSC_VER
36 #include <io.h>
37 #else
38 #include <unistd.h>
39 #endif
40 #include <sys/types.h>
41 #include <sys/stat.h>
42 #include <fcntl.h>
43 #include <errno.h>
44 
45 #include <algorithm>
46 #include <memory>
47 #ifndef _SHARED_PTR_H
48 #include <google/protobuf/stubs/shared_ptr.h>
49 #endif
50 
51 #include <google/protobuf/compiler/importer.h>
52 
53 #include <google/protobuf/compiler/parser.h>
54 #include <google/protobuf/io/tokenizer.h>
55 #include <google/protobuf/io/zero_copy_stream_impl.h>
56 #include <google/protobuf/stubs/strutil.h>
57 
58 namespace google {
59 namespace protobuf {
60 namespace compiler {
61 
62 #ifdef _WIN32
63 #ifndef F_OK
64 #define F_OK 00  // not defined by MSVC for whatever reason
65 #endif
66 #include <ctype.h>
67 #endif
68 
69 // Returns true if the text looks like a Windows-style absolute path, starting
70 // with a drive letter.  Example:  "C:\foo".  TODO(kenton):  Share this with
71 // copy in command_line_interface.cc?
IsWindowsAbsolutePath(const string & text)72 static bool IsWindowsAbsolutePath(const string& text) {
73 #if defined(_WIN32) || defined(__CYGWIN__)
74   return text.size() >= 3 && text[1] == ':' &&
75          isalpha(text[0]) &&
76          (text[2] == '/' || text[2] == '\\') &&
77          text.find_last_of(':') == 1;
78 #else
79   return false;
80 #endif
81 }
82 
~MultiFileErrorCollector()83 MultiFileErrorCollector::~MultiFileErrorCollector() {}
84 
85 // This class serves two purposes:
86 // - It implements the ErrorCollector interface (used by Tokenizer and Parser)
87 //   in terms of MultiFileErrorCollector, using a particular filename.
88 // - It lets us check if any errors have occurred.
89 class SourceTreeDescriptorDatabase::SingleFileErrorCollector
90     : public io::ErrorCollector {
91  public:
SingleFileErrorCollector(const string & filename,MultiFileErrorCollector * multi_file_error_collector)92   SingleFileErrorCollector(const string& filename,
93                            MultiFileErrorCollector* multi_file_error_collector)
94     : filename_(filename),
95       multi_file_error_collector_(multi_file_error_collector),
96       had_errors_(false) {}
~SingleFileErrorCollector()97   ~SingleFileErrorCollector() {}
98 
had_errors()99   bool had_errors() { return had_errors_; }
100 
101   // implements ErrorCollector ---------------------------------------
AddError(int line,int column,const string & message)102   void AddError(int line, int column, const string& message) {
103     if (multi_file_error_collector_ != NULL) {
104       multi_file_error_collector_->AddError(filename_, line, column, message);
105     }
106     had_errors_ = true;
107   }
108 
109  private:
110   string filename_;
111   MultiFileErrorCollector* multi_file_error_collector_;
112   bool had_errors_;
113 };
114 
115 // ===================================================================
116 
SourceTreeDescriptorDatabase(SourceTree * source_tree)117 SourceTreeDescriptorDatabase::SourceTreeDescriptorDatabase(
118     SourceTree* source_tree)
119   : source_tree_(source_tree),
120     error_collector_(NULL),
121     using_validation_error_collector_(false),
122     validation_error_collector_(this) {}
123 
~SourceTreeDescriptorDatabase()124 SourceTreeDescriptorDatabase::~SourceTreeDescriptorDatabase() {}
125 
FindFileByName(const string & filename,FileDescriptorProto * output)126 bool SourceTreeDescriptorDatabase::FindFileByName(
127     const string& filename, FileDescriptorProto* output) {
128   google::protobuf::scoped_ptr<io::ZeroCopyInputStream> input(source_tree_->Open(filename));
129   if (input == NULL) {
130     if (error_collector_ != NULL) {
131       error_collector_->AddError(filename, -1, 0,
132                                  source_tree_->GetLastErrorMessage());
133     }
134     return false;
135   }
136 
137   // Set up the tokenizer and parser.
138   SingleFileErrorCollector file_error_collector(filename, error_collector_);
139   io::Tokenizer tokenizer(input.get(), &file_error_collector);
140 
141   Parser parser;
142   if (error_collector_ != NULL) {
143     parser.RecordErrorsTo(&file_error_collector);
144   }
145   if (using_validation_error_collector_) {
146     parser.RecordSourceLocationsTo(&source_locations_);
147   }
148 
149   // Parse it.
150   output->set_name(filename);
151   return parser.Parse(&tokenizer, output) &&
152          !file_error_collector.had_errors();
153 }
154 
FindFileContainingSymbol(const string & symbol_name,FileDescriptorProto * output)155 bool SourceTreeDescriptorDatabase::FindFileContainingSymbol(
156     const string& symbol_name, FileDescriptorProto* output) {
157   return false;
158 }
159 
FindFileContainingExtension(const string & containing_type,int field_number,FileDescriptorProto * output)160 bool SourceTreeDescriptorDatabase::FindFileContainingExtension(
161     const string& containing_type, int field_number,
162     FileDescriptorProto* output) {
163   return false;
164 }
165 
166 // -------------------------------------------------------------------
167 
168 SourceTreeDescriptorDatabase::ValidationErrorCollector::
ValidationErrorCollector(SourceTreeDescriptorDatabase * owner)169 ValidationErrorCollector(SourceTreeDescriptorDatabase* owner)
170   : owner_(owner) {}
171 
172 SourceTreeDescriptorDatabase::ValidationErrorCollector::
~ValidationErrorCollector()173 ~ValidationErrorCollector() {}
174 
AddError(const string & filename,const string & element_name,const Message * descriptor,ErrorLocation location,const string & message)175 void SourceTreeDescriptorDatabase::ValidationErrorCollector::AddError(
176     const string& filename,
177     const string& element_name,
178     const Message* descriptor,
179     ErrorLocation location,
180     const string& message) {
181   if (owner_->error_collector_ == NULL) return;
182 
183   int line, column;
184   owner_->source_locations_.Find(descriptor, location, &line, &column);
185   owner_->error_collector_->AddError(filename, line, column, message);
186 }
187 
AddWarning(const string & filename,const string & element_name,const Message * descriptor,ErrorLocation location,const string & message)188 void SourceTreeDescriptorDatabase::ValidationErrorCollector::AddWarning(
189     const string& filename,
190     const string& element_name,
191     const Message* descriptor,
192     ErrorLocation location,
193     const string& message) {
194   if (owner_->error_collector_ == NULL) return;
195 
196   int line, column;
197   owner_->source_locations_.Find(descriptor, location, &line, &column);
198   owner_->error_collector_->AddWarning(filename, line, column, message);
199 }
200 
201 // ===================================================================
202 
Importer(SourceTree * source_tree,MultiFileErrorCollector * error_collector)203 Importer::Importer(SourceTree* source_tree,
204                    MultiFileErrorCollector* error_collector)
205   : database_(source_tree),
206     pool_(&database_, database_.GetValidationErrorCollector()) {
207   pool_.EnforceWeakDependencies(true);
208   database_.RecordErrorsTo(error_collector);
209 }
210 
~Importer()211 Importer::~Importer() {}
212 
Import(const string & filename)213 const FileDescriptor* Importer::Import(const string& filename) {
214   return pool_.FindFileByName(filename);
215 }
216 
AddUnusedImportTrackFile(const string & file_name)217 void Importer::AddUnusedImportTrackFile(const string& file_name) {
218   pool_.AddUnusedImportTrackFile(file_name);
219 }
220 
ClearUnusedImportTrackFiles()221 void Importer::ClearUnusedImportTrackFiles() {
222   pool_.ClearUnusedImportTrackFiles();
223 }
224 
225 // ===================================================================
226 
~SourceTree()227 SourceTree::~SourceTree() {}
228 
GetLastErrorMessage()229 string SourceTree::GetLastErrorMessage() {
230   return "File not found.";
231 }
232 
DiskSourceTree()233 DiskSourceTree::DiskSourceTree() {}
234 
~DiskSourceTree()235 DiskSourceTree::~DiskSourceTree() {}
236 
LastChar(const string & str)237 static inline char LastChar(const string& str) {
238   return str[str.size() - 1];
239 }
240 
241 // Given a path, returns an equivalent path with these changes:
242 // - On Windows, any backslashes are replaced with forward slashes.
243 // - Any instances of the directory "." are removed.
244 // - Any consecutive '/'s are collapsed into a single slash.
245 // Note that the resulting string may be empty.
246 //
247 // TODO(kenton):  It would be nice to handle "..", e.g. so that we can figure
248 //   out that "foo/bar.proto" is inside "baz/../foo".  However, if baz is a
249 //   symlink or doesn't exist, then things get complicated, and we can't
250 //   actually determine this without investigating the filesystem, probably
251 //   in non-portable ways.  So, we punt.
252 //
253 // TODO(kenton):  It would be nice to use realpath() here except that it
254 //   resolves symbolic links.  This could cause problems if people place
255 //   symbolic links in their source tree.  For example, if you executed:
256 //     protoc --proto_path=foo foo/bar/baz.proto
257 //   then if foo/bar is a symbolic link, foo/bar/baz.proto will canonicalize
258 //   to a path which does not appear to be under foo, and thus the compiler
259 //   will complain that baz.proto is not inside the --proto_path.
CanonicalizePath(string path)260 static string CanonicalizePath(string path) {
261 #ifdef _WIN32
262   // The Win32 API accepts forward slashes as a path delimiter even though
263   // backslashes are standard.  Let's avoid confusion and use only forward
264   // slashes.
265   if (HasPrefixString(path, "\\\\")) {
266     // Avoid converting two leading backslashes.
267     path = "\\\\" + StringReplace(path.substr(2), "\\", "/", true);
268   } else {
269     path = StringReplace(path, "\\", "/", true);
270   }
271 #endif
272 
273   vector<string> canonical_parts;
274   vector<string> parts = Split(
275       path, "/", true);  // Note:  Removes empty parts.
276   for (int i = 0; i < parts.size(); i++) {
277     if (parts[i] == ".") {
278       // Ignore.
279     } else {
280       canonical_parts.push_back(parts[i]);
281     }
282   }
283   string result = Join(canonical_parts, "/");
284   if (!path.empty() && path[0] == '/') {
285     // Restore leading slash.
286     result = '/' + result;
287   }
288   if (!path.empty() && LastChar(path) == '/' &&
289       !result.empty() && LastChar(result) != '/') {
290     // Restore trailing slash.
291     result += '/';
292   }
293   return result;
294 }
295 
ContainsParentReference(const string & path)296 static inline bool ContainsParentReference(const string& path) {
297   return path == ".." ||
298          HasPrefixString(path, "../") ||
299          HasSuffixString(path, "/..") ||
300          path.find("/../") != string::npos;
301 }
302 
303 // Maps a file from an old location to a new one.  Typically, old_prefix is
304 // a virtual path and new_prefix is its corresponding disk path.  Returns
305 // false if the filename did not start with old_prefix, otherwise replaces
306 // old_prefix with new_prefix and stores the result in *result.  Examples:
307 //   string result;
308 //   assert(ApplyMapping("foo/bar", "", "baz", &result));
309 //   assert(result == "baz/foo/bar");
310 //
311 //   assert(ApplyMapping("foo/bar", "foo", "baz", &result));
312 //   assert(result == "baz/bar");
313 //
314 //   assert(ApplyMapping("foo", "foo", "bar", &result));
315 //   assert(result == "bar");
316 //
317 //   assert(!ApplyMapping("foo/bar", "baz", "qux", &result));
318 //   assert(!ApplyMapping("foo/bar", "baz", "qux", &result));
319 //   assert(!ApplyMapping("foobar", "foo", "baz", &result));
ApplyMapping(const string & filename,const string & old_prefix,const string & new_prefix,string * result)320 static bool ApplyMapping(const string& filename,
321                          const string& old_prefix,
322                          const string& new_prefix,
323                          string* result) {
324   if (old_prefix.empty()) {
325     // old_prefix matches any relative path.
326     if (ContainsParentReference(filename)) {
327       // We do not allow the file name to use "..".
328       return false;
329     }
330     if (HasPrefixString(filename, "/") ||
331         IsWindowsAbsolutePath(filename)) {
332       // This is an absolute path, so it isn't matched by the empty string.
333       return false;
334     }
335     result->assign(new_prefix);
336     if (!result->empty()) result->push_back('/');
337     result->append(filename);
338     return true;
339   } else if (HasPrefixString(filename, old_prefix)) {
340     // old_prefix is a prefix of the filename.  Is it the whole filename?
341     if (filename.size() == old_prefix.size()) {
342       // Yep, it's an exact match.
343       *result = new_prefix;
344       return true;
345     } else {
346       // Not an exact match.  Is the next character a '/'?  Otherwise,
347       // this isn't actually a match at all.  E.g. the prefix "foo/bar"
348       // does not match the filename "foo/barbaz".
349       int after_prefix_start = -1;
350       if (filename[old_prefix.size()] == '/') {
351         after_prefix_start = old_prefix.size() + 1;
352       } else if (filename[old_prefix.size() - 1] == '/') {
353         // old_prefix is never empty, and canonicalized paths never have
354         // consecutive '/' characters.
355         after_prefix_start = old_prefix.size();
356       }
357       if (after_prefix_start != -1) {
358         // Yep.  So the prefixes are directories and the filename is a file
359         // inside them.
360         string after_prefix = filename.substr(after_prefix_start);
361         if (ContainsParentReference(after_prefix)) {
362           // We do not allow the file name to use "..".
363           return false;
364         }
365         result->assign(new_prefix);
366         if (!result->empty()) result->push_back('/');
367         result->append(after_prefix);
368         return true;
369       }
370     }
371   }
372 
373   return false;
374 }
375 
MapPath(const string & virtual_path,const string & disk_path)376 void DiskSourceTree::MapPath(const string& virtual_path,
377                              const string& disk_path) {
378   mappings_.push_back(Mapping(virtual_path, CanonicalizePath(disk_path)));
379 }
380 
381 DiskSourceTree::DiskFileToVirtualFileResult
DiskFileToVirtualFile(const string & disk_file,string * virtual_file,string * shadowing_disk_file)382 DiskSourceTree::DiskFileToVirtualFile(
383     const string& disk_file,
384     string* virtual_file,
385     string* shadowing_disk_file) {
386   int mapping_index = -1;
387   string canonical_disk_file = CanonicalizePath(disk_file);
388 
389   for (int i = 0; i < mappings_.size(); i++) {
390     // Apply the mapping in reverse.
391     if (ApplyMapping(canonical_disk_file, mappings_[i].disk_path,
392                      mappings_[i].virtual_path, virtual_file)) {
393       // Success.
394       mapping_index = i;
395       break;
396     }
397   }
398 
399   if (mapping_index == -1) {
400     return NO_MAPPING;
401   }
402 
403   // Iterate through all mappings with higher precedence and verify that none
404   // of them map this file to some other existing file.
405   for (int i = 0; i < mapping_index; i++) {
406     if (ApplyMapping(*virtual_file, mappings_[i].virtual_path,
407                      mappings_[i].disk_path, shadowing_disk_file)) {
408       if (access(shadowing_disk_file->c_str(), F_OK) >= 0) {
409         // File exists.
410         return SHADOWED;
411       }
412     }
413   }
414   shadowing_disk_file->clear();
415 
416   // Verify that we can open the file.  Note that this also has the side-effect
417   // of verifying that we are not canonicalizing away any non-existent
418   // directories.
419   google::protobuf::scoped_ptr<io::ZeroCopyInputStream> stream(OpenDiskFile(disk_file));
420   if (stream == NULL) {
421     return CANNOT_OPEN;
422   }
423 
424   return SUCCESS;
425 }
426 
VirtualFileToDiskFile(const string & virtual_file,string * disk_file)427 bool DiskSourceTree::VirtualFileToDiskFile(const string& virtual_file,
428                                            string* disk_file) {
429   google::protobuf::scoped_ptr<io::ZeroCopyInputStream> stream(
430       OpenVirtualFile(virtual_file, disk_file));
431   return stream != NULL;
432 }
433 
Open(const string & filename)434 io::ZeroCopyInputStream* DiskSourceTree::Open(const string& filename) {
435   return OpenVirtualFile(filename, NULL);
436 }
437 
GetLastErrorMessage()438 string DiskSourceTree::GetLastErrorMessage() {
439   return last_error_message_;
440 }
441 
OpenVirtualFile(const string & virtual_file,string * disk_file)442 io::ZeroCopyInputStream* DiskSourceTree::OpenVirtualFile(
443     const string& virtual_file,
444     string* disk_file) {
445   if (virtual_file != CanonicalizePath(virtual_file) ||
446       ContainsParentReference(virtual_file)) {
447     // We do not allow importing of paths containing things like ".." or
448     // consecutive slashes since the compiler expects files to be uniquely
449     // identified by file name.
450     last_error_message_ = "Backslashes, consecutive slashes, \".\", or \"..\" "
451                           "are not allowed in the virtual path";
452     return NULL;
453   }
454 
455   for (int i = 0; i < mappings_.size(); i++) {
456     string temp_disk_file;
457     if (ApplyMapping(virtual_file, mappings_[i].virtual_path,
458                      mappings_[i].disk_path, &temp_disk_file)) {
459       io::ZeroCopyInputStream* stream = OpenDiskFile(temp_disk_file);
460       if (stream != NULL) {
461         if (disk_file != NULL) {
462           *disk_file = temp_disk_file;
463         }
464         return stream;
465       }
466 
467       if (errno == EACCES) {
468         // The file exists but is not readable.
469         last_error_message_ = "Read access is denied for file: " +
470                               temp_disk_file;
471         return NULL;
472       }
473     }
474   }
475   last_error_message_ = "File not found.";
476   return NULL;
477 }
478 
OpenDiskFile(const string & filename)479 io::ZeroCopyInputStream* DiskSourceTree::OpenDiskFile(
480     const string& filename) {
481   int file_descriptor;
482   do {
483     file_descriptor = open(filename.c_str(), O_RDONLY);
484   } while (file_descriptor < 0 && errno == EINTR);
485   if (file_descriptor >= 0) {
486     io::FileInputStream* result = new io::FileInputStream(file_descriptor);
487     result->SetCloseOnDelete(true);
488     return result;
489   } else {
490     return NULL;
491   }
492 }
493 
494 }  // namespace compiler
495 }  // namespace protobuf
496 }  // namespace google
497