[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 1 | // Copyright (c) 2013 The Chromium Authors. All rights reserved. |
| 2 | // Use of this source code is governed by a BSD-style license that can be |
| 3 | // found in the LICENSE file. |
| 4 | |
| 5 | #include "tools/gn/filesystem_utils.h" |
| 6 | |
| 7 | #include "base/logging.h" |
[email protected] | f6b314d1 | 2013-08-23 21:07:47 | [diff] [blame^] | 8 | #include "base/strings/string_util.h" |
[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 9 | #include "base/strings/utf_string_conversions.h" |
| 10 | #include "build/build_config.h" |
| 11 | #include "tools/gn/location.h" |
| 12 | #include "tools/gn/source_dir.h" |
| 13 | |
| 14 | namespace { |
| 15 | |
| 16 | enum DotDisposition { |
| 17 | // The given dot is just part of a filename and is not special. |
| 18 | NOT_A_DIRECTORY, |
| 19 | |
| 20 | // The given dot is the current directory. |
| 21 | DIRECTORY_CUR, |
| 22 | |
| 23 | // The given dot is the first of a double dot that should take us up one. |
| 24 | DIRECTORY_UP |
| 25 | }; |
| 26 | |
| 27 | // When we find a dot, this function is called with the character following |
| 28 | // that dot to see what it is. The return value indicates what type this dot is |
| 29 | // (see above). This code handles the case where the dot is at the end of the |
| 30 | // input. |
| 31 | // |
| 32 | // |*consumed_len| will contain the number of characters in the input that |
| 33 | // express what we found. |
| 34 | DotDisposition ClassifyAfterDot(const std::string& path, |
| 35 | size_t after_dot, |
| 36 | size_t* consumed_len) { |
| 37 | if (after_dot == path.size()) { |
| 38 | // Single dot at the end. |
| 39 | *consumed_len = 1; |
| 40 | return DIRECTORY_CUR; |
| 41 | } |
| 42 | if (path[after_dot] == '/') { |
| 43 | // Single dot followed by a slash. |
| 44 | *consumed_len = 2; // Consume the slash |
| 45 | return DIRECTORY_CUR; |
| 46 | } |
| 47 | |
| 48 | if (path[after_dot] == '.') { |
| 49 | // Two dots. |
| 50 | if (after_dot + 1 == path.size()) { |
| 51 | // Double dot at the end. |
| 52 | *consumed_len = 2; |
| 53 | return DIRECTORY_UP; |
| 54 | } |
| 55 | if (path[after_dot + 1] == '/') { |
| 56 | // Double dot folowed by a slash. |
| 57 | *consumed_len = 3; |
| 58 | return DIRECTORY_UP; |
| 59 | } |
| 60 | } |
| 61 | |
| 62 | // The dots are followed by something else, not a directory. |
| 63 | *consumed_len = 1; |
| 64 | return NOT_A_DIRECTORY; |
| 65 | } |
| 66 | |
[email protected] | f6b314d1 | 2013-08-23 21:07:47 | [diff] [blame^] | 67 | #if defined(OS_WIN) |
| 68 | inline char NormalizeWindowsPathChar(char c) { |
| 69 | if (c == '/') |
| 70 | return '\\'; |
| 71 | return base::ToLowerASCII(c); |
| 72 | } |
| 73 | |
| 74 | // Attempts to do a case and slash-insensitive comparison of two 8-bit Windows |
| 75 | // paths. |
| 76 | bool AreAbsoluteWindowsPathsEqual(const base::StringPiece& a, |
| 77 | const base::StringPiece& b) { |
| 78 | if (a.size() != b.size()) |
| 79 | return false; |
| 80 | |
| 81 | // For now, just do a case-insensitive ASCII comparison. We could convert to |
| 82 | // UTF-16 and use ICU if necessary. Or maybe base::strcasecmp is good enough? |
| 83 | for (size_t i = 0; i < a.size(); i++) { |
| 84 | if (NormalizeWindowsPathChar(a[i]) != NormalizeWindowsPathChar(b[i])) |
| 85 | return false; |
| 86 | } |
| 87 | return true; |
| 88 | } |
| 89 | |
| 90 | bool DoesBeginWindowsDriveLetter(const base::StringPiece& path) { |
| 91 | if (path.size() < 3) |
| 92 | return false; |
| 93 | |
| 94 | // Check colon first, this will generally fail fastest. |
| 95 | if (path[1] != ':') |
| 96 | return false; |
| 97 | |
| 98 | // Check drive letter |
| 99 | if (!((path[0] >= 'A' && path[0] <= 'Z') || |
| 100 | path[0] >= 'a' && path[0] <= 'z')) |
| 101 | return false; |
| 102 | |
| 103 | if (path[2] != '/' && path[2] != '\\') |
| 104 | return false; |
| 105 | return true; |
| 106 | } |
| 107 | #endif |
| 108 | |
| 109 | } // namespace |
[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 110 | |
| 111 | SourceFileType GetSourceFileType(const SourceFile& file, |
| 112 | Settings::TargetOS os) { |
| 113 | base::StringPiece extension = FindExtension(&file.value()); |
| 114 | if (extension == "cc" || extension == "cpp" || extension == "cxx") |
| 115 | return SOURCE_CC; |
| 116 | if (extension == "h") |
| 117 | return SOURCE_H; |
| 118 | if (extension == "c") |
| 119 | return SOURCE_C; |
| 120 | |
| 121 | switch (os) { |
| 122 | case Settings::MAC: |
| 123 | if (extension == "m") |
| 124 | return SOURCE_M; |
| 125 | if (extension == "mm") |
| 126 | return SOURCE_MM; |
| 127 | break; |
| 128 | |
| 129 | case Settings::WIN: |
| 130 | if (extension == "rc") |
| 131 | return SOURCE_RC; |
[email protected] | c6f27f2 | 2013-08-21 21:44:59 | [diff] [blame] | 132 | // TODO(brettw) asm files. |
[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 133 | break; |
| 134 | |
| 135 | default: |
| 136 | break; |
| 137 | } |
| 138 | |
[email protected] | c6f27f2 | 2013-08-21 21:44:59 | [diff] [blame] | 139 | if (os != Settings::WIN) { |
| 140 | if (extension == "S") |
| 141 | return SOURCE_S; |
| 142 | } |
| 143 | |
[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 144 | return SOURCE_UNKNOWN; |
| 145 | } |
| 146 | |
| 147 | const char* GetExtensionForOutputType(Target::OutputType type, |
| 148 | Settings::TargetOS os) { |
| 149 | switch (os) { |
[email protected] | 8338eea | 2013-08-05 23:08:12 | [diff] [blame] | 150 | case Settings::MAC: |
| 151 | switch (type) { |
[email protected] | 8338eea | 2013-08-05 23:08:12 | [diff] [blame] | 152 | case Target::EXECUTABLE: |
| 153 | return ""; |
| 154 | case Target::SHARED_LIBRARY: |
| 155 | return "dylib"; |
| 156 | case Target::STATIC_LIBRARY: |
| 157 | return "a"; |
[email protected] | 8338eea | 2013-08-05 23:08:12 | [diff] [blame] | 158 | default: |
| 159 | NOTREACHED(); |
| 160 | } |
| 161 | break; |
| 162 | |
[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 163 | case Settings::WIN: |
| 164 | switch (type) { |
[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 165 | case Target::EXECUTABLE: |
| 166 | return "exe"; |
| 167 | case Target::SHARED_LIBRARY: |
| 168 | return "dll.lib"; // Extension of import library. |
| 169 | case Target::STATIC_LIBRARY: |
| 170 | return "lib"; |
[email protected] | c0822d7f | 2013-08-13 17:10:56 | [diff] [blame] | 171 | default: |
| 172 | NOTREACHED(); |
| 173 | } |
| 174 | break; |
| 175 | |
| 176 | case Settings::LINUX: |
| 177 | switch (type) { |
| 178 | case Target::EXECUTABLE: |
| 179 | return ""; |
| 180 | case Target::SHARED_LIBRARY: |
| 181 | return "so"; |
| 182 | case Target::STATIC_LIBRARY: |
| 183 | return "a"; |
[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 184 | default: |
| 185 | NOTREACHED(); |
| 186 | } |
| 187 | break; |
| 188 | |
| 189 | default: |
| 190 | NOTREACHED(); |
| 191 | } |
| 192 | return ""; |
| 193 | } |
| 194 | |
| 195 | std::string FilePathToUTF8(const base::FilePath& path) { |
| 196 | #if defined(OS_WIN) |
| 197 | return WideToUTF8(path.value()); |
| 198 | #else |
| 199 | return path.value(); |
| 200 | #endif |
| 201 | } |
| 202 | |
| 203 | base::FilePath UTF8ToFilePath(const base::StringPiece& sp) { |
| 204 | #if defined(OS_WIN) |
| 205 | return base::FilePath(UTF8ToWide(sp)); |
| 206 | #else |
| 207 | return base::FilePath(sp.as_string()); |
| 208 | #endif |
| 209 | } |
| 210 | |
| 211 | size_t FindExtensionOffset(const std::string& path) { |
| 212 | for (int i = static_cast<int>(path.size()); i >= 0; i--) { |
| 213 | if (path[i] == '/') |
| 214 | break; |
| 215 | if (path[i] == '.') |
| 216 | return i + 1; |
| 217 | } |
| 218 | return std::string::npos; |
| 219 | } |
| 220 | |
| 221 | base::StringPiece FindExtension(const std::string* path) { |
| 222 | size_t extension_offset = FindExtensionOffset(*path); |
| 223 | if (extension_offset == std::string::npos) |
| 224 | return base::StringPiece(); |
| 225 | return base::StringPiece(&path->data()[extension_offset], |
| 226 | path->size() - extension_offset); |
| 227 | } |
| 228 | |
| 229 | size_t FindFilenameOffset(const std::string& path) { |
| 230 | for (int i = static_cast<int>(path.size()) - 1; i >= 0; i--) { |
| 231 | if (path[i] == '/') |
| 232 | return i + 1; |
| 233 | } |
| 234 | return 0; // No filename found means everything was the filename. |
| 235 | } |
| 236 | |
| 237 | base::StringPiece FindFilename(const std::string* path) { |
| 238 | size_t filename_offset = FindFilenameOffset(*path); |
| 239 | if (filename_offset == 0) |
| 240 | return base::StringPiece(*path); // Everything is the file name. |
| 241 | return base::StringPiece(&(*path).data()[filename_offset], |
| 242 | path->size() - filename_offset); |
| 243 | } |
| 244 | |
| 245 | base::StringPiece FindFilenameNoExtension(const std::string* path) { |
| 246 | if (path->empty()) |
| 247 | return base::StringPiece(); |
| 248 | size_t filename_offset = FindFilenameOffset(*path); |
| 249 | size_t extension_offset = FindExtensionOffset(*path); |
| 250 | |
| 251 | size_t name_len; |
| 252 | if (extension_offset == std::string::npos) |
| 253 | name_len = path->size() - filename_offset; |
| 254 | else |
| 255 | name_len = extension_offset - filename_offset - 1; |
| 256 | |
| 257 | return base::StringPiece(&(*path).data()[filename_offset], name_len); |
| 258 | } |
| 259 | |
| 260 | void RemoveFilename(std::string* path) { |
| 261 | path->resize(FindFilenameOffset(*path)); |
| 262 | } |
| 263 | |
| 264 | bool EndsWithSlash(const std::string& s) { |
| 265 | return !s.empty() && s[s.size() - 1] == '/'; |
| 266 | } |
| 267 | |
| 268 | base::StringPiece FindDir(const std::string* path) { |
| 269 | size_t filename_offset = FindFilenameOffset(*path); |
| 270 | if (filename_offset == 0u) |
| 271 | return base::StringPiece(); |
| 272 | return base::StringPiece(path->data(), filename_offset); |
| 273 | } |
| 274 | |
| 275 | bool EnsureStringIsInOutputDir(const SourceDir& dir, |
| 276 | const std::string& str, |
| 277 | const Value& originating, |
| 278 | Err* err) { |
| 279 | // The last char of the dir will be a slash. We don't care if the input ends |
| 280 | // in a slash or not, so just compare up until there. |
| 281 | // |
| 282 | // This check will be wrong for all proper prefixes "e.g. "/output" will |
| 283 | // match "/out" but we don't really care since this is just a sanity check. |
| 284 | const std::string& dir_str = dir.value(); |
| 285 | if (str.compare(0, dir_str.length() - 1, dir_str, 0, dir_str.length() - 1) |
| 286 | != 0) { |
| 287 | *err = Err(originating, "File not inside output directory.", |
| 288 | "The given file should be in the output directory. Normally you would " |
| 289 | "specify\n\"$target_output_dir/foo\" or " |
| 290 | "\"$target_gen_dir/foo\". I interpreted this as\n\"" |
| 291 | + str + "\"."); |
| 292 | return false; |
| 293 | } |
| 294 | return true; |
| 295 | } |
| 296 | |
[email protected] | ac1128e | 2013-08-23 00:26:56 | [diff] [blame] | 297 | bool IsPathAbsolute(const base::StringPiece& path) { |
| 298 | if (path.empty()) |
| 299 | return false; |
| 300 | |
| 301 | if (path[0] != '/') { |
| 302 | #if defined(OS_WIN) |
| 303 | // Check for Windows system paths like "C:\foo". |
| 304 | if (path.size() > 2 && |
| 305 | path[1] == ':' && (path[2] == '/' || path[2] == '\\')) |
| 306 | return true; |
| 307 | #endif |
| 308 | return false; // Doesn't begin with a slash, is relative. |
| 309 | } |
| 310 | |
| 311 | if (path.size() > 1 && path[1] == '/') |
| 312 | return false; // Double slash at the beginning means source-relative. |
| 313 | |
| 314 | return true; |
| 315 | } |
| 316 | |
| 317 | bool MakeAbsolutePathRelativeIfPossible(const base::StringPiece& source_root, |
| 318 | const base::StringPiece& path, |
| 319 | std::string* dest) { |
| 320 | DCHECK(IsPathAbsolute(source_root)); |
| 321 | DCHECK(IsPathAbsolute(path)); |
| 322 | |
| 323 | dest->clear(); |
| 324 | |
| 325 | if (source_root.size() > path.size()) |
| 326 | return false; // The source root is longer: the path can never be inside. |
| 327 | |
| 328 | #if defined(OS_WIN) |
| 329 | // Source root should be canonical on Windows. |
| 330 | DCHECK(source_root.size() > 2 && source_root[0] != '/' && |
| 331 | source_root[1] == ':' && source_root[2] =='\\'); |
[email protected] | f6b314d1 | 2013-08-23 21:07:47 | [diff] [blame^] | 332 | |
| 333 | size_t after_common_index = std::string::npos; |
| 334 | if (DoesBeginWindowsDriveLetter(path)) { |
| 335 | // Handle "C:\foo" |
| 336 | if (AreAbsoluteWindowsPathsEqual(source_root, |
| 337 | path.substr(0, source_root.size()))) |
| 338 | after_common_index = source_root.size(); |
| 339 | else |
| 340 | return false; |
| 341 | } else if (path[0] == '/' && source_root.size() <= path.size() - 1 && |
| 342 | DoesBeginWindowsDriveLetter(path.substr(1))) { |
| 343 | // Handle "/C:/foo" |
| 344 | if (AreAbsoluteWindowsPathsEqual(source_root, |
| 345 | path.substr(1, source_root.size()))) |
| 346 | after_common_index = source_root.size() + 1; |
| 347 | else |
| 348 | return false; |
| 349 | } else { |
| 350 | return false; |
| 351 | } |
| 352 | |
| 353 | // If we get here, there's a match and after_common_index identifies the |
| 354 | // part after it. |
| 355 | |
| 356 | // The base may or may not have a trailing slash, so skip all slashes from |
| 357 | // the path after our prefix match. |
| 358 | size_t first_after_slash = after_common_index; |
| 359 | while (first_after_slash < path.size() && |
| 360 | (path[first_after_slash] == '/' || path[first_after_slash] == '\\')) |
| 361 | first_after_slash++; |
| 362 | |
| 363 | dest->assign("//"); // Result is source root relative. |
| 364 | dest->append(&path.data()[first_after_slash], |
| 365 | path.size() - first_after_slash); |
| 366 | return true; |
| 367 | |
[email protected] | ac1128e | 2013-08-23 00:26:56 | [diff] [blame] | 368 | #else |
[email protected] | f6b314d1 | 2013-08-23 21:07:47 | [diff] [blame^] | 369 | |
[email protected] | ac1128e | 2013-08-23 00:26:56 | [diff] [blame] | 370 | // On non-Windows this is easy. Since we know both are absolute, just do a |
| 371 | // prefix check. |
| 372 | if (path.substr(0, source_root.size()) == source_root) { |
[email protected] | ac1128e | 2013-08-23 00:26:56 | [diff] [blame] | 373 | // The base may or may not have a trailing slash, so skip all slashes from |
| 374 | // the path after our prefix match. |
| 375 | size_t first_after_slash = source_root.size(); |
| 376 | while (first_after_slash < path.size() && path[first_after_slash] == '/') |
| 377 | first_after_slash++; |
| 378 | |
[email protected] | f6b314d1 | 2013-08-23 21:07:47 | [diff] [blame^] | 379 | dest->assign("//"); // Result is source root relative. |
[email protected] | ac1128e | 2013-08-23 00:26:56 | [diff] [blame] | 380 | dest->append(&path.data()[first_after_slash], |
| 381 | path.size() - first_after_slash); |
| 382 | return true; |
| 383 | } |
[email protected] | ac1128e | 2013-08-23 00:26:56 | [diff] [blame] | 384 | return false; |
[email protected] | f6b314d1 | 2013-08-23 21:07:47 | [diff] [blame^] | 385 | #endif |
[email protected] | ac1128e | 2013-08-23 00:26:56 | [diff] [blame] | 386 | } |
| 387 | |
[email protected] | 96ea63d | 2013-07-30 10:17:07 | [diff] [blame] | 388 | std::string InvertDir(const SourceDir& path) { |
| 389 | const std::string value = path.value(); |
| 390 | if (value.empty()) |
| 391 | return std::string(); |
| 392 | |
| 393 | DCHECK(value[0] == '/'); |
| 394 | size_t begin_index = 1; |
| 395 | |
| 396 | // If the input begins with two slashes, skip over both (this is a |
| 397 | // source-relative dir). |
| 398 | if (value.size() > 1 && value[1] == '/') |
| 399 | begin_index = 2; |
| 400 | |
| 401 | std::string ret; |
| 402 | for (size_t i = begin_index; i < value.size(); i++) { |
| 403 | if (value[i] == '/') |
| 404 | ret.append("../"); |
| 405 | } |
| 406 | return ret; |
| 407 | } |
| 408 | |
| 409 | void NormalizePath(std::string* path) { |
| 410 | char* pathbuf = path->empty() ? NULL : &(*path)[0]; |
| 411 | |
| 412 | // top_index is the first character we can modify in the path. Anything |
| 413 | // before this indicates where the path is relative to. |
| 414 | size_t top_index = 0; |
| 415 | bool is_relative = true; |
| 416 | if (!path->empty() && pathbuf[0] == '/') { |
| 417 | is_relative = false; |
| 418 | |
| 419 | if (path->size() > 1 && pathbuf[1] == '/') { |
| 420 | // Two leading slashes, this is a path into the source dir. |
| 421 | top_index = 2; |
| 422 | } else { |
| 423 | // One leading slash, this is a system-absolute path. |
| 424 | top_index = 1; |
| 425 | } |
| 426 | } |
| 427 | |
| 428 | size_t dest_i = top_index; |
| 429 | for (size_t src_i = top_index; src_i < path->size(); /* nothing */) { |
| 430 | if (pathbuf[src_i] == '.') { |
| 431 | if (src_i == 0 || pathbuf[src_i - 1] == '/') { |
| 432 | // Slash followed by a dot, see if it's something special. |
| 433 | size_t consumed_len; |
| 434 | switch (ClassifyAfterDot(*path, src_i + 1, &consumed_len)) { |
| 435 | case NOT_A_DIRECTORY: |
| 436 | // Copy the dot to the output, it means nothing special. |
| 437 | pathbuf[dest_i++] = pathbuf[src_i++]; |
| 438 | break; |
| 439 | case DIRECTORY_CUR: |
| 440 | // Current directory, just skip the input. |
| 441 | src_i += consumed_len; |
| 442 | break; |
| 443 | case DIRECTORY_UP: |
| 444 | // Back up over previous directory component. If we're already |
| 445 | // at the top, preserve the "..". |
| 446 | if (dest_i > top_index) { |
| 447 | // The previous char was a slash, remove it. |
| 448 | dest_i--; |
| 449 | } |
| 450 | |
| 451 | if (dest_i == top_index) { |
| 452 | if (is_relative) { |
| 453 | // We're already at the beginning of a relative input, copy the |
| 454 | // ".." and continue. We need the trailing slash if there was |
| 455 | // one before (otherwise we're at the end of the input). |
| 456 | pathbuf[dest_i++] = '.'; |
| 457 | pathbuf[dest_i++] = '.'; |
| 458 | if (consumed_len == 3) |
| 459 | pathbuf[dest_i++] = '/'; |
| 460 | |
| 461 | // This also makes a new "root" that we can't delete by going |
| 462 | // up more levels. Otherwise "../.." would collapse to |
| 463 | // nothing. |
| 464 | top_index = dest_i; |
| 465 | } |
| 466 | // Otherwise we're at the beginning of an absolute path. Don't |
| 467 | // allow ".." to go up another level and just eat it. |
| 468 | } else { |
| 469 | // Just find the previous slash or the beginning of input. |
| 470 | while (dest_i > 0 && pathbuf[dest_i - 1] != '/') |
| 471 | dest_i--; |
| 472 | } |
| 473 | src_i += consumed_len; |
| 474 | } |
| 475 | } else { |
| 476 | // Dot not preceeded by a slash, copy it literally. |
| 477 | pathbuf[dest_i++] = pathbuf[src_i++]; |
| 478 | } |
| 479 | } else if (pathbuf[src_i] == '/') { |
| 480 | if (src_i > 0 && pathbuf[src_i - 1] == '/') { |
| 481 | // Two slashes in a row, skip over it. |
| 482 | src_i++; |
| 483 | } else { |
| 484 | // Just one slash, copy it. |
| 485 | pathbuf[dest_i++] = pathbuf[src_i++]; |
| 486 | } |
| 487 | } else { |
| 488 | // Input nothing special, just copy it. |
| 489 | pathbuf[dest_i++] = pathbuf[src_i++]; |
| 490 | } |
| 491 | } |
| 492 | path->resize(dest_i); |
| 493 | } |
| 494 | |
| 495 | void ConvertPathToSystem(std::string* path) { |
| 496 | #if defined(OS_WIN) |
| 497 | for (size_t i = 0; i < path->size(); i++) { |
| 498 | if ((*path)[i] == '/') |
| 499 | (*path)[i] = '\\'; |
| 500 | } |
| 501 | #endif |
| 502 | } |
| 503 | |
| 504 | std::string PathToSystem(const std::string& path) { |
| 505 | std::string ret(path); |
| 506 | ConvertPathToSystem(&ret); |
| 507 | return ret; |
| 508 | } |
| 509 | |