xref: /freebsd/sys/contrib/zstd/programs/util.c (revision fe815331bb40604ba31312acf7e4619674631777)
1 /*
2  * Copyright (c) 2016-2020, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
3  * All rights reserved.
4  *
5  * This source code is licensed under both the BSD-style license (found in the
6  * LICENSE file in the root directory of this source tree) and the GPLv2 (found
7  * in the COPYING file in the root directory of this source tree).
8  * You may select, at your option, one of the above-listed licenses.
9  */
10 
11 #if defined (__cplusplus)
12 extern "C" {
13 #endif
14 
15 
16 /*-****************************************
17 *  Dependencies
18 ******************************************/
19 #include "util.h"       /* note : ensure that platform.h is included first ! */
20 #include <stdlib.h>     /* malloc, realloc, free */
21 #include <stdio.h>      /* fprintf */
22 #include <time.h>       /* clock_t, clock, CLOCKS_PER_SEC, nanosleep */
23 #include <errno.h>
24 #include <assert.h>
25 
26 #if defined(_WIN32)
27 #  include <sys/utime.h>  /* utime */
28 #  include <io.h>         /* _chmod */
29 #else
30 #  include <unistd.h>     /* chown, stat */
31 #  if PLATFORM_POSIX_VERSION < 200809L
32 #    include <utime.h>    /* utime */
33 #  else
34 #    include <fcntl.h>    /* AT_FDCWD */
35 #    include <sys/stat.h> /* utimensat */
36 #  endif
37 #endif
38 
39 #if defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__)
40 #include <direct.h>     /* needed for _mkdir in windows */
41 #endif
42 
43 #if defined(__linux__) || (PLATFORM_POSIX_VERSION >= 200112L)  /* opendir, readdir require POSIX.1-2001 */
44 #  include <dirent.h>       /* opendir, readdir */
45 #  include <string.h>       /* strerror, memcpy */
46 #endif /* #ifdef _WIN32 */
47 
48 
49 /*-****************************************
50 *  Internal Macros
51 ******************************************/
52 
53 /* CONTROL is almost like an assert(), but is never disabled.
54  * It's designed for failures that may happen rarely,
55  * but we don't want to maintain a specific error code path for them,
56  * such as a malloc() returning NULL for example.
57  * Since it's always active, this macro can trigger side effects.
58  */
59 #define CONTROL(c)  {         \
60     if (!(c)) {               \
61         UTIL_DISPLAYLEVEL(1, "Error : %s, %i : %s",  \
62                           __FILE__, __LINE__, #c);   \
63         exit(1);              \
64 }   }
65 
66 /* console log */
67 #define UTIL_DISPLAY(...)         fprintf(stderr, __VA_ARGS__)
68 #define UTIL_DISPLAYLEVEL(l, ...) { if (g_utilDisplayLevel>=l) { UTIL_DISPLAY(__VA_ARGS__); } }
69 
70 /* A modified version of realloc().
71  * If UTIL_realloc() fails the original block is freed.
72  */
73 UTIL_STATIC void* UTIL_realloc(void *ptr, size_t size)
74 {
75     void *newptr = realloc(ptr, size);
76     if (newptr) return newptr;
77     free(ptr);
78     return NULL;
79 }
80 
81 #if defined(_MSC_VER)
82     #define chmod _chmod
83 #endif
84 
85 
86 /*-****************************************
87 *  Console log
88 ******************************************/
89 int g_utilDisplayLevel;
90 
91 
92 /*-*************************************
93 *  Constants
94 ***************************************/
95 #define LIST_SIZE_INCREASE   (8*1024)
96 #define MAX_FILE_OF_FILE_NAMES_SIZE (1<<20)*50
97 
98 
99 /*-*************************************
100 *  Functions
101 ***************************************/
102 
103 int UTIL_fileExist(const char* filename)
104 {
105     stat_t statbuf;
106 #if defined(_MSC_VER)
107     int const stat_error = _stat64(filename, &statbuf);
108 #else
109     int const stat_error = stat(filename, &statbuf);
110 #endif
111     return !stat_error;
112 }
113 
114 int UTIL_isRegularFile(const char* infilename)
115 {
116     stat_t statbuf;
117     return UTIL_getFileStat(infilename, &statbuf); /* Only need to know whether it is a regular file */
118 }
119 
120 int UTIL_getFileStat(const char* infilename, stat_t *statbuf)
121 {
122     int r;
123 #if defined(_MSC_VER)
124     r = _stat64(infilename, statbuf);
125     if (r || !(statbuf->st_mode & S_IFREG)) return 0;   /* No good... */
126 #else
127     r = stat(infilename, statbuf);
128     if (r || !S_ISREG(statbuf->st_mode)) return 0;   /* No good... */
129 #endif
130     return 1;
131 }
132 
133 /* like chmod, but avoid changing permission of /dev/null */
134 int UTIL_chmod(char const* filename, mode_t permissions)
135 {
136     if (!strcmp(filename, "/dev/null")) return 0;   /* pretend success, but don't change anything */
137     return chmod(filename, permissions);
138 }
139 
140 int UTIL_setFileStat(const char *filename, stat_t *statbuf)
141 {
142     int res = 0;
143 
144     if (!UTIL_isRegularFile(filename))
145         return -1;
146 
147     /* set access and modification times */
148     /* We check that st_mtime is a macro here in order to give us confidence
149      * that struct stat has a struct timespec st_mtim member. We need this
150      * check because there are some platforms that claim to be POSIX 2008
151      * compliant but which do not have st_mtim... */
152 #if (PLATFORM_POSIX_VERSION >= 200809L) && defined(st_mtime)
153     {
154         /* (atime, mtime) */
155         struct timespec timebuf[2] = { {0, UTIME_NOW} };
156         timebuf[1] = statbuf->st_mtim;
157         res += utimensat(AT_FDCWD, filename, timebuf, 0);
158     }
159 #else
160     {
161         struct utimbuf timebuf;
162         timebuf.actime = time(NULL);
163         timebuf.modtime = statbuf->st_mtime;
164         res += utime(filename, &timebuf);
165     }
166 #endif
167 
168 #if !defined(_WIN32)
169     res += chown(filename, statbuf->st_uid, statbuf->st_gid);  /* Copy ownership */
170 #endif
171 
172     res += UTIL_chmod(filename, statbuf->st_mode & 07777);  /* Copy file permissions */
173 
174     errno = 0;
175     return -res; /* number of errors is returned */
176 }
177 
178 int UTIL_isDirectory(const char* infilename)
179 {
180     stat_t statbuf;
181 #if defined(_MSC_VER)
182     int const r = _stat64(infilename, &statbuf);
183     if (!r && (statbuf.st_mode & _S_IFDIR)) return 1;
184 #else
185     int const r = stat(infilename, &statbuf);
186     if (!r && S_ISDIR(statbuf.st_mode)) return 1;
187 #endif
188     return 0;
189 }
190 
191 int UTIL_compareStr(const void *p1, const void *p2) {
192     return strcmp(* (char * const *) p1, * (char * const *) p2);
193 }
194 
195 int UTIL_isSameFile(const char* fName1, const char* fName2)
196 {
197     assert(fName1 != NULL); assert(fName2 != NULL);
198 #if defined(_MSC_VER) || defined(_WIN32)
199     /* note : Visual does not support file identification by inode.
200      *        inode does not work on Windows, even with a posix layer, like msys2.
201      *        The following work-around is limited to detecting exact name repetition only,
202      *        aka `filename` is considered different from `subdir/../filename` */
203     return !strcmp(fName1, fName2);
204 #else
205     {   stat_t file1Stat;
206         stat_t file2Stat;
207         return UTIL_getFileStat(fName1, &file1Stat)
208             && UTIL_getFileStat(fName2, &file2Stat)
209             && (file1Stat.st_dev == file2Stat.st_dev)
210             && (file1Stat.st_ino == file2Stat.st_ino);
211     }
212 #endif
213 }
214 
215 /* UTIL_isFIFO : distinguish named pipes */
216 int UTIL_isFIFO(const char* infilename)
217 {
218 /* macro guards, as defined in : https://linux.die.net/man/2/lstat */
219 #if PLATFORM_POSIX_VERSION >= 200112L
220     stat_t statbuf;
221     int const r = UTIL_getFileStat(infilename, &statbuf);
222     if (!r && S_ISFIFO(statbuf.st_mode)) return 1;
223 #endif
224     (void)infilename;
225     return 0;
226 }
227 
228 int UTIL_isLink(const char* infilename)
229 {
230 /* macro guards, as defined in : https://linux.die.net/man/2/lstat */
231 #if PLATFORM_POSIX_VERSION >= 200112L
232     stat_t statbuf;
233     int const r = lstat(infilename, &statbuf);
234     if (!r && S_ISLNK(statbuf.st_mode)) return 1;
235 #endif
236     (void)infilename;
237     return 0;
238 }
239 
240 U64 UTIL_getFileSize(const char* infilename)
241 {
242     if (!UTIL_isRegularFile(infilename)) return UTIL_FILESIZE_UNKNOWN;
243     {   int r;
244 #if defined(_MSC_VER)
245         struct __stat64 statbuf;
246         r = _stat64(infilename, &statbuf);
247         if (r || !(statbuf.st_mode & S_IFREG)) return UTIL_FILESIZE_UNKNOWN;
248 #elif defined(__MINGW32__) && defined (__MSVCRT__)
249         struct _stati64 statbuf;
250         r = _stati64(infilename, &statbuf);
251         if (r || !(statbuf.st_mode & S_IFREG)) return UTIL_FILESIZE_UNKNOWN;
252 #else
253         struct stat statbuf;
254         r = stat(infilename, &statbuf);
255         if (r || !S_ISREG(statbuf.st_mode)) return UTIL_FILESIZE_UNKNOWN;
256 #endif
257         return (U64)statbuf.st_size;
258     }
259 }
260 
261 
262 U64 UTIL_getTotalFileSize(const char* const * fileNamesTable, unsigned nbFiles)
263 {
264     U64 total = 0;
265     unsigned n;
266     for (n=0; n<nbFiles; n++) {
267         U64 const size = UTIL_getFileSize(fileNamesTable[n]);
268         if (size == UTIL_FILESIZE_UNKNOWN) return UTIL_FILESIZE_UNKNOWN;
269         total += size;
270     }
271     return total;
272 }
273 
274 
275 /* condition : @file must be valid, and not have reached its end.
276  * @return : length of line written into @buf, ended with `\0` instead of '\n',
277  *           or 0, if there is no new line */
278 static size_t readLineFromFile(char* buf, size_t len, FILE* file)
279 {
280     assert(!feof(file));
281     /* Work around Cygwin problem when len == 1 it returns NULL. */
282     if (len <= 1) return 0;
283     CONTROL( fgets(buf, (int) len, file) );
284     {   size_t linelen = strlen(buf);
285         if (strlen(buf)==0) return 0;
286         if (buf[linelen-1] == '\n') linelen--;
287         buf[linelen] = '\0';
288         return linelen+1;
289     }
290 }
291 
292 /* Conditions :
293  *   size of @inputFileName file must be < @dstCapacity
294  *   @dst must be initialized
295  * @return : nb of lines
296  *       or -1 if there's an error
297  */
298 static int
299 readLinesFromFile(void* dst, size_t dstCapacity,
300             const char* inputFileName)
301 {
302     int nbFiles = 0;
303     size_t pos = 0;
304     char* const buf = (char*)dst;
305     FILE* const inputFile = fopen(inputFileName, "r");
306 
307     assert(dst != NULL);
308 
309     if(!inputFile) {
310         if (g_utilDisplayLevel >= 1) perror("zstd:util:readLinesFromFile");
311         return -1;
312     }
313 
314     while ( !feof(inputFile) ) {
315         size_t const lineLength = readLineFromFile(buf+pos, dstCapacity-pos, inputFile);
316         if (lineLength == 0) break;
317         assert(pos + lineLength < dstCapacity);
318         pos += lineLength;
319         ++nbFiles;
320     }
321 
322     CONTROL( fclose(inputFile) == 0 );
323 
324     return nbFiles;
325 }
326 
327 /*Note: buf is not freed in case function successfully created table because filesTable->fileNames[0] = buf*/
328 FileNamesTable*
329 UTIL_createFileNamesTable_fromFileName(const char* inputFileName)
330 {
331     size_t nbFiles = 0;
332     char* buf;
333     size_t bufSize;
334     size_t pos = 0;
335 
336     if (!UTIL_fileExist(inputFileName) || !UTIL_isRegularFile(inputFileName))
337         return NULL;
338 
339     {   U64 const inputFileSize = UTIL_getFileSize(inputFileName);
340         if(inputFileSize > MAX_FILE_OF_FILE_NAMES_SIZE)
341             return NULL;
342         bufSize = (size_t)(inputFileSize + 1); /* (+1) to add '\0' at the end of last filename */
343     }
344 
345     buf = (char*) malloc(bufSize);
346     CONTROL( buf != NULL );
347 
348     {   int const ret_nbFiles = readLinesFromFile(buf, bufSize, inputFileName);
349 
350         if (ret_nbFiles <= 0) {
351           free(buf);
352           return NULL;
353         }
354         nbFiles = (size_t)ret_nbFiles;
355     }
356 
357     {   const char** filenamesTable = (const char**) malloc(nbFiles * sizeof(*filenamesTable));
358         CONTROL(filenamesTable != NULL);
359 
360         {   size_t fnb;
361             for (fnb = 0, pos = 0; fnb < nbFiles; fnb++) {
362                 filenamesTable[fnb] = buf+pos;
363                 pos += strlen(buf+pos)+1;  /* +1 for the finishing `\0` */
364         }   }
365         assert(pos <= bufSize);
366 
367         return UTIL_assembleFileNamesTable(filenamesTable, nbFiles, buf);
368     }
369 }
370 
371 static FileNamesTable*
372 UTIL_assembleFileNamesTable2(const char** filenames, size_t tableSize, size_t tableCapacity, char* buf)
373 {
374     FileNamesTable* const table = (FileNamesTable*) malloc(sizeof(*table));
375     CONTROL(table != NULL);
376     table->fileNames = filenames;
377     table->buf = buf;
378     table->tableSize = tableSize;
379     table->tableCapacity = tableCapacity;
380     return table;
381 }
382 
383 FileNamesTable*
384 UTIL_assembleFileNamesTable(const char** filenames, size_t tableSize, char* buf)
385 {
386     return UTIL_assembleFileNamesTable2(filenames, tableSize, tableSize, buf);
387 }
388 
389 void UTIL_freeFileNamesTable(FileNamesTable* table)
390 {
391     if (table==NULL) return;
392     free((void*)table->fileNames);
393     free(table->buf);
394     free(table);
395 }
396 
397 FileNamesTable* UTIL_allocateFileNamesTable(size_t tableSize)
398 {
399     const char** const fnTable = (const char**)malloc(tableSize * sizeof(*fnTable));
400     FileNamesTable* fnt;
401     if (fnTable==NULL) return NULL;
402     fnt = UTIL_assembleFileNamesTable(fnTable, tableSize, NULL);
403     fnt->tableSize = 0;   /* the table is empty */
404     return fnt;
405 }
406 
407 void UTIL_refFilename(FileNamesTable* fnt, const char* filename)
408 {
409     assert(fnt->tableSize < fnt->tableCapacity);
410     fnt->fileNames[fnt->tableSize] = filename;
411     fnt->tableSize++;
412 }
413 
414 static size_t getTotalTableSize(FileNamesTable* table)
415 {
416     size_t fnb = 0, totalSize = 0;
417     for(fnb = 0 ; fnb < table->tableSize && table->fileNames[fnb] ; ++fnb) {
418         totalSize += strlen(table->fileNames[fnb]) + 1; /* +1 to add '\0' at the end of each fileName */
419     }
420     return totalSize;
421 }
422 
423 FileNamesTable*
424 UTIL_mergeFileNamesTable(FileNamesTable* table1, FileNamesTable* table2)
425 {
426     unsigned newTableIdx = 0;
427     size_t pos = 0;
428     size_t newTotalTableSize;
429     char* buf;
430 
431     FileNamesTable* const newTable = UTIL_assembleFileNamesTable(NULL, 0, NULL);
432     CONTROL( newTable != NULL );
433 
434     newTotalTableSize = getTotalTableSize(table1) + getTotalTableSize(table2);
435 
436     buf = (char*) calloc(newTotalTableSize, sizeof(*buf));
437     CONTROL ( buf != NULL );
438 
439     newTable->buf = buf;
440     newTable->tableSize = table1->tableSize + table2->tableSize;
441     newTable->fileNames = (const char **) calloc(newTable->tableSize, sizeof(*(newTable->fileNames)));
442     CONTROL ( newTable->fileNames != NULL );
443 
444     {   unsigned idx1;
445         for( idx1=0 ; (idx1 < table1->tableSize) && table1->fileNames[idx1] && (pos < newTotalTableSize); ++idx1, ++newTableIdx) {
446             size_t const curLen = strlen(table1->fileNames[idx1]);
447             memcpy(buf+pos, table1->fileNames[idx1], curLen);
448             assert(newTableIdx <= newTable->tableSize);
449             newTable->fileNames[newTableIdx] = buf+pos;
450             pos += curLen+1;
451     }   }
452 
453     {   unsigned idx2;
454         for( idx2=0 ; (idx2 < table2->tableSize) && table2->fileNames[idx2] && (pos < newTotalTableSize) ; ++idx2, ++newTableIdx) {
455             size_t const curLen = strlen(table2->fileNames[idx2]);
456             memcpy(buf+pos, table2->fileNames[idx2], curLen);
457             assert(newTableIdx <= newTable->tableSize);
458             newTable->fileNames[newTableIdx] = buf+pos;
459             pos += curLen+1;
460     }   }
461     assert(pos <= newTotalTableSize);
462     newTable->tableSize = newTableIdx;
463 
464     UTIL_freeFileNamesTable(table1);
465     UTIL_freeFileNamesTable(table2);
466 
467     return newTable;
468 }
469 
470 #ifdef _WIN32
471 static int UTIL_prepareFileList(const char* dirName,
472                                 char** bufStart, size_t* pos,
473                                 char** bufEnd, int followLinks)
474 {
475     char* path;
476     size_t dirLength, pathLength;
477     int nbFiles = 0;
478     WIN32_FIND_DATAA cFile;
479     HANDLE hFile;
480 
481     dirLength = strlen(dirName);
482     path = (char*) malloc(dirLength + 3);
483     if (!path) return 0;
484 
485     memcpy(path, dirName, dirLength);
486     path[dirLength] = '\\';
487     path[dirLength+1] = '*';
488     path[dirLength+2] = 0;
489 
490     hFile=FindFirstFileA(path, &cFile);
491     if (hFile == INVALID_HANDLE_VALUE) {
492         UTIL_DISPLAYLEVEL(1, "Cannot open directory '%s'\n", dirName);
493         return 0;
494     }
495     free(path);
496 
497     do {
498         size_t const fnameLength = strlen(cFile.cFileName);
499         path = (char*) malloc(dirLength + fnameLength + 2);
500         if (!path) { FindClose(hFile); return 0; }
501         memcpy(path, dirName, dirLength);
502         path[dirLength] = '\\';
503         memcpy(path+dirLength+1, cFile.cFileName, fnameLength);
504         pathLength = dirLength+1+fnameLength;
505         path[pathLength] = 0;
506         if (cFile.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) {
507             if ( strcmp (cFile.cFileName, "..") == 0
508               || strcmp (cFile.cFileName, ".") == 0 )
509                 continue;
510             /* Recursively call "UTIL_prepareFileList" with the new path. */
511             nbFiles += UTIL_prepareFileList(path, bufStart, pos, bufEnd, followLinks);
512             if (*bufStart == NULL) { free(path); FindClose(hFile); return 0; }
513         } else if ( (cFile.dwFileAttributes & FILE_ATTRIBUTE_NORMAL)
514                  || (cFile.dwFileAttributes & FILE_ATTRIBUTE_ARCHIVE)
515                  || (cFile.dwFileAttributes & FILE_ATTRIBUTE_COMPRESSED) ) {
516             if (*bufStart + *pos + pathLength >= *bufEnd) {
517                 ptrdiff_t const newListSize = (*bufEnd - *bufStart) + LIST_SIZE_INCREASE;
518                 *bufStart = (char*)UTIL_realloc(*bufStart, newListSize);
519                 if (*bufStart == NULL) { free(path); FindClose(hFile); return 0; }
520                 *bufEnd = *bufStart + newListSize;
521             }
522             if (*bufStart + *pos + pathLength < *bufEnd) {
523                 memcpy(*bufStart + *pos, path, pathLength+1 /* include final \0 */);
524                 *pos += pathLength + 1;
525                 nbFiles++;
526         }   }
527         free(path);
528     } while (FindNextFileA(hFile, &cFile));
529 
530     FindClose(hFile);
531     return nbFiles;
532 }
533 
534 #elif defined(__linux__) || (PLATFORM_POSIX_VERSION >= 200112L)  /* opendir, readdir require POSIX.1-2001 */
535 
536 static int UTIL_prepareFileList(const char *dirName,
537                                 char** bufStart, size_t* pos,
538                                 char** bufEnd, int followLinks)
539 {
540     DIR* dir;
541     struct dirent * entry;
542     size_t dirLength;
543     int nbFiles = 0;
544 
545     if (!(dir = opendir(dirName))) {
546         UTIL_DISPLAYLEVEL(1, "Cannot open directory '%s': %s\n", dirName, strerror(errno));
547         return 0;
548     }
549 
550     dirLength = strlen(dirName);
551     errno = 0;
552     while ((entry = readdir(dir)) != NULL) {
553         char* path;
554         size_t fnameLength, pathLength;
555         if (strcmp (entry->d_name, "..") == 0 ||
556             strcmp (entry->d_name, ".") == 0) continue;
557         fnameLength = strlen(entry->d_name);
558         path = (char*) malloc(dirLength + fnameLength + 2);
559         if (!path) { closedir(dir); return 0; }
560         memcpy(path, dirName, dirLength);
561 
562         path[dirLength] = '/';
563         memcpy(path+dirLength+1, entry->d_name, fnameLength);
564         pathLength = dirLength+1+fnameLength;
565         path[pathLength] = 0;
566 
567         if (!followLinks && UTIL_isLink(path)) {
568             UTIL_DISPLAYLEVEL(2, "Warning : %s is a symbolic link, ignoring\n", path);
569             free(path);
570             continue;
571         }
572 
573         if (UTIL_isDirectory(path)) {
574             nbFiles += UTIL_prepareFileList(path, bufStart, pos, bufEnd, followLinks);  /* Recursively call "UTIL_prepareFileList" with the new path. */
575             if (*bufStart == NULL) { free(path); closedir(dir); return 0; }
576         } else {
577             if (*bufStart + *pos + pathLength >= *bufEnd) {
578                 ptrdiff_t newListSize = (*bufEnd - *bufStart) + LIST_SIZE_INCREASE;
579                 assert(newListSize >= 0);
580                 *bufStart = (char*)UTIL_realloc(*bufStart, (size_t)newListSize);
581                 *bufEnd = *bufStart + newListSize;
582                 if (*bufStart == NULL) { free(path); closedir(dir); return 0; }
583             }
584             if (*bufStart + *pos + pathLength < *bufEnd) {
585                 memcpy(*bufStart + *pos, path, pathLength + 1);  /* with final \0 */
586                 *pos += pathLength + 1;
587                 nbFiles++;
588         }   }
589         free(path);
590         errno = 0; /* clear errno after UTIL_isDirectory, UTIL_prepareFileList */
591     }
592 
593     if (errno != 0) {
594         UTIL_DISPLAYLEVEL(1, "readdir(%s) error: %s \n", dirName, strerror(errno));
595         free(*bufStart);
596         *bufStart = NULL;
597     }
598     closedir(dir);
599     return nbFiles;
600 }
601 
602 #else
603 
604 static int UTIL_prepareFileList(const char *dirName,
605                                 char** bufStart, size_t* pos,
606                                 char** bufEnd, int followLinks)
607 {
608     (void)bufStart; (void)bufEnd; (void)pos; (void)followLinks;
609     UTIL_DISPLAYLEVEL(1, "Directory %s ignored (compiled without _WIN32 or _POSIX_C_SOURCE) \n", dirName);
610     return 0;
611 }
612 
613 #endif /* #ifdef _WIN32 */
614 
615 int UTIL_isCompressedFile(const char *inputName, const char *extensionList[])
616 {
617   const char* ext = UTIL_getFileExtension(inputName);
618   while(*extensionList!=NULL)
619   {
620     const int isCompressedExtension = strcmp(ext,*extensionList);
621     if(isCompressedExtension==0)
622       return 1;
623     ++extensionList;
624   }
625    return 0;
626 }
627 
628 /*Utility function to get file extension from file */
629 const char* UTIL_getFileExtension(const char* infilename)
630 {
631    const char* extension = strrchr(infilename, '.');
632    if(!extension || extension==infilename) return "";
633    return extension;
634 }
635 
636 
637 FileNamesTable*
638 UTIL_createExpandedFNT(const char** inputNames, size_t nbIfns, int followLinks)
639 {
640     unsigned nbFiles;
641     char* buf = (char*)malloc(LIST_SIZE_INCREASE);
642     char* bufend = buf + LIST_SIZE_INCREASE;
643 
644     if (!buf) return NULL;
645 
646     {   size_t ifnNb, pos;
647         for (ifnNb=0, pos=0, nbFiles=0; ifnNb<nbIfns; ifnNb++) {
648             if (!UTIL_isDirectory(inputNames[ifnNb])) {
649                 size_t const len = strlen(inputNames[ifnNb]);
650                 if (buf + pos + len >= bufend) {
651                     ptrdiff_t newListSize = (bufend - buf) + LIST_SIZE_INCREASE;
652                     assert(newListSize >= 0);
653                     buf = (char*)UTIL_realloc(buf, (size_t)newListSize);
654                     if (!buf) return NULL;
655                     bufend = buf + newListSize;
656                 }
657                 if (buf + pos + len < bufend) {
658                     memcpy(buf+pos, inputNames[ifnNb], len+1);  /* including final \0 */
659                     pos += len + 1;
660                     nbFiles++;
661                 }
662             } else {
663                 nbFiles += (unsigned)UTIL_prepareFileList(inputNames[ifnNb], &buf, &pos, &bufend, followLinks);
664                 if (buf == NULL) return NULL;
665     }   }   }
666 
667     /* note : even if nbFiles==0, function returns a valid, though empty, FileNamesTable* object */
668 
669     {   size_t ifnNb, pos;
670         size_t const fntCapacity = nbFiles + 1;  /* minimum 1, allows adding one reference, typically stdin */
671         const char** const fileNamesTable = (const char**)malloc(fntCapacity * sizeof(*fileNamesTable));
672         if (!fileNamesTable) { free(buf); return NULL; }
673 
674         for (ifnNb = 0, pos = 0; ifnNb < nbFiles; ifnNb++) {
675             fileNamesTable[ifnNb] = buf + pos;
676             if (buf + pos > bufend) { free(buf); free((void*)fileNamesTable); return NULL; }
677             pos += strlen(fileNamesTable[ifnNb]) + 1;
678         }
679         return UTIL_assembleFileNamesTable2(fileNamesTable, nbFiles, fntCapacity, buf);
680     }
681 }
682 
683 
684 void UTIL_expandFNT(FileNamesTable** fnt, int followLinks)
685 {
686     FileNamesTable* const newFNT = UTIL_createExpandedFNT((*fnt)->fileNames, (*fnt)->tableSize, followLinks);
687     CONTROL(newFNT != NULL);
688     UTIL_freeFileNamesTable(*fnt);
689     *fnt = newFNT;
690 }
691 
692 FileNamesTable* UTIL_createFNT_fromROTable(const char** filenames, size_t nbFilenames)
693 {
694     size_t const sizeof_FNTable = nbFilenames * sizeof(*filenames);
695     const char** const newFNTable = (const char**)malloc(sizeof_FNTable);
696     if (newFNTable==NULL) return NULL;
697     memcpy((void*)newFNTable, filenames, sizeof_FNTable);  /* void* : mitigate a Visual compiler bug or limitation */
698     return UTIL_assembleFileNamesTable(newFNTable, nbFilenames, NULL);
699 }
700 
701 
702 /*-****************************************
703 *  count the number of physical cores
704 ******************************************/
705 
706 #if defined(_WIN32) || defined(WIN32)
707 
708 #include <windows.h>
709 
710 typedef BOOL(WINAPI* LPFN_GLPI)(PSYSTEM_LOGICAL_PROCESSOR_INFORMATION, PDWORD);
711 
712 int UTIL_countPhysicalCores(void)
713 {
714     static int numPhysicalCores = 0;
715     if (numPhysicalCores != 0) return numPhysicalCores;
716 
717     {   LPFN_GLPI glpi;
718         BOOL done = FALSE;
719         PSYSTEM_LOGICAL_PROCESSOR_INFORMATION buffer = NULL;
720         PSYSTEM_LOGICAL_PROCESSOR_INFORMATION ptr = NULL;
721         DWORD returnLength = 0;
722         size_t byteOffset = 0;
723 
724 #if defined(_MSC_VER)
725 /* Visual Studio does not like the following cast */
726 #   pragma warning( disable : 4054 )  /* conversion from function ptr to data ptr */
727 #   pragma warning( disable : 4055 )  /* conversion from data ptr to function ptr */
728 #endif
729         glpi = (LPFN_GLPI)(void*)GetProcAddress(GetModuleHandle(TEXT("kernel32")),
730                                                "GetLogicalProcessorInformation");
731 
732         if (glpi == NULL) {
733             goto failed;
734         }
735 
736         while(!done) {
737             DWORD rc = glpi(buffer, &returnLength);
738             if (FALSE == rc) {
739                 if (GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
740                     if (buffer)
741                         free(buffer);
742                     buffer = (PSYSTEM_LOGICAL_PROCESSOR_INFORMATION)malloc(returnLength);
743 
744                     if (buffer == NULL) {
745                         perror("zstd");
746                         exit(1);
747                     }
748                 } else {
749                     /* some other error */
750                     goto failed;
751                 }
752             } else {
753                 done = TRUE;
754         }   }
755 
756         ptr = buffer;
757 
758         while (byteOffset + sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION) <= returnLength) {
759 
760             if (ptr->Relationship == RelationProcessorCore) {
761                 numPhysicalCores++;
762             }
763 
764             ptr++;
765             byteOffset += sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION);
766         }
767 
768         free(buffer);
769 
770         return numPhysicalCores;
771     }
772 
773 failed:
774     /* try to fall back on GetSystemInfo */
775     {   SYSTEM_INFO sysinfo;
776         GetSystemInfo(&sysinfo);
777         numPhysicalCores = sysinfo.dwNumberOfProcessors;
778         if (numPhysicalCores == 0) numPhysicalCores = 1; /* just in case */
779     }
780     return numPhysicalCores;
781 }
782 
783 #elif defined(__APPLE__)
784 
785 #include <sys/sysctl.h>
786 
787 /* Use apple-provided syscall
788  * see: man 3 sysctl */
789 int UTIL_countPhysicalCores(void)
790 {
791     static S32 numPhysicalCores = 0; /* apple specifies int32_t */
792     if (numPhysicalCores != 0) return numPhysicalCores;
793 
794     {   size_t size = sizeof(S32);
795         int const ret = sysctlbyname("hw.physicalcpu", &numPhysicalCores, &size, NULL, 0);
796         if (ret != 0) {
797             if (errno == ENOENT) {
798                 /* entry not present, fall back on 1 */
799                 numPhysicalCores = 1;
800             } else {
801                 perror("zstd: can't get number of physical cpus");
802                 exit(1);
803             }
804         }
805 
806         return numPhysicalCores;
807     }
808 }
809 
810 #elif defined(__linux__)
811 
812 /* parse /proc/cpuinfo
813  * siblings / cpu cores should give hyperthreading ratio
814  * otherwise fall back on sysconf */
815 int UTIL_countPhysicalCores(void)
816 {
817     static int numPhysicalCores = 0;
818 
819     if (numPhysicalCores != 0) return numPhysicalCores;
820 
821     numPhysicalCores = (int)sysconf(_SC_NPROCESSORS_ONLN);
822     if (numPhysicalCores == -1) {
823         /* value not queryable, fall back on 1 */
824         return numPhysicalCores = 1;
825     }
826 
827     /* try to determine if there's hyperthreading */
828     {   FILE* const cpuinfo = fopen("/proc/cpuinfo", "r");
829 #define BUF_SIZE 80
830         char buff[BUF_SIZE];
831 
832         int siblings = 0;
833         int cpu_cores = 0;
834         int ratio = 1;
835 
836         if (cpuinfo == NULL) {
837             /* fall back on the sysconf value */
838             return numPhysicalCores;
839         }
840 
841         /* assume the cpu cores/siblings values will be constant across all
842          * present processors */
843         while (!feof(cpuinfo)) {
844             if (fgets(buff, BUF_SIZE, cpuinfo) != NULL) {
845                 if (strncmp(buff, "siblings", 8) == 0) {
846                     const char* const sep = strchr(buff, ':');
847                     if (sep == NULL || *sep == '\0') {
848                         /* formatting was broken? */
849                         goto failed;
850                     }
851 
852                     siblings = atoi(sep + 1);
853                 }
854                 if (strncmp(buff, "cpu cores", 9) == 0) {
855                     const char* const sep = strchr(buff, ':');
856                     if (sep == NULL || *sep == '\0') {
857                         /* formatting was broken? */
858                         goto failed;
859                     }
860 
861                     cpu_cores = atoi(sep + 1);
862                 }
863             } else if (ferror(cpuinfo)) {
864                 /* fall back on the sysconf value */
865                 goto failed;
866         }   }
867         if (siblings && cpu_cores) {
868             ratio = siblings / cpu_cores;
869         }
870 failed:
871         fclose(cpuinfo);
872         return numPhysicalCores = numPhysicalCores / ratio;
873     }
874 }
875 
876 #elif defined(__FreeBSD__)
877 
878 #include <sys/param.h>
879 #include <sys/sysctl.h>
880 
881 /* Use physical core sysctl when available
882  * see: man 4 smp, man 3 sysctl */
883 int UTIL_countPhysicalCores(void)
884 {
885     static int numPhysicalCores = 0; /* freebsd sysctl is native int sized */
886     if (numPhysicalCores != 0) return numPhysicalCores;
887 
888 #if __FreeBSD_version >= 1300008
889     {   size_t size = sizeof(numPhysicalCores);
890         int ret = sysctlbyname("kern.smp.cores", &numPhysicalCores, &size, NULL, 0);
891         if (ret == 0) return numPhysicalCores;
892         if (errno != ENOENT) {
893             perror("zstd: can't get number of physical cpus");
894             exit(1);
895         }
896         /* sysctl not present, fall through to older sysconf method */
897     }
898 #endif
899 
900     numPhysicalCores = (int)sysconf(_SC_NPROCESSORS_ONLN);
901     if (numPhysicalCores == -1) {
902         /* value not queryable, fall back on 1 */
903         numPhysicalCores = 1;
904     }
905     return numPhysicalCores;
906 }
907 
908 #elif defined(__NetBSD__) || defined(__OpenBSD__) || defined(__DragonFly__) || defined(__CYGWIN__)
909 
910 /* Use POSIX sysconf
911  * see: man 3 sysconf */
912 int UTIL_countPhysicalCores(void)
913 {
914     static int numPhysicalCores = 0;
915 
916     if (numPhysicalCores != 0) return numPhysicalCores;
917 
918     numPhysicalCores = (int)sysconf(_SC_NPROCESSORS_ONLN);
919     if (numPhysicalCores == -1) {
920         /* value not queryable, fall back on 1 */
921         return numPhysicalCores = 1;
922     }
923     return numPhysicalCores;
924 }
925 
926 #else
927 
928 int UTIL_countPhysicalCores(void)
929 {
930     /* assume 1 */
931     return 1;
932 }
933 
934 #endif
935 
936 #if defined (__cplusplus)
937 }
938 #endif
939