1 // © 2016 and later: Unicode, Inc. and others.
2 // License & terms of use: http://www.unicode.org/copyright.html
3 /*
4 ******************************************************************************
5 *
6 * Copyright (C) 1997-2016, International Business Machines
7 * Corporation and others. All Rights Reserved.
8 *
9 ******************************************************************************
10 *
11 * FILE NAME : putil.c (previously putil.cpp and ptypes.cpp)
12 *
13 * Date Name Description
14 * 04/14/97 aliu Creation.
15 * 04/24/97 aliu Added getDefaultDataDirectory() and
16 * getDefaultLocaleID().
17 * 04/28/97 aliu Rewritten to assume Unix and apply general methods
18 * for assumed case. Non-UNIX platforms must be
19 * special-cased. Rewrote numeric methods dealing
20 * with NaN and Infinity to be platform independent
21 * over all IEEE 754 platforms.
22 * 05/13/97 aliu Restored sign of timezone
23 * (semantics are hours West of GMT)
24 * 06/16/98 erm Added IEEE_754 stuff, cleaned up isInfinite, isNan,
25 * nextDouble..
26 * 07/22/98 stephen Added remainder, max, min, trunc
27 * 08/13/98 stephen Added isNegativeInfinity, isPositiveInfinity
28 * 08/24/98 stephen Added longBitsFromDouble
29 * 09/08/98 stephen Minor changes for Mac Port
30 * 03/02/99 stephen Removed openFile(). Added AS400 support.
31 * Fixed EBCDIC tables
32 * 04/15/99 stephen Converted to C.
33 * 06/28/99 stephen Removed mutex locking in u_isBigEndian().
34 * 08/04/99 jeffrey R. Added OS/2 changes
35 * 11/15/99 helena Integrated S/390 IEEE support.
36 * 04/26/01 Barry N. OS/400 support for uprv_getDefaultLocaleID
37 * 08/15/01 Steven H. OS/400 support for uprv_getDefaultCodepage
38 * 01/03/08 Steven L. Fake Time Support
39 ******************************************************************************
40 */
41
42 // Defines _XOPEN_SOURCE for access to POSIX functions.
43 // Must be before any other #includes.
44 #include "uposixdefs.h"
45
46 // First, the platform type. Need this for U_PLATFORM.
47 #include "unicode/platform.h"
48
49 #if U_PLATFORM == U_PF_MINGW && defined __STRICT_ANSI__
50 /* tzset isn't defined in strict ANSI on MinGW. */
51 #undef __STRICT_ANSI__
52 #endif
53
54 /*
55 * Cygwin with GCC requires inclusion of time.h after the above disabling strict asci mode statement.
56 */
57 #include <time.h>
58
59 #if !U_PLATFORM_USES_ONLY_WIN32_API
60 #include <sys/time.h>
61 #endif
62
63 /* include the rest of the ICU headers */
64 #include "unicode/putil.h"
65 #include "unicode/ustring.h"
66 #include "putilimp.h"
67 #include "uassert.h"
68 #include "umutex.h"
69 #include "cmemory.h"
70 #include "cstring.h"
71 #include "locmap.h"
72 #include "ucln_cmn.h"
73 #include "charstr.h"
74
75 /* Include standard headers. */
76 #include <stdio.h>
77 #include <stdlib.h>
78 #include <string.h>
79 #include <math.h>
80 #include <locale.h>
81 #include <float.h>
82
83 #ifndef U_COMMON_IMPLEMENTATION
84 #error U_COMMON_IMPLEMENTATION not set - must be set for all ICU source files in common/ - see https://unicode-org.github.io/icu/userguide/howtouseicu
85 #endif
86
87
88 /* include system headers */
89 #if U_PLATFORM_USES_ONLY_WIN32_API
90 /*
91 * TODO: U_PLATFORM_USES_ONLY_WIN32_API includes MinGW.
92 * Should Cygwin be included as well (U_PLATFORM_HAS_WIN32_API)
93 * to use native APIs as much as possible?
94 */
95 #ifndef WIN32_LEAN_AND_MEAN
96 # define WIN32_LEAN_AND_MEAN
97 #endif
98 # define VC_EXTRALEAN
99 # define NOUSER
100 # define NOSERVICE
101 # define NOIME
102 # define NOMCX
103 # include <windows.h>
104 # include "unicode/uloc.h"
105 # include "wintz.h"
106 #elif U_PLATFORM == U_PF_OS400
107 # include <float.h>
108 # include <qusec.h> /* error code structure */
109 # include <qusrjobi.h>
110 # include <qliept.h> /* EPT_CALL macro - this include must be after all other "QSYSINCs" */
111 # include <mih/testptr.h> /* For uprv_maximumPtr */
112 #elif U_PLATFORM == U_PF_OS390
113 # include "unicode/ucnv.h" /* Needed for UCNV_SWAP_LFNL_OPTION_STRING */
114 #elif U_PLATFORM_IS_DARWIN_BASED || U_PLATFORM_IS_LINUX_BASED || U_PLATFORM == U_PF_BSD || U_PLATFORM == U_PF_SOLARIS
115 # include <limits.h>
116 # include <unistd.h>
117 # if U_PLATFORM == U_PF_SOLARIS
118 # ifndef _XPG4_2
119 # define _XPG4_2
120 # endif
121 # elif U_PLATFORM == U_PF_ANDROID
122 # include <sys/system_properties.h>
123 # include <dlfcn.h>
124 # endif
125 #elif U_PLATFORM == U_PF_QNX
126 # include <sys/neutrino.h>
127 #endif
128
129
130 /*
131 * Only include langinfo.h if we have a way to get the codeset. If we later
132 * depend on more feature, we can test on U_HAVE_NL_LANGINFO.
133 *
134 */
135
136 #if U_HAVE_NL_LANGINFO_CODESET
137 #include <langinfo.h>
138 #endif
139
140 /**
141 * Simple things (presence of functions, etc) should just go in configure.in and be added to
142 * icucfg.h via autoheader.
143 */
144 #if U_PLATFORM_IMPLEMENTS_POSIX
145 # if U_PLATFORM == U_PF_OS400
146 # define HAVE_DLFCN_H 0
147 # define HAVE_DLOPEN 0
148 # else
149 # ifndef HAVE_DLFCN_H
150 # define HAVE_DLFCN_H 1
151 # endif
152 # ifndef HAVE_DLOPEN
153 # define HAVE_DLOPEN 1
154 # endif
155 # endif
156 # ifndef HAVE_GETTIMEOFDAY
157 # define HAVE_GETTIMEOFDAY 1
158 # endif
159 #else
160 # define HAVE_DLFCN_H 0
161 # define HAVE_DLOPEN 0
162 # define HAVE_GETTIMEOFDAY 0
163 #endif
164
165 U_NAMESPACE_USE
166
167 /* Define the extension for data files, again... */
168 #define DATA_TYPE "dat"
169
170 /* Leave this copyright notice here! */
171 static const char copyright[] = U_COPYRIGHT_STRING;
172
173 /* floating point implementations ------------------------------------------- */
174
175 /* We return QNAN rather than SNAN*/
176 #define SIGN 0x80000000U
177
178 /* Make it easy to define certain types of constants */
179 typedef union {
180 int64_t i64; /* This must be defined first in order to allow the initialization to work. This is a C89 feature. */
181 double d64;
182 } BitPatternConversion;
183 static const BitPatternConversion gNan = { (int64_t) INT64_C(0x7FF8000000000000) };
184 static const BitPatternConversion gInf = { (int64_t) INT64_C(0x7FF0000000000000) };
185
186 /*---------------------------------------------------------------------------
187 Platform utilities
188 Our general strategy is to assume we're on a POSIX platform. Platforms which
189 are non-POSIX must declare themselves so. The default POSIX implementation
190 will sometimes work for non-POSIX platforms as well (e.g., the NaN-related
191 functions).
192 ---------------------------------------------------------------------------*/
193
194 #if U_PLATFORM_USES_ONLY_WIN32_API || U_PLATFORM == U_PF_OS400
195 # undef U_POSIX_LOCALE
196 #else
197 # define U_POSIX_LOCALE 1
198 #endif
199
200 /*
201 WARNING! u_topNBytesOfDouble and u_bottomNBytesOfDouble
202 can't be properly optimized by the gcc compiler sometimes (i.e. gcc 3.2).
203 */
204 #if !IEEE_754
205 static char*
u_topNBytesOfDouble(double * d,int n)206 u_topNBytesOfDouble(double* d, int n)
207 {
208 #if U_IS_BIG_ENDIAN
209 return (char*)d;
210 #else
211 return (char*)(d + 1) - n;
212 #endif
213 }
214
215 static char*
u_bottomNBytesOfDouble(double * d,int n)216 u_bottomNBytesOfDouble(double* d, int n)
217 {
218 #if U_IS_BIG_ENDIAN
219 return (char*)(d + 1) - n;
220 #else
221 return (char*)d;
222 #endif
223 }
224 #endif /* !IEEE_754 */
225
226 #if IEEE_754
227 static UBool
u_signBit(double d)228 u_signBit(double d) {
229 uint8_t hiByte;
230 #if U_IS_BIG_ENDIAN
231 hiByte = *(uint8_t *)&d;
232 #else
233 hiByte = *(((uint8_t *)&d) + sizeof(double) - 1);
234 #endif
235 return (hiByte & 0x80) != 0;
236 }
237 #endif
238
239
240
241 #if defined (U_DEBUG_FAKETIME)
242 /* Override the clock to test things without having to move the system clock.
243 * Assumes POSIX gettimeofday() will function
244 */
245 UDate fakeClock_t0 = 0; /** Time to start the clock from **/
246 UDate fakeClock_dt = 0; /** Offset (fake time - real time) **/
247 UBool fakeClock_set = false; /** True if fake clock has spun up **/
248
getUTCtime_real()249 static UDate getUTCtime_real() {
250 struct timeval posixTime;
251 gettimeofday(&posixTime, nullptr);
252 return (UDate)(((int64_t)posixTime.tv_sec * U_MILLIS_PER_SECOND) + (posixTime.tv_usec/1000));
253 }
254
getUTCtime_fake()255 static UDate getUTCtime_fake() {
256 static UMutex fakeClockMutex;
257 umtx_lock(&fakeClockMutex);
258 if(!fakeClock_set) {
259 UDate real = getUTCtime_real();
260 const char *fake_start = getenv("U_FAKETIME_START");
261 if((fake_start!=nullptr) && (fake_start[0]!=0)) {
262 sscanf(fake_start,"%lf",&fakeClock_t0);
263 fakeClock_dt = fakeClock_t0 - real;
264 fprintf(stderr,"U_DEBUG_FAKETIME was set at compile time, so the ICU clock will start at a preset value\n"
265 "env variable U_FAKETIME_START=%.0f (%s) for an offset of %.0f ms from the current time %.0f\n",
266 fakeClock_t0, fake_start, fakeClock_dt, real);
267 } else {
268 fakeClock_dt = 0;
269 fprintf(stderr,"U_DEBUG_FAKETIME was set at compile time, but U_FAKETIME_START was not set.\n"
270 "Set U_FAKETIME_START to the number of milliseconds since 1/1/1970 to set the ICU clock.\n");
271 }
272 fakeClock_set = true;
273 }
274 umtx_unlock(&fakeClockMutex);
275
276 return getUTCtime_real() + fakeClock_dt;
277 }
278 #endif
279
280 #if U_PLATFORM_USES_ONLY_WIN32_API
281 typedef union {
282 int64_t int64;
283 FILETIME fileTime;
284 } FileTimeConversion; /* This is like a ULARGE_INTEGER */
285
286 /* Number of 100 nanoseconds from 1/1/1601 to 1/1/1970 */
287 #define EPOCH_BIAS INT64_C(116444736000000000)
288 #define HECTONANOSECOND_PER_MILLISECOND 10000
289
290 #endif
291
292 /*---------------------------------------------------------------------------
293 Universal Implementations
294 These are designed to work on all platforms. Try these, and if they
295 don't work on your platform, then special case your platform with new
296 implementations.
297 ---------------------------------------------------------------------------*/
298
299 U_CAPI UDate U_EXPORT2
uprv_getUTCtime()300 uprv_getUTCtime()
301 {
302 #if defined(U_DEBUG_FAKETIME)
303 return getUTCtime_fake(); /* Hook for overriding the clock */
304 #else
305 return uprv_getRawUTCtime();
306 #endif
307 }
308
309 /* Return UTC (GMT) time measured in milliseconds since 0:00 on 1/1/70.*/
310 U_CAPI UDate U_EXPORT2
uprv_getRawUTCtime()311 uprv_getRawUTCtime()
312 {
313 #if U_PLATFORM_USES_ONLY_WIN32_API
314
315 FileTimeConversion winTime;
316 GetSystemTimeAsFileTime(&winTime.fileTime);
317 return (UDate)((winTime.int64 - EPOCH_BIAS) / HECTONANOSECOND_PER_MILLISECOND);
318 #else
319
320 #if HAVE_GETTIMEOFDAY
321 struct timeval posixTime;
322 gettimeofday(&posixTime, nullptr);
323 return (UDate)(((int64_t)posixTime.tv_sec * U_MILLIS_PER_SECOND) + (posixTime.tv_usec/1000));
324 #else
325 time_t epochtime;
326 time(&epochtime);
327 return (UDate)epochtime * U_MILLIS_PER_SECOND;
328 #endif
329
330 #endif
331 }
332
333 /*-----------------------------------------------------------------------------
334 IEEE 754
335 These methods detect and return NaN and infinity values for doubles
336 conforming to IEEE 754. Platforms which support this standard include X86,
337 Mac 680x0, Mac PowerPC, AIX RS/6000, and most others.
338 If this doesn't work on your platform, you have non-IEEE floating-point, and
339 will need to code your own versions. A naive implementation is to return 0.0
340 for getNaN and getInfinity, and false for isNaN and isInfinite.
341 ---------------------------------------------------------------------------*/
342
343 U_CAPI UBool U_EXPORT2
uprv_isNaN(double number)344 uprv_isNaN(double number)
345 {
346 #if IEEE_754
347 BitPatternConversion convertedNumber;
348 convertedNumber.d64 = number;
349 /* Infinity is 0x7FF0000000000000U. Anything greater than that is a NaN */
350 return (UBool)((convertedNumber.i64 & U_INT64_MAX) > gInf.i64);
351
352 #elif U_PLATFORM == U_PF_OS390
353 uint32_t highBits = *(uint32_t*)u_topNBytesOfDouble(&number,
354 sizeof(uint32_t));
355 uint32_t lowBits = *(uint32_t*)u_bottomNBytesOfDouble(&number,
356 sizeof(uint32_t));
357
358 return ((highBits & 0x7F080000L) == 0x7F080000L) &&
359 (lowBits == 0x00000000L);
360
361 #else
362 /* If your platform doesn't support IEEE 754 but *does* have an NaN value,*/
363 /* you'll need to replace this default implementation with what's correct*/
364 /* for your platform.*/
365 return number != number;
366 #endif
367 }
368
369 U_CAPI UBool U_EXPORT2
uprv_isInfinite(double number)370 uprv_isInfinite(double number)
371 {
372 #if IEEE_754
373 BitPatternConversion convertedNumber;
374 convertedNumber.d64 = number;
375 /* Infinity is exactly 0x7FF0000000000000U. */
376 return (UBool)((convertedNumber.i64 & U_INT64_MAX) == gInf.i64);
377 #elif U_PLATFORM == U_PF_OS390
378 uint32_t highBits = *(uint32_t*)u_topNBytesOfDouble(&number,
379 sizeof(uint32_t));
380 uint32_t lowBits = *(uint32_t*)u_bottomNBytesOfDouble(&number,
381 sizeof(uint32_t));
382
383 return ((highBits & ~SIGN) == 0x70FF0000L) && (lowBits == 0x00000000L);
384
385 #else
386 /* If your platform doesn't support IEEE 754 but *does* have an infinity*/
387 /* value, you'll need to replace this default implementation with what's*/
388 /* correct for your platform.*/
389 return number == (2.0 * number);
390 #endif
391 }
392
393 U_CAPI UBool U_EXPORT2
uprv_isPositiveInfinity(double number)394 uprv_isPositiveInfinity(double number)
395 {
396 #if IEEE_754 || U_PLATFORM == U_PF_OS390
397 return (UBool)(number > 0 && uprv_isInfinite(number));
398 #else
399 return uprv_isInfinite(number);
400 #endif
401 }
402
403 U_CAPI UBool U_EXPORT2
uprv_isNegativeInfinity(double number)404 uprv_isNegativeInfinity(double number)
405 {
406 #if IEEE_754 || U_PLATFORM == U_PF_OS390
407 return (UBool)(number < 0 && uprv_isInfinite(number));
408
409 #else
410 uint32_t highBits = *(uint32_t*)u_topNBytesOfDouble(&number,
411 sizeof(uint32_t));
412 return((highBits & SIGN) && uprv_isInfinite(number));
413
414 #endif
415 }
416
417 U_CAPI double U_EXPORT2
uprv_getNaN()418 uprv_getNaN()
419 {
420 #if IEEE_754 || U_PLATFORM == U_PF_OS390
421 return gNan.d64;
422 #else
423 /* If your platform doesn't support IEEE 754 but *does* have an NaN value,*/
424 /* you'll need to replace this default implementation with what's correct*/
425 /* for your platform.*/
426 return 0.0;
427 #endif
428 }
429
430 U_CAPI double U_EXPORT2
uprv_getInfinity()431 uprv_getInfinity()
432 {
433 #if IEEE_754 || U_PLATFORM == U_PF_OS390
434 return gInf.d64;
435 #else
436 /* If your platform doesn't support IEEE 754 but *does* have an infinity*/
437 /* value, you'll need to replace this default implementation with what's*/
438 /* correct for your platform.*/
439 return 0.0;
440 #endif
441 }
442
443 U_CAPI double U_EXPORT2
uprv_floor(double x)444 uprv_floor(double x)
445 {
446 return floor(x);
447 }
448
449 U_CAPI double U_EXPORT2
uprv_ceil(double x)450 uprv_ceil(double x)
451 {
452 return ceil(x);
453 }
454
455 U_CAPI double U_EXPORT2
uprv_round(double x)456 uprv_round(double x)
457 {
458 return uprv_floor(x + 0.5);
459 }
460
461 U_CAPI double U_EXPORT2
uprv_fabs(double x)462 uprv_fabs(double x)
463 {
464 return fabs(x);
465 }
466
467 U_CAPI double U_EXPORT2
uprv_modf(double x,double * y)468 uprv_modf(double x, double* y)
469 {
470 return modf(x, y);
471 }
472
473 U_CAPI double U_EXPORT2
uprv_fmod(double x,double y)474 uprv_fmod(double x, double y)
475 {
476 return fmod(x, y);
477 }
478
479 U_CAPI double U_EXPORT2
uprv_pow(double x,double y)480 uprv_pow(double x, double y)
481 {
482 /* This is declared as "double pow(double x, double y)" */
483 return pow(x, y);
484 }
485
486 U_CAPI double U_EXPORT2
uprv_pow10(int32_t x)487 uprv_pow10(int32_t x)
488 {
489 return pow(10.0, (double)x);
490 }
491
492 U_CAPI double U_EXPORT2
uprv_fmax(double x,double y)493 uprv_fmax(double x, double y)
494 {
495 #if IEEE_754
496 /* first handle NaN*/
497 if(uprv_isNaN(x) || uprv_isNaN(y))
498 return uprv_getNaN();
499
500 /* check for -0 and 0*/
501 if(x == 0.0 && y == 0.0 && u_signBit(x))
502 return y;
503
504 #endif
505
506 /* this should work for all flt point w/o NaN and Inf special cases */
507 return (x > y ? x : y);
508 }
509
510 U_CAPI double U_EXPORT2
uprv_fmin(double x,double y)511 uprv_fmin(double x, double y)
512 {
513 #if IEEE_754
514 /* first handle NaN*/
515 if(uprv_isNaN(x) || uprv_isNaN(y))
516 return uprv_getNaN();
517
518 /* check for -0 and 0*/
519 if(x == 0.0 && y == 0.0 && u_signBit(y))
520 return y;
521
522 #endif
523
524 /* this should work for all flt point w/o NaN and Inf special cases */
525 return (x > y ? y : x);
526 }
527
528 U_CAPI UBool U_EXPORT2
uprv_add32_overflow(int32_t a,int32_t b,int32_t * res)529 uprv_add32_overflow(int32_t a, int32_t b, int32_t* res) {
530 // NOTE: Some compilers (GCC, Clang) have primitives available, like __builtin_add_overflow.
531 // This function could be optimized by calling one of those primitives.
532 auto a64 = static_cast<int64_t>(a);
533 auto b64 = static_cast<int64_t>(b);
534 int64_t res64 = a64 + b64;
535 *res = static_cast<int32_t>(res64);
536 return res64 != *res;
537 }
538
539 U_CAPI UBool U_EXPORT2
uprv_mul32_overflow(int32_t a,int32_t b,int32_t * res)540 uprv_mul32_overflow(int32_t a, int32_t b, int32_t* res) {
541 // NOTE: Some compilers (GCC, Clang) have primitives available, like __builtin_mul_overflow.
542 // This function could be optimized by calling one of those primitives.
543 auto a64 = static_cast<int64_t>(a);
544 auto b64 = static_cast<int64_t>(b);
545 int64_t res64 = a64 * b64;
546 *res = static_cast<int32_t>(res64);
547 return res64 != *res;
548 }
549
550 /**
551 * Truncates the given double.
552 * trunc(3.3) = 3.0, trunc (-3.3) = -3.0
553 * This is different than calling floor() or ceil():
554 * floor(3.3) = 3, floor(-3.3) = -4
555 * ceil(3.3) = 4, ceil(-3.3) = -3
556 */
557 U_CAPI double U_EXPORT2
uprv_trunc(double d)558 uprv_trunc(double d)
559 {
560 #if IEEE_754
561 /* handle error cases*/
562 if(uprv_isNaN(d))
563 return uprv_getNaN();
564 if(uprv_isInfinite(d))
565 return uprv_getInfinity();
566
567 if(u_signBit(d)) /* Signbit() picks up -0.0; d<0 does not. */
568 return ceil(d);
569 else
570 return floor(d);
571
572 #else
573 return d >= 0 ? floor(d) : ceil(d);
574
575 #endif
576 }
577
578 /**
579 * Return the largest positive number that can be represented by an integer
580 * type of arbitrary bit length.
581 */
582 U_CAPI double U_EXPORT2
uprv_maxMantissa()583 uprv_maxMantissa()
584 {
585 return pow(2.0, DBL_MANT_DIG + 1.0) - 1.0;
586 }
587
588 U_CAPI double U_EXPORT2
uprv_log(double d)589 uprv_log(double d)
590 {
591 return log(d);
592 }
593
594 U_CAPI void * U_EXPORT2
uprv_maximumPtr(void * base)595 uprv_maximumPtr(void * base)
596 {
597 #if U_PLATFORM == U_PF_OS400
598 /*
599 * With the provided function we should never be out of range of a given segment
600 * (a traditional/typical segment that is). Our segments have 5 bytes for the
601 * id and 3 bytes for the offset. The key is that the casting takes care of
602 * only retrieving the offset portion minus x1000. Hence, the smallest offset
603 * seen in a program is x001000 and when casted to an int would be 0.
604 * That's why we can only add 0xffefff. Otherwise, we would exceed the segment.
605 *
606 * Currently, 16MB is the current addressing limitation on i5/OS if the activation is
607 * non-TERASPACE. If it is TERASPACE it is 2GB - 4k(header information).
608 * This function determines the activation based on the pointer that is passed in and
609 * calculates the appropriate maximum available size for
610 * each pointer type (TERASPACE and non-TERASPACE)
611 *
612 * Unlike other operating systems, the pointer model isn't determined at
613 * compile time on i5/OS.
614 */
615 if ((base != nullptr) && (_TESTPTR(base, _C_TERASPACE_CHECK))) {
616 /* if it is a TERASPACE pointer the max is 2GB - 4k */
617 return ((void *)(((char *)base)-((uint32_t)(base))+((uint32_t)0x7fffefff)));
618 }
619 /* otherwise 16MB since nullptr ptr is not checkable or the ptr is not TERASPACE */
620 return ((void *)(((char *)base)-((uint32_t)(base))+((uint32_t)0xffefff)));
621
622 #else
623 return U_MAX_PTR(base);
624 #endif
625 }
626
627 /*---------------------------------------------------------------------------
628 Platform-specific Implementations
629 Try these, and if they don't work on your platform, then special case your
630 platform with new implementations.
631 ---------------------------------------------------------------------------*/
632
633 /* Generic time zone layer -------------------------------------------------- */
634
635 /* Time zone utilities */
636 U_CAPI void U_EXPORT2
uprv_tzset()637 uprv_tzset()
638 {
639 #if defined(U_TZSET)
640 U_TZSET();
641 #else
642 /* no initialization*/
643 #endif
644 }
645
646 U_CAPI int32_t U_EXPORT2
uprv_timezone()647 uprv_timezone()
648 {
649 #ifdef U_TIMEZONE
650 return U_TIMEZONE;
651 #else
652 time_t t, t1, t2;
653 struct tm tmrec;
654 int32_t tdiff = 0;
655
656 time(&t);
657 uprv_memcpy( &tmrec, localtime(&t), sizeof(tmrec) );
658 #if U_PLATFORM != U_PF_IPHONE
659 UBool dst_checked = (tmrec.tm_isdst != 0); /* daylight savings time is checked*/
660 #endif
661 t1 = mktime(&tmrec); /* local time in seconds*/
662 uprv_memcpy( &tmrec, gmtime(&t), sizeof(tmrec) );
663 t2 = mktime(&tmrec); /* GMT (or UTC) in seconds*/
664 tdiff = t2 - t1;
665
666 #if U_PLATFORM != U_PF_IPHONE
667 /* imitate NT behaviour, which returns same timezone offset to GMT for
668 winter and summer.
669 This does not work on all platforms. For instance, on glibc on Linux
670 and on Mac OS 10.5, tdiff calculated above remains the same
671 regardless of whether DST is in effect or not. iOS is another
672 platform where this does not work. Linux + glibc and Mac OS 10.5
673 have U_TIMEZONE defined so that this code is not reached.
674 */
675 if (dst_checked)
676 tdiff += 3600;
677 #endif
678 return tdiff;
679 #endif
680 }
681
682 /* Note that U_TZNAME does *not* have to be tzname, but if it is,
683 some platforms need to have it declared here. */
684
685 #if defined(U_TZNAME) && (U_PLATFORM == U_PF_IRIX || U_PLATFORM_IS_DARWIN_BASED)
686 /* RS6000 and others reject char **tzname. */
687 extern U_IMPORT char *U_TZNAME[];
688 #endif
689
690 #if !UCONFIG_NO_FILE_IO && ((U_PLATFORM_IS_DARWIN_BASED && (U_PLATFORM != U_PF_IPHONE || defined(U_TIMEZONE))) || U_PLATFORM_IS_LINUX_BASED || U_PLATFORM == U_PF_BSD || U_PLATFORM == U_PF_SOLARIS)
691 /* These platforms are likely to use Olson timezone IDs. */
692 /* common targets of the symbolic link at TZDEFAULT are:
693 * "/usr/share/zoneinfo/<olsonID>" default, older Linux distros, macOS to 10.12
694 * "../usr/share/zoneinfo/<olsonID>" newer Linux distros: Red Hat Enterprise Linux 7, Ubuntu 16, SuSe Linux 12
695 * "/usr/share/lib/zoneinfo/<olsonID>" Solaris
696 * "../usr/share/lib/zoneinfo/<olsonID>" Solaris
697 * "/var/db/timezone/zoneinfo/<olsonID>" macOS 10.13
698 * To avoid checking lots of paths, just check that the target path
699 * before the <olsonID> ends with "/zoneinfo/", and the <olsonID> is valid.
700 */
701
702 #define CHECK_LOCALTIME_LINK 1
703 #if U_PLATFORM_IS_DARWIN_BASED
704 #include <tzfile.h>
705 #define TZZONEINFO (TZDIR "/")
706 #elif U_PLATFORM == U_PF_SOLARIS
707 #define TZDEFAULT "/etc/localtime"
708 #define TZZONEINFO "/usr/share/lib/zoneinfo/"
709 #define TZ_ENV_CHECK "localtime"
710 #else
711 #define TZDEFAULT "/etc/localtime"
712 #define TZZONEINFO "/usr/share/zoneinfo/"
713 #endif
714 #define TZZONEINFOTAIL "/zoneinfo/"
715 #if U_HAVE_DIRENT_H
716 #define TZFILE_SKIP "posixrules" /* tz file to skip when searching. */
717 /* Some Linux distributions have 'localtime' in /usr/share/zoneinfo
718 symlinked to /etc/localtime, which makes searchForTZFile return
719 'localtime' when it's the first match. */
720 #define TZFILE_SKIP2 "localtime"
721 #define SEARCH_TZFILE
722 #include <dirent.h> /* Needed to search through system timezone files */
723 #endif
724 static char gTimeZoneBuffer[PATH_MAX];
725 static char *gTimeZoneBufferPtr = nullptr;
726 #endif
727
728 #if !U_PLATFORM_USES_ONLY_WIN32_API
729 #define isNonDigit(ch) (ch < '0' || '9' < ch)
730 #define isDigit(ch) ('0' <= ch && ch <= '9')
isValidOlsonID(const char * id)731 static UBool isValidOlsonID(const char *id) {
732 int32_t idx = 0;
733 int32_t idxMax = 0;
734
735 /* Determine if this is something like Iceland (Olson ID)
736 or AST4ADT (non-Olson ID) */
737 while (id[idx] && isNonDigit(id[idx]) && id[idx] != ',') {
738 idx++;
739 }
740
741 /* Allow at maximum 2 numbers at the end of the id to support zone id's
742 like GMT+11. */
743 idxMax = idx + 2;
744 while (id[idx] && isDigit(id[idx]) && idx < idxMax) {
745 idx++;
746 }
747
748 /* If we went through the whole string, then it might be okay.
749 The timezone is sometimes set to "CST-7CDT", "CST6CDT5,J129,J131/19:30",
750 "GRNLNDST3GRNLNDDT" or similar, so we cannot use it.
751 The rest of the time it could be an Olson ID. George */
752 return (UBool)(id[idx] == 0
753 || uprv_strcmp(id, "PST8PDT") == 0
754 || uprv_strcmp(id, "MST7MDT") == 0
755 || uprv_strcmp(id, "CST6CDT") == 0
756 || uprv_strcmp(id, "EST5EDT") == 0);
757 }
758
759 /* On some Unix-like OS, 'posix' subdirectory in
760 /usr/share/zoneinfo replicates the top-level contents. 'right'
761 subdirectory has the same set of files, but individual files
762 are different from those in the top-level directory or 'posix'
763 because 'right' has files for TAI (Int'l Atomic Time) while 'posix'
764 has files for UTC.
765 When the first match for /etc/localtime is in either of them
766 (usually in posix because 'right' has different file contents),
767 or TZ environment variable points to one of them, createTimeZone
768 fails because, say, 'posix/America/New_York' is not an Olson
769 timezone id ('America/New_York' is). So, we have to skip
770 'posix/' and 'right/' at the beginning. */
skipZoneIDPrefix(const char ** id)771 static void skipZoneIDPrefix(const char** id) {
772 if (uprv_strncmp(*id, "posix/", 6) == 0
773 || uprv_strncmp(*id, "right/", 6) == 0)
774 {
775 *id += 6;
776 }
777 }
778 #endif
779
780 #if defined(U_TZNAME) && !U_PLATFORM_USES_ONLY_WIN32_API
781
782 #define CONVERT_HOURS_TO_SECONDS(offset) (int32_t)(offset*3600)
783 typedef struct OffsetZoneMapping {
784 int32_t offsetSeconds;
785 int32_t daylightType; /* 0=U_DAYLIGHT_NONE, 1=daylight in June-U_DAYLIGHT_JUNE, 2=daylight in December=U_DAYLIGHT_DECEMBER*/
786 const char *stdID;
787 const char *dstID;
788 const char *olsonID;
789 } OffsetZoneMapping;
790
791 enum { U_DAYLIGHT_NONE=0,U_DAYLIGHT_JUNE=1,U_DAYLIGHT_DECEMBER=2 };
792
793 /*
794 This list tries to disambiguate a set of abbreviated timezone IDs and offsets
795 and maps it to an Olson ID.
796 Before adding anything to this list, take a look at
797 icu/source/tools/tzcode/tz.alias
798 Sometimes no daylight savings (0) is important to define due to aliases.
799 This list can be tested with icu/source/test/compat/tzone.pl
800 More values could be added to daylightType to increase precision.
801 */
802 static const struct OffsetZoneMapping OFFSET_ZONE_MAPPINGS[] = {
803 {-45900, 2, "CHAST", "CHADT", "Pacific/Chatham"},
804 {-43200, 1, "PETT", "PETST", "Asia/Kamchatka"},
805 {-43200, 2, "NZST", "NZDT", "Pacific/Auckland"},
806 {-43200, 1, "ANAT", "ANAST", "Asia/Anadyr"},
807 {-39600, 1, "MAGT", "MAGST", "Asia/Magadan"},
808 {-37800, 2, "LHST", "LHST", "Australia/Lord_Howe"},
809 {-36000, 2, "EST", "EST", "Australia/Sydney"},
810 {-36000, 1, "SAKT", "SAKST", "Asia/Sakhalin"},
811 {-36000, 1, "VLAT", "VLAST", "Asia/Vladivostok"},
812 {-34200, 2, "CST", "CST", "Australia/South"},
813 {-32400, 1, "YAKT", "YAKST", "Asia/Yakutsk"},
814 {-32400, 1, "CHOT", "CHOST", "Asia/Choibalsan"},
815 {-31500, 2, "CWST", "CWST", "Australia/Eucla"},
816 {-28800, 1, "IRKT", "IRKST", "Asia/Irkutsk"},
817 {-28800, 1, "ULAT", "ULAST", "Asia/Ulaanbaatar"},
818 {-28800, 2, "WST", "WST", "Australia/West"},
819 {-25200, 1, "HOVT", "HOVST", "Asia/Hovd"},
820 {-25200, 1, "KRAT", "KRAST", "Asia/Krasnoyarsk"},
821 {-21600, 1, "NOVT", "NOVST", "Asia/Novosibirsk"},
822 {-21600, 1, "OMST", "OMSST", "Asia/Omsk"},
823 {-18000, 1, "YEKT", "YEKST", "Asia/Yekaterinburg"},
824 {-14400, 1, "SAMT", "SAMST", "Europe/Samara"},
825 {-14400, 1, "AMT", "AMST", "Asia/Yerevan"},
826 {-14400, 1, "AZT", "AZST", "Asia/Baku"},
827 {-10800, 1, "AST", "ADT", "Asia/Baghdad"},
828 {-10800, 1, "MSK", "MSD", "Europe/Moscow"},
829 {-10800, 1, "VOLT", "VOLST", "Europe/Volgograd"},
830 {-7200, 0, "EET", "CEST", "Africa/Tripoli"},
831 {-7200, 1, "EET", "EEST", "Europe/Athens"}, /* Conflicts with Africa/Cairo */
832 {-7200, 1, "IST", "IDT", "Asia/Jerusalem"},
833 {-3600, 0, "CET", "WEST", "Africa/Algiers"},
834 {-3600, 2, "WAT", "WAST", "Africa/Windhoek"},
835 {0, 1, "GMT", "IST", "Europe/Dublin"},
836 {0, 1, "GMT", "BST", "Europe/London"},
837 {0, 0, "WET", "WEST", "Africa/Casablanca"},
838 {0, 0, "WET", "WET", "Africa/El_Aaiun"},
839 {3600, 1, "AZOT", "AZOST", "Atlantic/Azores"},
840 {3600, 1, "EGT", "EGST", "America/Scoresbysund"},
841 {10800, 1, "PMST", "PMDT", "America/Miquelon"},
842 {10800, 2, "UYT", "UYST", "America/Montevideo"},
843 {10800, 1, "WGT", "WGST", "America/Godthab"},
844 {10800, 2, "BRT", "BRST", "Brazil/East"},
845 {12600, 1, "NST", "NDT", "America/St_Johns"},
846 {14400, 1, "AST", "ADT", "Canada/Atlantic"},
847 {14400, 2, "AMT", "AMST", "America/Cuiaba"},
848 {14400, 2, "CLT", "CLST", "Chile/Continental"},
849 {14400, 2, "FKT", "FKST", "Atlantic/Stanley"},
850 {14400, 2, "PYT", "PYST", "America/Asuncion"},
851 {18000, 1, "CST", "CDT", "America/Havana"},
852 {18000, 1, "EST", "EDT", "US/Eastern"}, /* Conflicts with America/Grand_Turk */
853 {21600, 2, "EAST", "EASST", "Chile/EasterIsland"},
854 {21600, 0, "CST", "MDT", "Canada/Saskatchewan"},
855 {21600, 0, "CST", "CDT", "America/Guatemala"},
856 {21600, 1, "CST", "CDT", "US/Central"}, /* Conflicts with Mexico/General */
857 {25200, 1, "MST", "MDT", "US/Mountain"}, /* Conflicts with Mexico/BajaSur */
858 {28800, 0, "PST", "PST", "Pacific/Pitcairn"},
859 {28800, 1, "PST", "PDT", "US/Pacific"}, /* Conflicts with Mexico/BajaNorte */
860 {32400, 1, "AKST", "AKDT", "US/Alaska"},
861 {36000, 1, "HAST", "HADT", "US/Aleutian"}
862 };
863
864 /*#define DEBUG_TZNAME*/
865
remapShortTimeZone(const char * stdID,const char * dstID,int32_t daylightType,int32_t offset)866 static const char* remapShortTimeZone(const char *stdID, const char *dstID, int32_t daylightType, int32_t offset)
867 {
868 int32_t idx;
869 #ifdef DEBUG_TZNAME
870 fprintf(stderr, "TZ=%s std=%s dst=%s daylight=%d offset=%d\n", getenv("TZ"), stdID, dstID, daylightType, offset);
871 #endif
872 for (idx = 0; idx < UPRV_LENGTHOF(OFFSET_ZONE_MAPPINGS); idx++)
873 {
874 if (offset == OFFSET_ZONE_MAPPINGS[idx].offsetSeconds
875 && daylightType == OFFSET_ZONE_MAPPINGS[idx].daylightType
876 && strcmp(OFFSET_ZONE_MAPPINGS[idx].stdID, stdID) == 0
877 && strcmp(OFFSET_ZONE_MAPPINGS[idx].dstID, dstID) == 0)
878 {
879 return OFFSET_ZONE_MAPPINGS[idx].olsonID;
880 }
881 }
882 return nullptr;
883 }
884 #endif
885
886 #ifdef SEARCH_TZFILE
887 #define MAX_READ_SIZE 512
888
889 typedef struct DefaultTZInfo {
890 char* defaultTZBuffer;
891 int64_t defaultTZFileSize;
892 FILE* defaultTZFilePtr;
893 UBool defaultTZstatus;
894 int32_t defaultTZPosition;
895 } DefaultTZInfo;
896
897 /*
898 * This method compares the two files given to see if they are a match.
899 * It is currently use to compare two TZ files.
900 */
compareBinaryFiles(const char * defaultTZFileName,const char * TZFileName,DefaultTZInfo * tzInfo)901 static UBool compareBinaryFiles(const char* defaultTZFileName, const char* TZFileName, DefaultTZInfo* tzInfo) {
902 FILE* file;
903 int64_t sizeFile;
904 int64_t sizeFileLeft;
905 int32_t sizeFileRead;
906 int32_t sizeFileToRead;
907 char bufferFile[MAX_READ_SIZE];
908 UBool result = true;
909
910 if (tzInfo->defaultTZFilePtr == nullptr) {
911 tzInfo->defaultTZFilePtr = fopen(defaultTZFileName, "r");
912 }
913 file = fopen(TZFileName, "r");
914
915 tzInfo->defaultTZPosition = 0; /* reset position to begin search */
916
917 if (file != nullptr && tzInfo->defaultTZFilePtr != nullptr) {
918 /* First check that the file size are equal. */
919 if (tzInfo->defaultTZFileSize == 0) {
920 fseek(tzInfo->defaultTZFilePtr, 0, SEEK_END);
921 tzInfo->defaultTZFileSize = ftell(tzInfo->defaultTZFilePtr);
922 }
923 fseek(file, 0, SEEK_END);
924 sizeFile = ftell(file);
925 sizeFileLeft = sizeFile;
926
927 if (sizeFile != tzInfo->defaultTZFileSize) {
928 result = false;
929 } else {
930 /* Store the data from the files in separate buffers and
931 * compare each byte to determine equality.
932 */
933 if (tzInfo->defaultTZBuffer == nullptr) {
934 rewind(tzInfo->defaultTZFilePtr);
935 tzInfo->defaultTZBuffer = (char*)uprv_malloc(sizeof(char) * tzInfo->defaultTZFileSize);
936 sizeFileRead = fread(tzInfo->defaultTZBuffer, 1, tzInfo->defaultTZFileSize, tzInfo->defaultTZFilePtr);
937 }
938 rewind(file);
939 while(sizeFileLeft > 0) {
940 uprv_memset(bufferFile, 0, MAX_READ_SIZE);
941 sizeFileToRead = sizeFileLeft < MAX_READ_SIZE ? sizeFileLeft : MAX_READ_SIZE;
942
943 sizeFileRead = fread(bufferFile, 1, sizeFileToRead, file);
944 if (memcmp(tzInfo->defaultTZBuffer + tzInfo->defaultTZPosition, bufferFile, sizeFileRead) != 0) {
945 result = false;
946 break;
947 }
948 sizeFileLeft -= sizeFileRead;
949 tzInfo->defaultTZPosition += sizeFileRead;
950 }
951 }
952 } else {
953 result = false;
954 }
955
956 if (file != nullptr) {
957 fclose(file);
958 }
959
960 return result;
961 }
962
963
964 /* dirent also lists two entries: "." and ".." that we can safely ignore. */
965 #define SKIP1 "."
966 #define SKIP2 ".."
967 static UBool U_CALLCONV putil_cleanup();
968 static CharString *gSearchTZFileResult = nullptr;
969
970 /*
971 * This method recursively traverses the directory given for a matching TZ file and returns the first match.
972 * This function is not thread safe - it uses a global, gSearchTZFileResult, to hold its results.
973 */
searchForTZFile(const char * path,DefaultTZInfo * tzInfo)974 static char* searchForTZFile(const char* path, DefaultTZInfo* tzInfo) {
975 DIR* dirp = nullptr;
976 struct dirent* dirEntry = nullptr;
977 char* result = nullptr;
978 UErrorCode status = U_ZERO_ERROR;
979
980 /* Save the current path */
981 CharString curpath(path, -1, status);
982 if (U_FAILURE(status)) {
983 goto cleanupAndReturn;
984 }
985
986 dirp = opendir(path);
987 if (dirp == nullptr) {
988 goto cleanupAndReturn;
989 }
990
991 if (gSearchTZFileResult == nullptr) {
992 gSearchTZFileResult = new CharString;
993 if (gSearchTZFileResult == nullptr) {
994 goto cleanupAndReturn;
995 }
996 ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
997 }
998
999 /* Check each entry in the directory. */
1000 while((dirEntry = readdir(dirp)) != nullptr) {
1001 const char* dirName = dirEntry->d_name;
1002 if (uprv_strcmp(dirName, SKIP1) != 0 && uprv_strcmp(dirName, SKIP2) != 0
1003 && uprv_strcmp(TZFILE_SKIP, dirName) != 0 && uprv_strcmp(TZFILE_SKIP2, dirName) != 0) {
1004 /* Create a newpath with the new entry to test each entry in the directory. */
1005 CharString newpath(curpath, status);
1006 newpath.append(dirName, -1, status);
1007 if (U_FAILURE(status)) {
1008 break;
1009 }
1010
1011 DIR* subDirp = nullptr;
1012 if ((subDirp = opendir(newpath.data())) != nullptr) {
1013 /* If this new path is a directory, make a recursive call with the newpath. */
1014 closedir(subDirp);
1015 newpath.append('/', status);
1016 if (U_FAILURE(status)) {
1017 break;
1018 }
1019 result = searchForTZFile(newpath.data(), tzInfo);
1020 /*
1021 Have to get out here. Otherwise, we'd keep looking
1022 and return the first match in the top-level directory
1023 if there's a match in the top-level. If not, this function
1024 would return nullptr and set gTimeZoneBufferPtr to nullptr in initDefault().
1025 It worked without this in most cases because we have a fallback of calling
1026 localtime_r to figure out the default timezone.
1027 */
1028 if (result != nullptr)
1029 break;
1030 } else {
1031 if(compareBinaryFiles(TZDEFAULT, newpath.data(), tzInfo)) {
1032 int32_t amountToSkip = sizeof(TZZONEINFO) - 1;
1033 if (amountToSkip > newpath.length()) {
1034 amountToSkip = newpath.length();
1035 }
1036 const char* zoneid = newpath.data() + amountToSkip;
1037 skipZoneIDPrefix(&zoneid);
1038 gSearchTZFileResult->clear();
1039 gSearchTZFileResult->append(zoneid, -1, status);
1040 if (U_FAILURE(status)) {
1041 break;
1042 }
1043 result = gSearchTZFileResult->data();
1044 /* Get out after the first one found. */
1045 break;
1046 }
1047 }
1048 }
1049 }
1050
1051 cleanupAndReturn:
1052 if (dirp) {
1053 closedir(dirp);
1054 }
1055 return result;
1056 }
1057 #endif
1058
1059 #if U_PLATFORM == U_PF_ANDROID
1060 typedef int(system_property_read_callback)(const prop_info* info,
1061 void (*callback)(void* cookie,
1062 const char* name,
1063 const char* value,
1064 uint32_t serial),
1065 void* cookie);
1066 typedef int(system_property_get)(const char*, char*);
1067
1068 static char gAndroidTimeZone[PROP_VALUE_MAX] = { '\0' };
1069
u_property_read(void * cookie,const char * name,const char * value,uint32_t serial)1070 static void u_property_read(void* cookie, const char* name, const char* value,
1071 uint32_t serial) {
1072 uprv_strcpy((char* )cookie, value);
1073 }
1074 #endif
1075
1076 U_CAPI void U_EXPORT2
uprv_tzname_clear_cache()1077 uprv_tzname_clear_cache()
1078 {
1079 #if U_PLATFORM == U_PF_ANDROID
1080 /* Android's timezone is stored in system property. */
1081 gAndroidTimeZone[0] = '\0';
1082 void* libc = dlopen("libc.so", RTLD_NOLOAD);
1083 if (libc) {
1084 /* Android API 26+ has new API to get system property and old API
1085 * (__system_property_get) is deprecated */
1086 system_property_read_callback* property_read_callback =
1087 (system_property_read_callback*)dlsym(
1088 libc, "__system_property_read_callback");
1089 if (property_read_callback) {
1090 const prop_info* info =
1091 __system_property_find("persist.sys.timezone");
1092 if (info) {
1093 property_read_callback(info, &u_property_read, gAndroidTimeZone);
1094 }
1095 } else {
1096 system_property_get* property_get =
1097 (system_property_get*)dlsym(libc, "__system_property_get");
1098 if (property_get) {
1099 property_get("persist.sys.timezone", gAndroidTimeZone);
1100 }
1101 }
1102 dlclose(libc);
1103 }
1104 #endif
1105
1106 #if defined(CHECK_LOCALTIME_LINK) && !defined(DEBUG_SKIP_LOCALTIME_LINK)
1107 gTimeZoneBufferPtr = nullptr;
1108 #endif
1109 }
1110
1111 U_CAPI const char* U_EXPORT2
uprv_tzname(int n)1112 uprv_tzname(int n)
1113 {
1114 (void)n; // Avoid unreferenced parameter warning.
1115 const char *tzid = nullptr;
1116 #if U_PLATFORM_USES_ONLY_WIN32_API
1117 tzid = uprv_detectWindowsTimeZone();
1118
1119 if (tzid != nullptr) {
1120 return tzid;
1121 }
1122
1123 #ifndef U_TZNAME
1124 // The return value is free'd in timezone.cpp on Windows because
1125 // the other code path returns a pointer to a heap location.
1126 // If we don't have a name already, then tzname wouldn't be any
1127 // better, so just fall back.
1128 return uprv_strdup("");
1129 #endif // !U_TZNAME
1130
1131 #else
1132
1133 /*#if U_PLATFORM_IS_DARWIN_BASED
1134 int ret;
1135
1136 tzid = getenv("TZFILE");
1137 if (tzid != nullptr) {
1138 return tzid;
1139 }
1140 #endif*/
1141
1142 /* This code can be temporarily disabled to test tzname resolution later on. */
1143 #ifndef DEBUG_TZNAME
1144 #if U_PLATFORM == U_PF_ANDROID
1145 tzid = gAndroidTimeZone;
1146 #else
1147 tzid = getenv("TZ");
1148 #endif
1149 if (tzid != nullptr && isValidOlsonID(tzid)
1150 #if U_PLATFORM == U_PF_SOLARIS
1151 /* Don't misinterpret TZ "localtime" on Solaris as a time zone name. */
1152 && uprv_strcmp(tzid, TZ_ENV_CHECK) != 0
1153 #endif
1154 ) {
1155 /* The colon forces tzset() to treat the remainder as zoneinfo path */
1156 if (tzid[0] == ':') {
1157 tzid++;
1158 }
1159 /* This might be a good Olson ID. */
1160 skipZoneIDPrefix(&tzid);
1161 return tzid;
1162 }
1163 /* else U_TZNAME will give a better result. */
1164 #endif
1165
1166 #if defined(CHECK_LOCALTIME_LINK) && !defined(DEBUG_SKIP_LOCALTIME_LINK)
1167 /* Caller must handle threading issues */
1168 if (gTimeZoneBufferPtr == nullptr) {
1169 /*
1170 This is a trick to look at the name of the link to get the Olson ID
1171 because the tzfile contents is underspecified.
1172 This isn't guaranteed to work because it may not be a symlink.
1173 */
1174 int32_t ret = (int32_t)readlink(TZDEFAULT, gTimeZoneBuffer, sizeof(gTimeZoneBuffer)-1);
1175 if (0 < ret) {
1176 int32_t tzZoneInfoTailLen = uprv_strlen(TZZONEINFOTAIL);
1177 char * tzZoneInfoTailPtr = uprv_strstr(gTimeZoneBuffer, TZZONEINFOTAIL);
1178
1179 if (tzZoneInfoTailPtr != nullptr
1180 && isValidOlsonID(tzZoneInfoTailPtr + tzZoneInfoTailLen))
1181 {
1182 return (gTimeZoneBufferPtr = tzZoneInfoTailPtr + tzZoneInfoTailLen);
1183 }
1184 } else {
1185 #if defined(SEARCH_TZFILE)
1186 DefaultTZInfo* tzInfo = (DefaultTZInfo*)uprv_malloc(sizeof(DefaultTZInfo));
1187 if (tzInfo != nullptr) {
1188 tzInfo->defaultTZBuffer = nullptr;
1189 tzInfo->defaultTZFileSize = 0;
1190 tzInfo->defaultTZFilePtr = nullptr;
1191 tzInfo->defaultTZstatus = false;
1192 tzInfo->defaultTZPosition = 0;
1193
1194 gTimeZoneBufferPtr = searchForTZFile(TZZONEINFO, tzInfo);
1195
1196 /* Free previously allocated memory */
1197 if (tzInfo->defaultTZBuffer != nullptr) {
1198 uprv_free(tzInfo->defaultTZBuffer);
1199 }
1200 if (tzInfo->defaultTZFilePtr != nullptr) {
1201 fclose(tzInfo->defaultTZFilePtr);
1202 }
1203 uprv_free(tzInfo);
1204 }
1205
1206 if (gTimeZoneBufferPtr != nullptr && isValidOlsonID(gTimeZoneBufferPtr)) {
1207 return gTimeZoneBufferPtr;
1208 }
1209 #endif
1210 }
1211 }
1212 else {
1213 return gTimeZoneBufferPtr;
1214 }
1215 #endif
1216 #endif
1217
1218 #ifdef U_TZNAME
1219 #if U_PLATFORM_USES_ONLY_WIN32_API
1220 /* The return value is free'd in timezone.cpp on Windows because
1221 * the other code path returns a pointer to a heap location. */
1222 return uprv_strdup(U_TZNAME[n]);
1223 #else
1224 /*
1225 U_TZNAME is usually a non-unique abbreviation, which isn't normally usable.
1226 So we remap the abbreviation to an olson ID.
1227
1228 Since Windows exposes a little more timezone information,
1229 we normally don't use this code on Windows because
1230 uprv_detectWindowsTimeZone should have already given the correct answer.
1231 */
1232 {
1233 struct tm juneSol, decemberSol;
1234 int daylightType;
1235 static const time_t juneSolstice=1182478260; /*2007-06-21 18:11 UT*/
1236 static const time_t decemberSolstice=1198332540; /*2007-12-22 06:09 UT*/
1237
1238 /* This probing will tell us when daylight savings occurs. */
1239 localtime_r(&juneSolstice, &juneSol);
1240 localtime_r(&decemberSolstice, &decemberSol);
1241 if(decemberSol.tm_isdst > 0) {
1242 daylightType = U_DAYLIGHT_DECEMBER;
1243 } else if(juneSol.tm_isdst > 0) {
1244 daylightType = U_DAYLIGHT_JUNE;
1245 } else {
1246 daylightType = U_DAYLIGHT_NONE;
1247 }
1248 tzid = remapShortTimeZone(U_TZNAME[0], U_TZNAME[1], daylightType, uprv_timezone());
1249 if (tzid != nullptr) {
1250 return tzid;
1251 }
1252 }
1253 return U_TZNAME[n];
1254 #endif
1255 #else
1256 return "";
1257 #endif
1258 }
1259
1260 /* Get and set the ICU data directory --------------------------------------- */
1261
1262 static icu::UInitOnce gDataDirInitOnce {};
1263 static char *gDataDirectory = nullptr;
1264
1265 UInitOnce gTimeZoneFilesInitOnce {};
1266 static CharString *gTimeZoneFilesDirectory = nullptr;
1267
1268 #if U_POSIX_LOCALE || U_PLATFORM_USES_ONLY_WIN32_API
1269 static const char *gCorrectedPOSIXLocale = nullptr; /* Sometimes heap allocated */
1270 static bool gCorrectedPOSIXLocaleHeapAllocated = false;
1271 #endif
1272
putil_cleanup()1273 static UBool U_CALLCONV putil_cleanup()
1274 {
1275 if (gDataDirectory && *gDataDirectory) {
1276 uprv_free(gDataDirectory);
1277 }
1278 gDataDirectory = nullptr;
1279 gDataDirInitOnce.reset();
1280
1281 delete gTimeZoneFilesDirectory;
1282 gTimeZoneFilesDirectory = nullptr;
1283 gTimeZoneFilesInitOnce.reset();
1284
1285 #ifdef SEARCH_TZFILE
1286 delete gSearchTZFileResult;
1287 gSearchTZFileResult = nullptr;
1288 #endif
1289
1290 #if U_POSIX_LOCALE || U_PLATFORM_USES_ONLY_WIN32_API
1291 if (gCorrectedPOSIXLocale && gCorrectedPOSIXLocaleHeapAllocated) {
1292 uprv_free(const_cast<char *>(gCorrectedPOSIXLocale));
1293 gCorrectedPOSIXLocale = nullptr;
1294 gCorrectedPOSIXLocaleHeapAllocated = false;
1295 }
1296 #endif
1297 return true;
1298 }
1299
1300 /*
1301 * Set the data directory.
1302 * Make a copy of the passed string, and set the global data dir to point to it.
1303 */
1304 U_CAPI void U_EXPORT2
u_setDataDirectory(const char * directory)1305 u_setDataDirectory(const char *directory) {
1306 char *newDataDir;
1307 int32_t length;
1308
1309 if(directory==nullptr || *directory==0) {
1310 /* A small optimization to prevent the malloc and copy when the
1311 shared library is used, and this is a way to make sure that nullptr
1312 is never returned.
1313 */
1314 newDataDir = (char *)"";
1315 }
1316 else {
1317 length=(int32_t)uprv_strlen(directory);
1318 newDataDir = (char *)uprv_malloc(length + 2);
1319 /* Exit out if newDataDir could not be created. */
1320 if (newDataDir == nullptr) {
1321 return;
1322 }
1323 uprv_strcpy(newDataDir, directory);
1324
1325 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR)
1326 {
1327 char *p;
1328 while((p = uprv_strchr(newDataDir, U_FILE_ALT_SEP_CHAR)) != nullptr) {
1329 *p = U_FILE_SEP_CHAR;
1330 }
1331 }
1332 #endif
1333 }
1334
1335 if (gDataDirectory && *gDataDirectory) {
1336 uprv_free(gDataDirectory);
1337 }
1338 gDataDirectory = newDataDir;
1339 ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
1340 }
1341
1342 U_CAPI UBool U_EXPORT2
uprv_pathIsAbsolute(const char * path)1343 uprv_pathIsAbsolute(const char *path)
1344 {
1345 if(!path || !*path) {
1346 return false;
1347 }
1348
1349 if(*path == U_FILE_SEP_CHAR) {
1350 return true;
1351 }
1352
1353 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR)
1354 if(*path == U_FILE_ALT_SEP_CHAR) {
1355 return true;
1356 }
1357 #endif
1358
1359 #if U_PLATFORM_USES_ONLY_WIN32_API
1360 if( (((path[0] >= 'A') && (path[0] <= 'Z')) ||
1361 ((path[0] >= 'a') && (path[0] <= 'z'))) &&
1362 path[1] == ':' ) {
1363 return true;
1364 }
1365 #endif
1366
1367 return false;
1368 }
1369
1370 /* Backup setting of ICU_DATA_DIR_PREFIX_ENV_VAR
1371 (needed for some Darwin ICU build environments) */
1372 #if U_PLATFORM_IS_DARWIN_BASED && defined(TARGET_OS_SIMULATOR) && TARGET_OS_SIMULATOR
1373 # if !defined(ICU_DATA_DIR_PREFIX_ENV_VAR)
1374 # define ICU_DATA_DIR_PREFIX_ENV_VAR "IPHONE_SIMULATOR_ROOT"
1375 # endif
1376 #endif
1377
1378 #if defined(ICU_DATA_DIR_WINDOWS)
1379 // Helper function to get the ICU Data Directory under the Windows directory location.
getIcuDataDirectoryUnderWindowsDirectory(char * directoryBuffer,UINT bufferLength)1380 static BOOL U_CALLCONV getIcuDataDirectoryUnderWindowsDirectory(char* directoryBuffer, UINT bufferLength)
1381 {
1382 wchar_t windowsPath[MAX_PATH];
1383 char windowsPathUtf8[MAX_PATH];
1384
1385 UINT length = GetSystemWindowsDirectoryW(windowsPath, UPRV_LENGTHOF(windowsPath));
1386 if ((length > 0) && (length < (UPRV_LENGTHOF(windowsPath) - 1))) {
1387 // Convert UTF-16 to a UTF-8 string.
1388 UErrorCode status = U_ZERO_ERROR;
1389 int32_t windowsPathUtf8Len = 0;
1390 u_strToUTF8(windowsPathUtf8, static_cast<int32_t>(UPRV_LENGTHOF(windowsPathUtf8)),
1391 &windowsPathUtf8Len, reinterpret_cast<const char16_t*>(windowsPath), -1, &status);
1392
1393 if (U_SUCCESS(status) && (status != U_STRING_NOT_TERMINATED_WARNING) &&
1394 (windowsPathUtf8Len < (UPRV_LENGTHOF(windowsPathUtf8) - 1))) {
1395 // Ensure it always has a separator, so we can append the ICU data path.
1396 if (windowsPathUtf8[windowsPathUtf8Len - 1] != U_FILE_SEP_CHAR) {
1397 windowsPathUtf8[windowsPathUtf8Len++] = U_FILE_SEP_CHAR;
1398 windowsPathUtf8[windowsPathUtf8Len] = '\0';
1399 }
1400 // Check if the concatenated string will fit.
1401 if ((windowsPathUtf8Len + UPRV_LENGTHOF(ICU_DATA_DIR_WINDOWS)) < bufferLength) {
1402 uprv_strcpy(directoryBuffer, windowsPathUtf8);
1403 uprv_strcat(directoryBuffer, ICU_DATA_DIR_WINDOWS);
1404 return true;
1405 }
1406 }
1407 }
1408
1409 return false;
1410 }
1411 #endif
1412
dataDirectoryInitFn()1413 static void U_CALLCONV dataDirectoryInitFn() {
1414 /* If we already have the directory, then return immediately. Will happen if user called
1415 * u_setDataDirectory().
1416 */
1417 if (gDataDirectory) {
1418 return;
1419 }
1420
1421 const char *path = nullptr;
1422 #if defined(ICU_DATA_DIR_PREFIX_ENV_VAR)
1423 char datadir_path_buffer[PATH_MAX];
1424 #endif
1425
1426 /*
1427 When ICU_NO_USER_DATA_OVERRIDE is defined, users aren't allowed to
1428 override ICU's data with the ICU_DATA environment variable. This prevents
1429 problems where multiple custom copies of ICU's specific version of data
1430 are installed on a system. Either the application must define the data
1431 directory with u_setDataDirectory, define ICU_DATA_DIR when compiling
1432 ICU, set the data with udata_setCommonData or trust that all of the
1433 required data is contained in ICU's data library that contains
1434 the entry point defined by U_ICUDATA_ENTRY_POINT.
1435
1436 There may also be some platforms where environment variables
1437 are not allowed.
1438 */
1439 # if !defined(ICU_NO_USER_DATA_OVERRIDE) && !UCONFIG_NO_FILE_IO
1440 /* First try to get the environment variable */
1441 # if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP does not support getenv
1442 path=getenv("ICU_DATA");
1443 # endif
1444 # endif
1445
1446 /* ICU_DATA_DIR may be set as a compile option.
1447 * U_ICU_DATA_DEFAULT_DIR is provided and is set by ICU at compile time
1448 * and is used only when data is built in archive mode eliminating the need
1449 * for ICU_DATA_DIR to be set. U_ICU_DATA_DEFAULT_DIR is set to the installation
1450 * directory of the data dat file. Users should use ICU_DATA_DIR if they want to
1451 * set their own path.
1452 */
1453 #if defined(ICU_DATA_DIR) || defined(U_ICU_DATA_DEFAULT_DIR)
1454 if(path==nullptr || *path==0) {
1455 # if defined(ICU_DATA_DIR_PREFIX_ENV_VAR)
1456 const char *prefix = getenv(ICU_DATA_DIR_PREFIX_ENV_VAR);
1457 # endif
1458 # ifdef ICU_DATA_DIR
1459 path=ICU_DATA_DIR;
1460 # else
1461 path=U_ICU_DATA_DEFAULT_DIR;
1462 # endif
1463 # if defined(ICU_DATA_DIR_PREFIX_ENV_VAR)
1464 if (prefix != nullptr) {
1465 snprintf(datadir_path_buffer, sizeof(datadir_path_buffer), "%s%s", prefix, path);
1466 path=datadir_path_buffer;
1467 }
1468 # endif
1469 }
1470 #endif
1471
1472 #if defined(ICU_DATA_DIR_WINDOWS)
1473 char datadir_path_buffer[MAX_PATH];
1474 if (getIcuDataDirectoryUnderWindowsDirectory(datadir_path_buffer, UPRV_LENGTHOF(datadir_path_buffer))) {
1475 path = datadir_path_buffer;
1476 }
1477 #endif
1478
1479 if(path==nullptr) {
1480 /* It looks really bad, set it to something. */
1481 path = "";
1482 }
1483
1484 u_setDataDirectory(path);
1485 return;
1486 }
1487
1488 U_CAPI const char * U_EXPORT2
u_getDataDirectory()1489 u_getDataDirectory() {
1490 umtx_initOnce(gDataDirInitOnce, &dataDirectoryInitFn);
1491 return gDataDirectory;
1492 }
1493
setTimeZoneFilesDir(const char * path,UErrorCode & status)1494 static void setTimeZoneFilesDir(const char *path, UErrorCode &status) {
1495 if (U_FAILURE(status)) {
1496 return;
1497 }
1498 gTimeZoneFilesDirectory->clear();
1499 gTimeZoneFilesDirectory->append(path, status);
1500 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR)
1501 char *p = gTimeZoneFilesDirectory->data();
1502 while ((p = uprv_strchr(p, U_FILE_ALT_SEP_CHAR)) != nullptr) {
1503 *p = U_FILE_SEP_CHAR;
1504 }
1505 #endif
1506 }
1507
1508 #define TO_STRING(x) TO_STRING_2(x)
1509 #define TO_STRING_2(x) #x
1510
TimeZoneDataDirInitFn(UErrorCode & status)1511 static void U_CALLCONV TimeZoneDataDirInitFn(UErrorCode &status) {
1512 U_ASSERT(gTimeZoneFilesDirectory == nullptr);
1513 ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
1514 gTimeZoneFilesDirectory = new CharString();
1515 if (gTimeZoneFilesDirectory == nullptr) {
1516 status = U_MEMORY_ALLOCATION_ERROR;
1517 return;
1518 }
1519
1520 const char *dir = "";
1521
1522 #if defined(ICU_TIMEZONE_FILES_DIR_PREFIX_ENV_VAR)
1523 char timezonefilesdir_path_buffer[PATH_MAX];
1524 const char *prefix = getenv(ICU_TIMEZONE_FILES_DIR_PREFIX_ENV_VAR);
1525 #endif
1526
1527 #if U_PLATFORM_HAS_WINUWP_API == 1
1528 // The UWP version does not support the environment variable setting.
1529
1530 # if defined(ICU_DATA_DIR_WINDOWS)
1531 // When using the Windows system data, we can possibly pick up time zone data from the Windows directory.
1532 char datadir_path_buffer[MAX_PATH];
1533 if (getIcuDataDirectoryUnderWindowsDirectory(datadir_path_buffer, UPRV_LENGTHOF(datadir_path_buffer))) {
1534 dir = datadir_path_buffer;
1535 }
1536 # endif
1537
1538 #else
1539 dir = getenv("ICU_TIMEZONE_FILES_DIR");
1540 #endif // U_PLATFORM_HAS_WINUWP_API
1541
1542 #if defined(U_TIMEZONE_FILES_DIR)
1543 if (dir == nullptr) {
1544 // Build time configuration setting.
1545 dir = TO_STRING(U_TIMEZONE_FILES_DIR);
1546 }
1547 #endif
1548
1549 if (dir == nullptr) {
1550 dir = "";
1551 }
1552
1553 #if defined(ICU_TIMEZONE_FILES_DIR_PREFIX_ENV_VAR)
1554 if (prefix != nullptr) {
1555 snprintf(timezonefilesdir_path_buffer, sizeof(timezonefilesdir_path_buffer), "%s%s", prefix, dir);
1556 dir = timezonefilesdir_path_buffer;
1557 }
1558 #endif
1559
1560 setTimeZoneFilesDir(dir, status);
1561 }
1562
1563
1564 U_CAPI const char * U_EXPORT2
u_getTimeZoneFilesDirectory(UErrorCode * status)1565 u_getTimeZoneFilesDirectory(UErrorCode *status) {
1566 umtx_initOnce(gTimeZoneFilesInitOnce, &TimeZoneDataDirInitFn, *status);
1567 return U_SUCCESS(*status) ? gTimeZoneFilesDirectory->data() : "";
1568 }
1569
1570 U_CAPI void U_EXPORT2
u_setTimeZoneFilesDirectory(const char * path,UErrorCode * status)1571 u_setTimeZoneFilesDirectory(const char *path, UErrorCode *status) {
1572 umtx_initOnce(gTimeZoneFilesInitOnce, &TimeZoneDataDirInitFn, *status);
1573 setTimeZoneFilesDir(path, *status);
1574
1575 // Note: this function does some extra churn, first setting based on the
1576 // environment, then immediately replacing with the value passed in.
1577 // The logic is simpler that way, and performance shouldn't be an issue.
1578 }
1579
1580
1581 #if U_POSIX_LOCALE
1582 /* A helper function used by uprv_getPOSIXIDForDefaultLocale and
1583 * uprv_getPOSIXIDForDefaultCodepage. Returns the posix locale id for
1584 * LC_CTYPE and LC_MESSAGES. It doesn't support other locale categories.
1585 */
uprv_getPOSIXIDForCategory(int category)1586 static const char *uprv_getPOSIXIDForCategory(int category)
1587 {
1588 const char* posixID = nullptr;
1589 if (category == LC_MESSAGES || category == LC_CTYPE) {
1590 /*
1591 * On Solaris two different calls to setlocale can result in
1592 * different values. Only get this value once.
1593 *
1594 * We must check this first because an application can set this.
1595 *
1596 * LC_ALL can't be used because it's platform dependent. The LANG
1597 * environment variable seems to affect LC_CTYPE variable by default.
1598 * Here is what setlocale(LC_ALL, nullptr) can return.
1599 * HPUX can return 'C C C C C C C'
1600 * Solaris can return /en_US/C/C/C/C/C on the second try.
1601 * Linux can return LC_CTYPE=C;LC_NUMERIC=C;...
1602 *
1603 * The default codepage detection also needs to use LC_CTYPE.
1604 *
1605 * Do not call setlocale(LC_*, "")! Using an empty string instead
1606 * of nullptr, will modify the libc behavior.
1607 */
1608 posixID = setlocale(category, nullptr);
1609 if ((posixID == 0)
1610 || (uprv_strcmp("C", posixID) == 0)
1611 || (uprv_strcmp("POSIX", posixID) == 0))
1612 {
1613 /* Maybe we got some garbage. Try something more reasonable */
1614 posixID = getenv("LC_ALL");
1615 /* Solaris speaks POSIX - See IEEE Std 1003.1-2008
1616 * This is needed to properly handle empty env. variables
1617 */
1618 #if U_PLATFORM == U_PF_SOLARIS
1619 if ((posixID == 0) || (posixID[0] == '\0')) {
1620 posixID = getenv(category == LC_MESSAGES ? "LC_MESSAGES" : "LC_CTYPE");
1621 if ((posixID == 0) || (posixID[0] == '\0')) {
1622 #else
1623 if (posixID == 0) {
1624 posixID = getenv(category == LC_MESSAGES ? "LC_MESSAGES" : "LC_CTYPE");
1625 if (posixID == 0) {
1626 #endif
1627 posixID = getenv("LANG");
1628 }
1629 }
1630 }
1631 }
1632 if ((posixID==0)
1633 || (uprv_strcmp("C", posixID) == 0)
1634 || (uprv_strcmp("POSIX", posixID) == 0))
1635 {
1636 /* Nothing worked. Give it a nice POSIX default value. */
1637 posixID = "en_US_POSIX";
1638 // Note: this test will not catch 'C.UTF-8',
1639 // that will be handled in uprv_getDefaultLocaleID().
1640 // Leave this mapping here for the uprv_getPOSIXIDForDefaultCodepage()
1641 // caller which expects to see "en_US_POSIX" in many branches.
1642 }
1643 return posixID;
1644 }
1645
1646 /* Return just the POSIX id for the default locale, whatever happens to be in
1647 * it. It gets the value from LC_MESSAGES and indirectly from LC_ALL and LANG.
1648 */
1649 static const char *uprv_getPOSIXIDForDefaultLocale()
1650 {
1651 static const char* posixID = nullptr;
1652 if (posixID == 0) {
1653 posixID = uprv_getPOSIXIDForCategory(LC_MESSAGES);
1654 }
1655 return posixID;
1656 }
1657
1658 #if !U_CHARSET_IS_UTF8
1659 /* Return just the POSIX id for the default codepage, whatever happens to be in
1660 * it. It gets the value from LC_CTYPE and indirectly from LC_ALL and LANG.
1661 */
1662 static const char *uprv_getPOSIXIDForDefaultCodepage()
1663 {
1664 static const char* posixID = nullptr;
1665 if (posixID == 0) {
1666 posixID = uprv_getPOSIXIDForCategory(LC_CTYPE);
1667 }
1668 return posixID;
1669 }
1670 #endif
1671 #endif
1672
1673 /* NOTE: The caller should handle thread safety */
1674 U_CAPI const char* U_EXPORT2
1675 uprv_getDefaultLocaleID()
1676 {
1677 #if U_POSIX_LOCALE
1678 /*
1679 Note that: (a '!' means the ID is improper somehow)
1680 LC_ALL ----> default_loc codepage
1681 --------------------------------------------------------
1682 ab.CD ab CD
1683 ab@CD ab__CD -
1684 ab@CD.EF ab__CD EF
1685
1686 ab_CD.EF@GH ab_CD_GH EF
1687
1688 Some 'improper' ways to do the same as above:
1689 ! ab_CD@GH.EF ab_CD_GH EF
1690 ! ab_CD.EF@GH.IJ ab_CD_GH EF
1691 ! ab_CD@ZZ.EF@GH.IJ ab_CD_GH EF
1692
1693 _CD@GH _CD_GH -
1694 _CD.EF@GH _CD_GH EF
1695
1696 The variant cannot have dots in it.
1697 The 'rightmost' variant (@xxx) wins.
1698 The leftmost codepage (.xxx) wins.
1699 */
1700 const char* posixID = uprv_getPOSIXIDForDefaultLocale();
1701
1702 /* Format: (no spaces)
1703 ll [ _CC ] [ . MM ] [ @ VV]
1704
1705 l = lang, C = ctry, M = charmap, V = variant
1706 */
1707
1708 if (gCorrectedPOSIXLocale != nullptr) {
1709 return gCorrectedPOSIXLocale;
1710 }
1711
1712 // Copy the ID into owned memory.
1713 // Over-allocate in case we replace "C" with "en_US_POSIX" (+10), + null termination
1714 char *correctedPOSIXLocale = static_cast<char *>(uprv_malloc(uprv_strlen(posixID) + 10 + 1));
1715 if (correctedPOSIXLocale == nullptr) {
1716 return nullptr;
1717 }
1718 uprv_strcpy(correctedPOSIXLocale, posixID);
1719
1720 char *limit;
1721 if ((limit = uprv_strchr(correctedPOSIXLocale, '.')) != nullptr) {
1722 *limit = 0;
1723 }
1724 if ((limit = uprv_strchr(correctedPOSIXLocale, '@')) != nullptr) {
1725 *limit = 0;
1726 }
1727
1728 if ((uprv_strcmp("C", correctedPOSIXLocale) == 0) // no @ variant
1729 || (uprv_strcmp("POSIX", correctedPOSIXLocale) == 0)) {
1730 // Raw input was C.* or POSIX.*, Give it a nice POSIX default value.
1731 // (The "C"/"POSIX" case is handled in uprv_getPOSIXIDForCategory())
1732 uprv_strcpy(correctedPOSIXLocale, "en_US_POSIX");
1733 }
1734
1735 /* Note that we scan the *uncorrected* ID. */
1736 const char *p;
1737 if ((p = uprv_strrchr(posixID, '@')) != nullptr) {
1738 p++;
1739
1740 /* Take care of any special cases here.. */
1741 if (!uprv_strcmp(p, "nynorsk")) {
1742 p = "NY";
1743 /* Don't worry about no__NY. In practice, it won't appear. */
1744 }
1745
1746 if (uprv_strchr(correctedPOSIXLocale,'_') == nullptr) {
1747 uprv_strcat(correctedPOSIXLocale, "__"); /* aa@b -> aa__b (note this can make the new locale 1 char longer) */
1748 }
1749 else {
1750 uprv_strcat(correctedPOSIXLocale, "_"); /* aa_CC@b -> aa_CC_b */
1751 }
1752
1753 const char *q;
1754 if ((q = uprv_strchr(p, '.')) != nullptr) {
1755 /* How big will the resulting string be? */
1756 int32_t len = (int32_t)(uprv_strlen(correctedPOSIXLocale) + (q-p));
1757 uprv_strncat(correctedPOSIXLocale, p, q-p); // do not include charset
1758 correctedPOSIXLocale[len] = 0;
1759 }
1760 else {
1761 /* Anything following the @ sign */
1762 uprv_strcat(correctedPOSIXLocale, p);
1763 }
1764
1765 /* Should there be a map from 'no@nynorsk' -> no_NO_NY here?
1766 * How about 'russian' -> 'ru'?
1767 * Many of the other locales using ISO codes will be handled by the
1768 * canonicalization functions in uloc_getDefault.
1769 */
1770 }
1771
1772 if (gCorrectedPOSIXLocale == nullptr) {
1773 gCorrectedPOSIXLocale = correctedPOSIXLocale;
1774 gCorrectedPOSIXLocaleHeapAllocated = true;
1775 ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
1776 correctedPOSIXLocale = nullptr;
1777 }
1778 posixID = gCorrectedPOSIXLocale;
1779
1780 if (correctedPOSIXLocale != nullptr) { /* Was already set - clean up. */
1781 uprv_free(correctedPOSIXLocale);
1782 }
1783
1784 return posixID;
1785
1786 #elif U_PLATFORM_USES_ONLY_WIN32_API
1787 #define POSIX_LOCALE_CAPACITY 64
1788 UErrorCode status = U_ZERO_ERROR;
1789 char *correctedPOSIXLocale = nullptr;
1790
1791 // If we have already figured this out just use the cached value
1792 if (gCorrectedPOSIXLocale != nullptr) {
1793 return gCorrectedPOSIXLocale;
1794 }
1795
1796 // No cached value, need to determine the current value
1797 static WCHAR windowsLocale[LOCALE_NAME_MAX_LENGTH] = {};
1798 int length = GetLocaleInfoEx(LOCALE_NAME_USER_DEFAULT, LOCALE_SNAME, windowsLocale, LOCALE_NAME_MAX_LENGTH);
1799
1800 // Now we should have a Windows locale name that needs converted to the POSIX style.
1801 if (length > 0) // If length is 0, then the GetLocaleInfoEx failed.
1802 {
1803 // First we need to go from UTF-16 to char (and also convert from _ to - while we're at it.)
1804 char modifiedWindowsLocale[LOCALE_NAME_MAX_LENGTH] = {};
1805
1806 int32_t i;
1807 for (i = 0; i < UPRV_LENGTHOF(modifiedWindowsLocale); i++)
1808 {
1809 if (windowsLocale[i] == '_')
1810 {
1811 modifiedWindowsLocale[i] = '-';
1812 }
1813 else
1814 {
1815 modifiedWindowsLocale[i] = static_cast<char>(windowsLocale[i]);
1816 }
1817
1818 if (modifiedWindowsLocale[i] == '\0')
1819 {
1820 break;
1821 }
1822 }
1823
1824 if (i >= UPRV_LENGTHOF(modifiedWindowsLocale))
1825 {
1826 // Ran out of room, can't really happen, maybe we'll be lucky about a matching
1827 // locale when tags are dropped
1828 modifiedWindowsLocale[UPRV_LENGTHOF(modifiedWindowsLocale) - 1] = '\0';
1829 }
1830
1831 // Now normalize the resulting name
1832 correctedPOSIXLocale = static_cast<char *>(uprv_malloc(POSIX_LOCALE_CAPACITY + 1));
1833 /* TODO: Should we just exit on memory allocation failure? */
1834 if (correctedPOSIXLocale)
1835 {
1836 int32_t posixLen = uloc_canonicalize(modifiedWindowsLocale, correctedPOSIXLocale, POSIX_LOCALE_CAPACITY, &status);
1837 if (U_SUCCESS(status))
1838 {
1839 *(correctedPOSIXLocale + posixLen) = 0;
1840 gCorrectedPOSIXLocale = correctedPOSIXLocale;
1841 gCorrectedPOSIXLocaleHeapAllocated = true;
1842 ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
1843 }
1844 else
1845 {
1846 uprv_free(correctedPOSIXLocale);
1847 }
1848 }
1849 }
1850
1851 // If unable to find a locale we can agree upon, use en-US by default
1852 if (gCorrectedPOSIXLocale == nullptr) {
1853 gCorrectedPOSIXLocale = "en_US";
1854 }
1855 return gCorrectedPOSIXLocale;
1856
1857 #elif U_PLATFORM == U_PF_OS400
1858 /* locales are process scoped and are by definition thread safe */
1859 static char correctedLocale[64];
1860 const char *localeID = getenv("LC_ALL");
1861 char *p;
1862
1863 if (localeID == nullptr)
1864 localeID = getenv("LANG");
1865 if (localeID == nullptr)
1866 localeID = setlocale(LC_ALL, nullptr);
1867 /* Make sure we have something... */
1868 if (localeID == nullptr)
1869 return "en_US_POSIX";
1870
1871 /* Extract the locale name from the path. */
1872 if((p = uprv_strrchr(localeID, '/')) != nullptr)
1873 {
1874 /* Increment p to start of locale name. */
1875 p++;
1876 localeID = p;
1877 }
1878
1879 /* Copy to work location. */
1880 uprv_strcpy(correctedLocale, localeID);
1881
1882 /* Strip off the '.locale' extension. */
1883 if((p = uprv_strchr(correctedLocale, '.')) != nullptr) {
1884 *p = 0;
1885 }
1886
1887 /* Upper case the locale name. */
1888 T_CString_toUpperCase(correctedLocale);
1889
1890 /* See if we are using the POSIX locale. Any of the
1891 * following are equivalent and use the same QLGPGCMA
1892 * (POSIX) locale.
1893 * QLGPGCMA2 means UCS2
1894 * QLGPGCMA_4 means UTF-32
1895 * QLGPGCMA_8 means UTF-8
1896 */
1897 if ((uprv_strcmp("C", correctedLocale) == 0) ||
1898 (uprv_strcmp("POSIX", correctedLocale) == 0) ||
1899 (uprv_strncmp("QLGPGCMA", correctedLocale, 8) == 0))
1900 {
1901 uprv_strcpy(correctedLocale, "en_US_POSIX");
1902 }
1903 else
1904 {
1905 int16_t LocaleLen;
1906
1907 /* Lower case the lang portion. */
1908 for(p = correctedLocale; *p != 0 && *p != '_'; p++)
1909 {
1910 *p = uprv_tolower(*p);
1911 }
1912
1913 /* Adjust for Euro. After '_E' add 'URO'. */
1914 LocaleLen = uprv_strlen(correctedLocale);
1915 if (correctedLocale[LocaleLen - 2] == '_' &&
1916 correctedLocale[LocaleLen - 1] == 'E')
1917 {
1918 uprv_strcat(correctedLocale, "URO");
1919 }
1920
1921 /* If using Lotus-based locale then convert to
1922 * equivalent non Lotus.
1923 */
1924 else if (correctedLocale[LocaleLen - 2] == '_' &&
1925 correctedLocale[LocaleLen - 1] == 'L')
1926 {
1927 correctedLocale[LocaleLen - 2] = 0;
1928 }
1929
1930 /* There are separate simplified and traditional
1931 * locales called zh_HK_S and zh_HK_T.
1932 */
1933 else if (uprv_strncmp(correctedLocale, "zh_HK", 5) == 0)
1934 {
1935 uprv_strcpy(correctedLocale, "zh_HK");
1936 }
1937
1938 /* A special zh_CN_GBK locale...
1939 */
1940 else if (uprv_strcmp(correctedLocale, "zh_CN_GBK") == 0)
1941 {
1942 uprv_strcpy(correctedLocale, "zh_CN");
1943 }
1944
1945 }
1946
1947 return correctedLocale;
1948 #endif
1949
1950 }
1951
1952 #if !U_CHARSET_IS_UTF8
1953 #if U_POSIX_LOCALE
1954 /*
1955 Due to various platform differences, one platform may specify a charset,
1956 when they really mean a different charset. Remap the names so that they are
1957 compatible with ICU. Only conflicting/ambiguous aliases should be resolved
1958 here. Before adding anything to this function, please consider adding unique
1959 names to the ICU alias table in the data directory.
1960 */
1961 static const char*
1962 remapPlatformDependentCodepage(const char *locale, const char *name) {
1963 if (locale != nullptr && *locale == 0) {
1964 /* Make sure that an empty locale is handled the same way. */
1965 locale = nullptr;
1966 }
1967 if (name == nullptr) {
1968 return nullptr;
1969 }
1970 #if U_PLATFORM == U_PF_AIX
1971 if (uprv_strcmp(name, "IBM-943") == 0) {
1972 /* Use the ASCII compatible ibm-943 */
1973 name = "Shift-JIS";
1974 }
1975 else if (uprv_strcmp(name, "IBM-1252") == 0) {
1976 /* Use the windows-1252 that contains the Euro */
1977 name = "IBM-5348";
1978 }
1979 #elif U_PLATFORM == U_PF_SOLARIS
1980 if (locale != nullptr && uprv_strcmp(name, "EUC") == 0) {
1981 /* Solaris underspecifies the "EUC" name. */
1982 if (uprv_strcmp(locale, "zh_CN") == 0) {
1983 name = "EUC-CN";
1984 }
1985 else if (uprv_strcmp(locale, "zh_TW") == 0) {
1986 name = "EUC-TW";
1987 }
1988 else if (uprv_strcmp(locale, "ko_KR") == 0) {
1989 name = "EUC-KR";
1990 }
1991 }
1992 else if (uprv_strcmp(name, "eucJP") == 0) {
1993 /*
1994 ibm-954 is the best match.
1995 ibm-33722 is the default for eucJP (similar to Windows).
1996 */
1997 name = "eucjis";
1998 }
1999 else if (uprv_strcmp(name, "646") == 0) {
2000 /*
2001 * The default codepage given by Solaris is 646 but the C library routines treat it as if it was
2002 * ISO-8859-1 instead of US-ASCII(646).
2003 */
2004 name = "ISO-8859-1";
2005 }
2006 #elif U_PLATFORM_IS_DARWIN_BASED
2007 if (locale == nullptr && *name == 0) {
2008 /*
2009 No locale was specified, and an empty name was passed in.
2010 This usually indicates that nl_langinfo didn't return valid information.
2011 Mac OS X uses UTF-8 by default (especially the locale data and console).
2012 */
2013 name = "UTF-8";
2014 }
2015 else if (uprv_strcmp(name, "CP949") == 0) {
2016 /* Remap CP949 to a similar codepage to avoid issues with backslash and won symbol. */
2017 name = "EUC-KR";
2018 }
2019 else if (locale != nullptr && uprv_strcmp(locale, "en_US_POSIX") != 0 && uprv_strcmp(name, "US-ASCII") == 0) {
2020 /*
2021 * For non C/POSIX locale, default the code page to UTF-8 instead of US-ASCII.
2022 */
2023 name = "UTF-8";
2024 }
2025 #elif U_PLATFORM == U_PF_BSD
2026 if (uprv_strcmp(name, "CP949") == 0) {
2027 /* Remap CP949 to a similar codepage to avoid issues with backslash and won symbol. */
2028 name = "EUC-KR";
2029 }
2030 #elif U_PLATFORM == U_PF_HPUX
2031 if (locale != nullptr && uprv_strcmp(locale, "zh_HK") == 0 && uprv_strcmp(name, "big5") == 0) {
2032 /* HP decided to extend big5 as hkbig5 even though it's not compatible :-( */
2033 /* zh_TW.big5 is not the same charset as zh_HK.big5! */
2034 name = "hkbig5";
2035 }
2036 else if (uprv_strcmp(name, "eucJP") == 0) {
2037 /*
2038 ibm-1350 is the best match, but unavailable.
2039 ibm-954 is mostly a superset of ibm-1350.
2040 ibm-33722 is the default for eucJP (similar to Windows).
2041 */
2042 name = "eucjis";
2043 }
2044 #elif U_PLATFORM == U_PF_LINUX
2045 if (locale != nullptr && uprv_strcmp(name, "euc") == 0) {
2046 /* Linux underspecifies the "EUC" name. */
2047 if (uprv_strcmp(locale, "korean") == 0) {
2048 name = "EUC-KR";
2049 }
2050 else if (uprv_strcmp(locale, "japanese") == 0) {
2051 /* See comment below about eucJP */
2052 name = "eucjis";
2053 }
2054 }
2055 else if (uprv_strcmp(name, "eucjp") == 0) {
2056 /*
2057 ibm-1350 is the best match, but unavailable.
2058 ibm-954 is mostly a superset of ibm-1350.
2059 ibm-33722 is the default for eucJP (similar to Windows).
2060 */
2061 name = "eucjis";
2062 }
2063 else if (locale != nullptr && uprv_strcmp(locale, "en_US_POSIX") != 0 &&
2064 (uprv_strcmp(name, "ANSI_X3.4-1968") == 0 || uprv_strcmp(name, "US-ASCII") == 0)) {
2065 /*
2066 * For non C/POSIX locale, default the code page to UTF-8 instead of US-ASCII.
2067 */
2068 name = "UTF-8";
2069 }
2070 /*
2071 * Linux returns ANSI_X3.4-1968 for C/POSIX, but the call site takes care of
2072 * it by falling back to 'US-ASCII' when nullptr is returned from this
2073 * function. So, we don't have to worry about it here.
2074 */
2075 #endif
2076 /* return nullptr when "" is passed in */
2077 if (*name == 0) {
2078 name = nullptr;
2079 }
2080 return name;
2081 }
2082
2083 static const char*
2084 getCodepageFromPOSIXID(const char *localeName, char * buffer, int32_t buffCapacity)
2085 {
2086 char localeBuf[100];
2087 const char *name = nullptr;
2088 char *variant = nullptr;
2089
2090 if (localeName != nullptr && (name = (uprv_strchr(localeName, '.'))) != nullptr) {
2091 size_t localeCapacity = uprv_min(sizeof(localeBuf), (name-localeName)+1);
2092 uprv_strncpy(localeBuf, localeName, localeCapacity);
2093 localeBuf[localeCapacity-1] = 0; /* ensure NUL termination */
2094 name = uprv_strncpy(buffer, name+1, buffCapacity);
2095 buffer[buffCapacity-1] = 0; /* ensure NUL termination */
2096 if ((variant = const_cast<char *>(uprv_strchr(name, '@'))) != nullptr) {
2097 *variant = 0;
2098 }
2099 name = remapPlatformDependentCodepage(localeBuf, name);
2100 }
2101 return name;
2102 }
2103 #endif
2104
2105 static const char*
2106 int_getDefaultCodepage()
2107 {
2108 #if U_PLATFORM == U_PF_OS400
2109 uint32_t ccsid = 37; /* Default to ibm-37 */
2110 static char codepage[64];
2111 Qwc_JOBI0400_t jobinfo;
2112 Qus_EC_t error = { sizeof(Qus_EC_t) }; /* SPI error code */
2113
2114 EPT_CALL(QUSRJOBI)(&jobinfo, sizeof(jobinfo), "JOBI0400",
2115 "* ", " ", &error);
2116
2117 if (error.Bytes_Available == 0) {
2118 if (jobinfo.Coded_Char_Set_ID != 0xFFFF) {
2119 ccsid = (uint32_t)jobinfo.Coded_Char_Set_ID;
2120 }
2121 else if (jobinfo.Default_Coded_Char_Set_Id != 0xFFFF) {
2122 ccsid = (uint32_t)jobinfo.Default_Coded_Char_Set_Id;
2123 }
2124 /* else use the default */
2125 }
2126 snprintf(codepage, sizeof(codepage), "ibm-%d", ccsid);
2127 return codepage;
2128
2129 #elif U_PLATFORM == U_PF_OS390
2130 static char codepage[64];
2131
2132 strncpy(codepage, nl_langinfo(CODESET),63-strlen(UCNV_SWAP_LFNL_OPTION_STRING));
2133 strcat(codepage,UCNV_SWAP_LFNL_OPTION_STRING);
2134 codepage[63] = 0; /* NUL terminate */
2135
2136 return codepage;
2137
2138 #elif U_PLATFORM_USES_ONLY_WIN32_API
2139 static char codepage[64];
2140 DWORD codepageNumber = 0;
2141
2142 #if U_PLATFORM_HAS_WINUWP_API == 1
2143 // UWP doesn't have a direct API to get the default ACP as Microsoft would rather
2144 // have folks use Unicode than a "system" code page, however this is the same
2145 // codepage as the system default locale codepage. (FWIW, the system locale is
2146 // ONLY used for codepage, it should never be used for anything else)
2147 GetLocaleInfoEx(LOCALE_NAME_SYSTEM_DEFAULT, LOCALE_IDEFAULTANSICODEPAGE | LOCALE_RETURN_NUMBER,
2148 (LPWSTR)&codepageNumber, sizeof(codepageNumber) / sizeof(WCHAR));
2149 #else
2150 // Win32 apps can call GetACP
2151 codepageNumber = GetACP();
2152 #endif
2153 // Special case for UTF-8
2154 if (codepageNumber == 65001)
2155 {
2156 return "UTF-8";
2157 }
2158 // Windows codepages can look like windows-1252, so format the found number
2159 // the numbers are eclectic, however all valid system code pages, besides UTF-8
2160 // are between 3 and 19999
2161 if (codepageNumber > 0 && codepageNumber < 20000)
2162 {
2163 snprintf(codepage, sizeof(codepage), "windows-%ld", codepageNumber);
2164 return codepage;
2165 }
2166 // If the codepage number call failed then return UTF-8
2167 return "UTF-8";
2168
2169 #elif U_POSIX_LOCALE
2170 static char codesetName[100];
2171 const char *localeName = nullptr;
2172 const char *name = nullptr;
2173
2174 localeName = uprv_getPOSIXIDForDefaultCodepage();
2175 uprv_memset(codesetName, 0, sizeof(codesetName));
2176 /* On Solaris nl_langinfo returns C locale values unless setlocale
2177 * was called earlier.
2178 */
2179 #if (U_HAVE_NL_LANGINFO_CODESET && U_PLATFORM != U_PF_SOLARIS)
2180 /* When available, check nl_langinfo first because it usually gives more
2181 useful names. It depends on LC_CTYPE.
2182 nl_langinfo may use the same buffer as setlocale. */
2183 {
2184 const char *codeset = nl_langinfo(U_NL_LANGINFO_CODESET);
2185 #if U_PLATFORM_IS_DARWIN_BASED || U_PLATFORM_IS_LINUX_BASED
2186 /*
2187 * On Linux and MacOSX, ensure that default codepage for non C/POSIX locale is UTF-8
2188 * instead of ASCII.
2189 */
2190 if (uprv_strcmp(localeName, "en_US_POSIX") != 0) {
2191 codeset = remapPlatformDependentCodepage(localeName, codeset);
2192 } else
2193 #endif
2194 {
2195 codeset = remapPlatformDependentCodepage(nullptr, codeset);
2196 }
2197
2198 if (codeset != nullptr) {
2199 uprv_strncpy(codesetName, codeset, sizeof(codesetName));
2200 codesetName[sizeof(codesetName)-1] = 0;
2201 return codesetName;
2202 }
2203 }
2204 #endif
2205
2206 /* Use setlocale in a nice way, and then check some environment variables.
2207 Maybe the application used setlocale already.
2208 */
2209 uprv_memset(codesetName, 0, sizeof(codesetName));
2210 name = getCodepageFromPOSIXID(localeName, codesetName, sizeof(codesetName));
2211 if (name) {
2212 /* if we can find the codeset name from setlocale, return that. */
2213 return name;
2214 }
2215
2216 if (*codesetName == 0)
2217 {
2218 /* Everything failed. Return US ASCII (ISO 646). */
2219 (void)uprv_strcpy(codesetName, "US-ASCII");
2220 }
2221 return codesetName;
2222 #else
2223 return "US-ASCII";
2224 #endif
2225 }
2226
2227
2228 U_CAPI const char* U_EXPORT2
2229 uprv_getDefaultCodepage()
2230 {
2231 static char const *name = nullptr;
2232 umtx_lock(nullptr);
2233 if (name == nullptr) {
2234 name = int_getDefaultCodepage();
2235 }
2236 umtx_unlock(nullptr);
2237 return name;
2238 }
2239 #endif /* !U_CHARSET_IS_UTF8 */
2240
2241
2242 /* end of platform-specific implementation -------------- */
2243
2244 /* version handling --------------------------------------------------------- */
2245
2246 U_CAPI void U_EXPORT2
2247 u_versionFromString(UVersionInfo versionArray, const char *versionString) {
2248 char *end;
2249 uint16_t part=0;
2250
2251 if(versionArray==nullptr) {
2252 return;
2253 }
2254
2255 if(versionString!=nullptr) {
2256 for(;;) {
2257 versionArray[part]=(uint8_t)uprv_strtoul(versionString, &end, 10);
2258 if(end==versionString || ++part==U_MAX_VERSION_LENGTH || *end!=U_VERSION_DELIMITER) {
2259 break;
2260 }
2261 versionString=end+1;
2262 }
2263 }
2264
2265 while(part<U_MAX_VERSION_LENGTH) {
2266 versionArray[part++]=0;
2267 }
2268 }
2269
2270 U_CAPI void U_EXPORT2
2271 u_versionFromUString(UVersionInfo versionArray, const char16_t *versionString) {
2272 if(versionArray!=nullptr && versionString!=nullptr) {
2273 char versionChars[U_MAX_VERSION_STRING_LENGTH+1];
2274 int32_t len = u_strlen(versionString);
2275 if(len>U_MAX_VERSION_STRING_LENGTH) {
2276 len = U_MAX_VERSION_STRING_LENGTH;
2277 }
2278 u_UCharsToChars(versionString, versionChars, len);
2279 versionChars[len]=0;
2280 u_versionFromString(versionArray, versionChars);
2281 }
2282 }
2283
2284 U_CAPI void U_EXPORT2
2285 u_versionToString(const UVersionInfo versionArray, char *versionString) {
2286 uint16_t count, part;
2287 uint8_t field;
2288
2289 if(versionString==nullptr) {
2290 return;
2291 }
2292
2293 if(versionArray==nullptr) {
2294 versionString[0]=0;
2295 return;
2296 }
2297
2298 /* count how many fields need to be written */
2299 for(count=4; count>0 && versionArray[count-1]==0; --count) {
2300 }
2301
2302 if(count <= 1) {
2303 count = 2;
2304 }
2305
2306 /* write the first part */
2307 /* write the decimal field value */
2308 field=versionArray[0];
2309 if(field>=100) {
2310 *versionString++=(char)('0'+field/100);
2311 field%=100;
2312 }
2313 if(field>=10) {
2314 *versionString++=(char)('0'+field/10);
2315 field%=10;
2316 }
2317 *versionString++=(char)('0'+field);
2318
2319 /* write the following parts */
2320 for(part=1; part<count; ++part) {
2321 /* write a dot first */
2322 *versionString++=U_VERSION_DELIMITER;
2323
2324 /* write the decimal field value */
2325 field=versionArray[part];
2326 if(field>=100) {
2327 *versionString++=(char)('0'+field/100);
2328 field%=100;
2329 }
2330 if(field>=10) {
2331 *versionString++=(char)('0'+field/10);
2332 field%=10;
2333 }
2334 *versionString++=(char)('0'+field);
2335 }
2336
2337 /* NUL-terminate */
2338 *versionString=0;
2339 }
2340
2341 U_CAPI void U_EXPORT2
2342 u_getVersion(UVersionInfo versionArray) {
2343 (void)copyright; // Suppress unused variable warning from clang.
2344 u_versionFromString(versionArray, U_ICU_VERSION);
2345 }
2346
2347 /**
2348 * icucfg.h dependent code
2349 */
2350
2351 #if U_ENABLE_DYLOAD && HAVE_DLOPEN && !U_PLATFORM_USES_ONLY_WIN32_API
2352
2353 #if HAVE_DLFCN_H
2354 #ifdef __MVS__
2355 #ifndef __SUSV3
2356 #define __SUSV3 1
2357 #endif
2358 #endif
2359 #include <dlfcn.h>
2360 #endif /* HAVE_DLFCN_H */
2361
2362 U_CAPI void * U_EXPORT2
2363 uprv_dl_open(const char *libName, UErrorCode *status) {
2364 void *ret = nullptr;
2365 if(U_FAILURE(*status)) return ret;
2366 ret = dlopen(libName, RTLD_NOW|RTLD_GLOBAL);
2367 if(ret==nullptr) {
2368 #ifdef U_TRACE_DYLOAD
2369 printf("dlerror on dlopen(%s): %s\n", libName, dlerror());
2370 #endif
2371 *status = U_MISSING_RESOURCE_ERROR;
2372 }
2373 return ret;
2374 }
2375
2376 U_CAPI void U_EXPORT2
2377 uprv_dl_close(void *lib, UErrorCode *status) {
2378 if(U_FAILURE(*status)) return;
2379 dlclose(lib);
2380 }
2381
2382 U_CAPI UVoidFunction* U_EXPORT2
2383 uprv_dlsym_func(void *lib, const char* sym, UErrorCode *status) {
2384 union {
2385 UVoidFunction *fp;
2386 void *vp;
2387 } uret;
2388 uret.fp = nullptr;
2389 if(U_FAILURE(*status)) return uret.fp;
2390 uret.vp = dlsym(lib, sym);
2391 if(uret.vp == nullptr) {
2392 #ifdef U_TRACE_DYLOAD
2393 printf("dlerror on dlsym(%p,%s): %s\n", lib,sym, dlerror());
2394 #endif
2395 *status = U_MISSING_RESOURCE_ERROR;
2396 }
2397 return uret.fp;
2398 }
2399
2400 #elif U_ENABLE_DYLOAD && U_PLATFORM_USES_ONLY_WIN32_API && !U_PLATFORM_HAS_WINUWP_API
2401
2402 /* Windows API implementation. */
2403 // Note: UWP does not expose/allow these APIs, so the UWP version gets the null implementation. */
2404
2405 U_CAPI void * U_EXPORT2
2406 uprv_dl_open(const char *libName, UErrorCode *status) {
2407 HMODULE lib = nullptr;
2408
2409 if(U_FAILURE(*status)) return nullptr;
2410
2411 lib = LoadLibraryA(libName);
2412
2413 if(lib==nullptr) {
2414 *status = U_MISSING_RESOURCE_ERROR;
2415 }
2416
2417 return (void*)lib;
2418 }
2419
2420 U_CAPI void U_EXPORT2
2421 uprv_dl_close(void *lib, UErrorCode *status) {
2422 HMODULE handle = (HMODULE)lib;
2423 if(U_FAILURE(*status)) return;
2424
2425 FreeLibrary(handle);
2426
2427 return;
2428 }
2429
2430 U_CAPI UVoidFunction* U_EXPORT2
2431 uprv_dlsym_func(void *lib, const char* sym, UErrorCode *status) {
2432 HMODULE handle = (HMODULE)lib;
2433 UVoidFunction* addr = nullptr;
2434
2435 if(U_FAILURE(*status) || lib==nullptr) return nullptr;
2436
2437 addr = (UVoidFunction*)GetProcAddress(handle, sym);
2438
2439 if(addr==nullptr) {
2440 DWORD lastError = GetLastError();
2441 if(lastError == ERROR_PROC_NOT_FOUND) {
2442 *status = U_MISSING_RESOURCE_ERROR;
2443 } else {
2444 *status = U_UNSUPPORTED_ERROR; /* other unknown error. */
2445 }
2446 }
2447
2448 return addr;
2449 }
2450
2451 #else
2452
2453 /* No dynamic loading, null (nonexistent) implementation. */
2454
2455 U_CAPI void * U_EXPORT2
2456 uprv_dl_open(const char *libName, UErrorCode *status) {
2457 (void)libName;
2458 if(U_FAILURE(*status)) return nullptr;
2459 *status = U_UNSUPPORTED_ERROR;
2460 return nullptr;
2461 }
2462
2463 U_CAPI void U_EXPORT2
2464 uprv_dl_close(void *lib, UErrorCode *status) {
2465 (void)lib;
2466 if(U_FAILURE(*status)) return;
2467 *status = U_UNSUPPORTED_ERROR;
2468 return;
2469 }
2470
2471 U_CAPI UVoidFunction* U_EXPORT2
2472 uprv_dlsym_func(void *lib, const char* sym, UErrorCode *status) {
2473 (void)lib;
2474 (void)sym;
2475 if(U_SUCCESS(*status)) {
2476 *status = U_UNSUPPORTED_ERROR;
2477 }
2478 return (UVoidFunction*)nullptr;
2479 }
2480
2481 #endif
2482
2483 /*
2484 * Hey, Emacs, please set the following:
2485 *
2486 * Local Variables:
2487 * indent-tabs-mode: nil
2488 * End:
2489 *
2490 */
2491