• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 const 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         char *ret = realpath(TZDEFAULT, gTimeZoneBuffer);
1175         if (ret != nullptr && uprv_strcmp(TZDEFAULT, gTimeZoneBuffer) != 0) {
1176             int32_t tzZoneInfoTailLen = uprv_strlen(TZZONEINFOTAIL);
1177             const char *tzZoneInfoTailPtr = uprv_strstr(gTimeZoneBuffer, TZZONEINFOTAIL);
1178             // MacOS14 has the realpath as something like
1179             // /usr/share/zoneinfo.default/Australia/Melbourne
1180             // which will not have "/zoneinfo/" in the path.
1181             // Therefore if we fail, we fall back to read the link which is
1182             // /var/db/timezone/zoneinfo/Australia/Melbourne
1183             // We also fall back to reading the link if the realpath leads to something like
1184             // /usr/share/zoneinfo/posixrules
1185             if (tzZoneInfoTailPtr == nullptr ||
1186                     uprv_strcmp(tzZoneInfoTailPtr + tzZoneInfoTailLen, "posixrules") == 0) {
1187                 ssize_t size = readlink(TZDEFAULT, gTimeZoneBuffer, sizeof(gTimeZoneBuffer)-1);
1188                 if (size > 0) {
1189                     gTimeZoneBuffer[size] = 0;
1190                     tzZoneInfoTailPtr = uprv_strstr(gTimeZoneBuffer, TZZONEINFOTAIL);
1191                 }
1192             }
1193             if (tzZoneInfoTailPtr != nullptr) {
1194                 tzZoneInfoTailPtr += tzZoneInfoTailLen;
1195                 skipZoneIDPrefix(&tzZoneInfoTailPtr);
1196                 if (isValidOlsonID(tzZoneInfoTailPtr)) {
1197                     return (gTimeZoneBufferPtr = tzZoneInfoTailPtr);
1198                 }
1199             }
1200         } else {
1201 #if defined(SEARCH_TZFILE)
1202             DefaultTZInfo* tzInfo = (DefaultTZInfo*)uprv_malloc(sizeof(DefaultTZInfo));
1203             if (tzInfo != nullptr) {
1204                 tzInfo->defaultTZBuffer = nullptr;
1205                 tzInfo->defaultTZFileSize = 0;
1206                 tzInfo->defaultTZFilePtr = nullptr;
1207                 tzInfo->defaultTZstatus = false;
1208                 tzInfo->defaultTZPosition = 0;
1209 
1210                 gTimeZoneBufferPtr = searchForTZFile(TZZONEINFO, tzInfo);
1211 
1212                 /* Free previously allocated memory */
1213                 if (tzInfo->defaultTZBuffer != nullptr) {
1214                     uprv_free(tzInfo->defaultTZBuffer);
1215                 }
1216                 if (tzInfo->defaultTZFilePtr != nullptr) {
1217                     fclose(tzInfo->defaultTZFilePtr);
1218                 }
1219                 uprv_free(tzInfo);
1220             }
1221 
1222             if (gTimeZoneBufferPtr != nullptr && isValidOlsonID(gTimeZoneBufferPtr)) {
1223                 return gTimeZoneBufferPtr;
1224             }
1225 #endif
1226         }
1227     }
1228     else {
1229         return gTimeZoneBufferPtr;
1230     }
1231 #endif
1232 #endif
1233 
1234 #ifdef U_TZNAME
1235 #if U_PLATFORM_USES_ONLY_WIN32_API
1236     /* The return value is free'd in timezone.cpp on Windows because
1237      * the other code path returns a pointer to a heap location. */
1238     return uprv_strdup(U_TZNAME[n]);
1239 #else
1240     /*
1241     U_TZNAME is usually a non-unique abbreviation, which isn't normally usable.
1242     So we remap the abbreviation to an olson ID.
1243 
1244     Since Windows exposes a little more timezone information,
1245     we normally don't use this code on Windows because
1246     uprv_detectWindowsTimeZone should have already given the correct answer.
1247     */
1248     {
1249         struct tm juneSol, decemberSol;
1250         int daylightType;
1251         static const time_t juneSolstice=1182478260; /*2007-06-21 18:11 UT*/
1252         static const time_t decemberSolstice=1198332540; /*2007-12-22 06:09 UT*/
1253 
1254         /* This probing will tell us when daylight savings occurs.  */
1255         localtime_r(&juneSolstice, &juneSol);
1256         localtime_r(&decemberSolstice, &decemberSol);
1257         if(decemberSol.tm_isdst > 0) {
1258           daylightType = U_DAYLIGHT_DECEMBER;
1259         } else if(juneSol.tm_isdst > 0) {
1260           daylightType = U_DAYLIGHT_JUNE;
1261         } else {
1262           daylightType = U_DAYLIGHT_NONE;
1263         }
1264         tzid = remapShortTimeZone(U_TZNAME[0], U_TZNAME[1], daylightType, uprv_timezone());
1265         if (tzid != nullptr) {
1266             return tzid;
1267         }
1268     }
1269     return U_TZNAME[n];
1270 #endif
1271 #else
1272     return "";
1273 #endif
1274 }
1275 
1276 /* Get and set the ICU data directory --------------------------------------- */
1277 
1278 static icu::UInitOnce gDataDirInitOnce {};
1279 static char *gDataDirectory = nullptr;
1280 
1281 UInitOnce gTimeZoneFilesInitOnce {};
1282 static CharString *gTimeZoneFilesDirectory = nullptr;
1283 
1284 #if U_POSIX_LOCALE || U_PLATFORM_USES_ONLY_WIN32_API
1285  static const char *gCorrectedPOSIXLocale = nullptr; /* Sometimes heap allocated */
1286  static bool gCorrectedPOSIXLocaleHeapAllocated = false;
1287 #endif
1288 
putil_cleanup()1289 static UBool U_CALLCONV putil_cleanup()
1290 {
1291     if (gDataDirectory && *gDataDirectory) {
1292         uprv_free(gDataDirectory);
1293     }
1294     gDataDirectory = nullptr;
1295     gDataDirInitOnce.reset();
1296 
1297     delete gTimeZoneFilesDirectory;
1298     gTimeZoneFilesDirectory = nullptr;
1299     gTimeZoneFilesInitOnce.reset();
1300 
1301 #ifdef SEARCH_TZFILE
1302     delete gSearchTZFileResult;
1303     gSearchTZFileResult = nullptr;
1304 #endif
1305 
1306 #if U_POSIX_LOCALE || U_PLATFORM_USES_ONLY_WIN32_API
1307     if (gCorrectedPOSIXLocale && gCorrectedPOSIXLocaleHeapAllocated) {
1308         uprv_free(const_cast<char *>(gCorrectedPOSIXLocale));
1309         gCorrectedPOSIXLocale = nullptr;
1310         gCorrectedPOSIXLocaleHeapAllocated = false;
1311     }
1312 #endif
1313     return true;
1314 }
1315 
1316 /*
1317  * Set the data directory.
1318  *    Make a copy of the passed string, and set the global data dir to point to it.
1319  */
1320 U_CAPI void U_EXPORT2
u_setDataDirectory(const char * directory)1321 u_setDataDirectory(const char *directory) {
1322     char *newDataDir;
1323     int32_t length;
1324 
1325     if(directory==nullptr || *directory==0) {
1326         /* A small optimization to prevent the malloc and copy when the
1327         shared library is used, and this is a way to make sure that nullptr
1328         is never returned.
1329         */
1330         newDataDir = (char *)"";
1331     }
1332     else {
1333         length=(int32_t)uprv_strlen(directory);
1334         newDataDir = (char *)uprv_malloc(length + 2);
1335         /* Exit out if newDataDir could not be created. */
1336         if (newDataDir == nullptr) {
1337             return;
1338         }
1339         uprv_strcpy(newDataDir, directory);
1340 
1341 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR)
1342         {
1343             char *p;
1344             while((p = uprv_strchr(newDataDir, U_FILE_ALT_SEP_CHAR)) != nullptr) {
1345                 *p = U_FILE_SEP_CHAR;
1346             }
1347         }
1348 #endif
1349     }
1350 
1351     if (gDataDirectory && *gDataDirectory) {
1352         uprv_free(gDataDirectory);
1353     }
1354     gDataDirectory = newDataDir;
1355     ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
1356 }
1357 
1358 U_CAPI UBool U_EXPORT2
uprv_pathIsAbsolute(const char * path)1359 uprv_pathIsAbsolute(const char *path)
1360 {
1361   if(!path || !*path) {
1362     return false;
1363   }
1364 
1365   if(*path == U_FILE_SEP_CHAR) {
1366     return true;
1367   }
1368 
1369 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR)
1370   if(*path == U_FILE_ALT_SEP_CHAR) {
1371     return true;
1372   }
1373 #endif
1374 
1375 #if U_PLATFORM_USES_ONLY_WIN32_API
1376   if( (((path[0] >= 'A') && (path[0] <= 'Z')) ||
1377        ((path[0] >= 'a') && (path[0] <= 'z'))) &&
1378       path[1] == ':' ) {
1379     return true;
1380   }
1381 #endif
1382 
1383   return false;
1384 }
1385 
1386 /* Backup setting of ICU_DATA_DIR_PREFIX_ENV_VAR
1387    (needed for some Darwin ICU build environments) */
1388 #if U_PLATFORM_IS_DARWIN_BASED && defined(TARGET_OS_SIMULATOR) && TARGET_OS_SIMULATOR
1389 # if !defined(ICU_DATA_DIR_PREFIX_ENV_VAR)
1390 #  define ICU_DATA_DIR_PREFIX_ENV_VAR "IPHONE_SIMULATOR_ROOT"
1391 # endif
1392 #endif
1393 
1394 #if defined(ICU_DATA_DIR_WINDOWS)
1395 // Helper function to get the ICU Data Directory under the Windows directory location.
getIcuDataDirectoryUnderWindowsDirectory(char * directoryBuffer,UINT bufferLength)1396 static BOOL U_CALLCONV getIcuDataDirectoryUnderWindowsDirectory(char* directoryBuffer, UINT bufferLength)
1397 {
1398     wchar_t windowsPath[MAX_PATH];
1399     char windowsPathUtf8[MAX_PATH];
1400 
1401     UINT length = GetSystemWindowsDirectoryW(windowsPath, UPRV_LENGTHOF(windowsPath));
1402     if ((length > 0) && (length < (UPRV_LENGTHOF(windowsPath) - 1))) {
1403         // Convert UTF-16 to a UTF-8 string.
1404         UErrorCode status = U_ZERO_ERROR;
1405         int32_t windowsPathUtf8Len = 0;
1406         u_strToUTF8(windowsPathUtf8, static_cast<int32_t>(UPRV_LENGTHOF(windowsPathUtf8)),
1407             &windowsPathUtf8Len, reinterpret_cast<const char16_t*>(windowsPath), -1, &status);
1408 
1409         if (U_SUCCESS(status) && (status != U_STRING_NOT_TERMINATED_WARNING) &&
1410             (windowsPathUtf8Len < (UPRV_LENGTHOF(windowsPathUtf8) - 1))) {
1411             // Ensure it always has a separator, so we can append the ICU data path.
1412             if (windowsPathUtf8[windowsPathUtf8Len - 1] != U_FILE_SEP_CHAR) {
1413                 windowsPathUtf8[windowsPathUtf8Len++] = U_FILE_SEP_CHAR;
1414                 windowsPathUtf8[windowsPathUtf8Len] = '\0';
1415             }
1416             // Check if the concatenated string will fit.
1417             if ((windowsPathUtf8Len + UPRV_LENGTHOF(ICU_DATA_DIR_WINDOWS)) < bufferLength) {
1418                 uprv_strcpy(directoryBuffer, windowsPathUtf8);
1419                 uprv_strcat(directoryBuffer, ICU_DATA_DIR_WINDOWS);
1420                 return true;
1421             }
1422         }
1423     }
1424 
1425     return false;
1426 }
1427 #endif
1428 
dataDirectoryInitFn()1429 static void U_CALLCONV dataDirectoryInitFn() {
1430     /* If we already have the directory, then return immediately. Will happen if user called
1431      * u_setDataDirectory().
1432      */
1433     if (gDataDirectory) {
1434         return;
1435     }
1436 
1437     const char *path = nullptr;
1438 #if defined(ICU_DATA_DIR_PREFIX_ENV_VAR)
1439     char datadir_path_buffer[PATH_MAX];
1440 #endif
1441 
1442     /*
1443     When ICU_NO_USER_DATA_OVERRIDE is defined, users aren't allowed to
1444     override ICU's data with the ICU_DATA environment variable. This prevents
1445     problems where multiple custom copies of ICU's specific version of data
1446     are installed on a system. Either the application must define the data
1447     directory with u_setDataDirectory, define ICU_DATA_DIR when compiling
1448     ICU, set the data with udata_setCommonData or trust that all of the
1449     required data is contained in ICU's data library that contains
1450     the entry point defined by U_ICUDATA_ENTRY_POINT.
1451 
1452     There may also be some platforms where environment variables
1453     are not allowed.
1454     */
1455 #   if !defined(ICU_NO_USER_DATA_OVERRIDE) && !UCONFIG_NO_FILE_IO
1456     /* First try to get the environment variable */
1457 #     if U_PLATFORM_HAS_WINUWP_API == 0  // Windows UWP does not support getenv
1458         path=getenv("ICU_DATA");
1459 #     endif
1460 #   endif
1461 
1462     /* ICU_DATA_DIR may be set as a compile option.
1463      * U_ICU_DATA_DEFAULT_DIR is provided and is set by ICU at compile time
1464      * and is used only when data is built in archive mode eliminating the need
1465      * for ICU_DATA_DIR to be set. U_ICU_DATA_DEFAULT_DIR is set to the installation
1466      * directory of the data dat file. Users should use ICU_DATA_DIR if they want to
1467      * set their own path.
1468      */
1469 #if defined(ICU_DATA_DIR) || defined(U_ICU_DATA_DEFAULT_DIR)
1470     if(path==nullptr || *path==0) {
1471 # if defined(ICU_DATA_DIR_PREFIX_ENV_VAR)
1472         const char *prefix = getenv(ICU_DATA_DIR_PREFIX_ENV_VAR);
1473 # endif
1474 # ifdef ICU_DATA_DIR
1475         path=ICU_DATA_DIR;
1476 # else
1477         path=U_ICU_DATA_DEFAULT_DIR;
1478 # endif
1479 # if defined(ICU_DATA_DIR_PREFIX_ENV_VAR)
1480         if (prefix != nullptr) {
1481             snprintf(datadir_path_buffer, sizeof(datadir_path_buffer), "%s%s", prefix, path);
1482             path=datadir_path_buffer;
1483         }
1484 # endif
1485     }
1486 #endif
1487 
1488 #if defined(ICU_DATA_DIR_WINDOWS)
1489     char datadir_path_buffer[MAX_PATH];
1490     if (getIcuDataDirectoryUnderWindowsDirectory(datadir_path_buffer, UPRV_LENGTHOF(datadir_path_buffer))) {
1491         path = datadir_path_buffer;
1492     }
1493 #endif
1494 
1495     if(path==nullptr) {
1496         /* It looks really bad, set it to something. */
1497         path = "";
1498     }
1499 
1500     u_setDataDirectory(path);
1501 }
1502 
1503 U_CAPI const char * U_EXPORT2
u_getDataDirectory()1504 u_getDataDirectory() {
1505     umtx_initOnce(gDataDirInitOnce, &dataDirectoryInitFn);
1506     return gDataDirectory;
1507 }
1508 
setTimeZoneFilesDir(const char * path,UErrorCode & status)1509 static void setTimeZoneFilesDir(const char *path, UErrorCode &status) {
1510     if (U_FAILURE(status)) {
1511         return;
1512     }
1513     gTimeZoneFilesDirectory->clear();
1514     gTimeZoneFilesDirectory->append(path, status);
1515 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR)
1516     char *p = gTimeZoneFilesDirectory->data();
1517     while ((p = uprv_strchr(p, U_FILE_ALT_SEP_CHAR)) != nullptr) {
1518         *p = U_FILE_SEP_CHAR;
1519     }
1520 #endif
1521 }
1522 
1523 #define TO_STRING(x) TO_STRING_2(x)
1524 #define TO_STRING_2(x) #x
1525 
TimeZoneDataDirInitFn(UErrorCode & status)1526 static void U_CALLCONV TimeZoneDataDirInitFn(UErrorCode &status) {
1527     U_ASSERT(gTimeZoneFilesDirectory == nullptr);
1528     ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
1529     gTimeZoneFilesDirectory = new CharString();
1530     if (gTimeZoneFilesDirectory == nullptr) {
1531         status = U_MEMORY_ALLOCATION_ERROR;
1532         return;
1533     }
1534 
1535     const char *dir = "";
1536 
1537 #if defined(ICU_TIMEZONE_FILES_DIR_PREFIX_ENV_VAR)
1538     char timezonefilesdir_path_buffer[PATH_MAX];
1539     const char *prefix = getenv(ICU_TIMEZONE_FILES_DIR_PREFIX_ENV_VAR);
1540 #endif
1541 
1542 #if U_PLATFORM_HAS_WINUWP_API == 1
1543 // The UWP version does not support the environment variable setting.
1544 
1545 # if defined(ICU_DATA_DIR_WINDOWS)
1546     // When using the Windows system data, we can possibly pick up time zone data from the Windows directory.
1547     char datadir_path_buffer[MAX_PATH];
1548     if (getIcuDataDirectoryUnderWindowsDirectory(datadir_path_buffer, UPRV_LENGTHOF(datadir_path_buffer))) {
1549         dir = datadir_path_buffer;
1550     }
1551 # endif
1552 
1553 #else
1554     dir = getenv("ICU_TIMEZONE_FILES_DIR");
1555 #endif // U_PLATFORM_HAS_WINUWP_API
1556 
1557 #if defined(U_TIMEZONE_FILES_DIR)
1558     if (dir == nullptr) {
1559         // Build time configuration setting.
1560         dir = TO_STRING(U_TIMEZONE_FILES_DIR);
1561     }
1562 #endif
1563 
1564     if (dir == nullptr) {
1565         dir = "";
1566     }
1567 
1568 #if defined(ICU_TIMEZONE_FILES_DIR_PREFIX_ENV_VAR)
1569     if (prefix != nullptr) {
1570         snprintf(timezonefilesdir_path_buffer, sizeof(timezonefilesdir_path_buffer), "%s%s", prefix, dir);
1571         dir = timezonefilesdir_path_buffer;
1572     }
1573 #endif
1574 
1575     setTimeZoneFilesDir(dir, status);
1576 }
1577 
1578 
1579 U_CAPI const char * U_EXPORT2
u_getTimeZoneFilesDirectory(UErrorCode * status)1580 u_getTimeZoneFilesDirectory(UErrorCode *status) {
1581     umtx_initOnce(gTimeZoneFilesInitOnce, &TimeZoneDataDirInitFn, *status);
1582     return U_SUCCESS(*status) ? gTimeZoneFilesDirectory->data() : "";
1583 }
1584 
1585 U_CAPI void U_EXPORT2
u_setTimeZoneFilesDirectory(const char * path,UErrorCode * status)1586 u_setTimeZoneFilesDirectory(const char *path, UErrorCode *status) {
1587     umtx_initOnce(gTimeZoneFilesInitOnce, &TimeZoneDataDirInitFn, *status);
1588     setTimeZoneFilesDir(path, *status);
1589 
1590     // Note: this function does some extra churn, first setting based on the
1591     //       environment, then immediately replacing with the value passed in.
1592     //       The logic is simpler that way, and performance shouldn't be an issue.
1593 }
1594 
1595 
1596 #if U_POSIX_LOCALE
1597 /* A helper function used by uprv_getPOSIXIDForDefaultLocale and
1598  * uprv_getPOSIXIDForDefaultCodepage. Returns the posix locale id for
1599  * LC_CTYPE and LC_MESSAGES. It doesn't support other locale categories.
1600  */
uprv_getPOSIXIDForCategory(int category)1601 static const char *uprv_getPOSIXIDForCategory(int category)
1602 {
1603     const char* posixID = nullptr;
1604     if (category == LC_MESSAGES || category == LC_CTYPE) {
1605         /*
1606         * On Solaris two different calls to setlocale can result in
1607         * different values. Only get this value once.
1608         *
1609         * We must check this first because an application can set this.
1610         *
1611         * LC_ALL can't be used because it's platform dependent. The LANG
1612         * environment variable seems to affect LC_CTYPE variable by default.
1613         * Here is what setlocale(LC_ALL, nullptr) can return.
1614         * HPUX can return 'C C C C C C C'
1615         * Solaris can return /en_US/C/C/C/C/C on the second try.
1616         * Linux can return LC_CTYPE=C;LC_NUMERIC=C;...
1617         *
1618         * The default codepage detection also needs to use LC_CTYPE.
1619         *
1620         * Do not call setlocale(LC_*, "")! Using an empty string instead
1621         * of nullptr, will modify the libc behavior.
1622         */
1623         posixID = setlocale(category, nullptr);
1624         if ((posixID == nullptr)
1625             || (uprv_strcmp("C", posixID) == 0)
1626             || (uprv_strcmp("POSIX", posixID) == 0))
1627         {
1628             /* Maybe we got some garbage.  Try something more reasonable */
1629             posixID = getenv("LC_ALL");
1630             /* Solaris speaks POSIX -  See IEEE Std 1003.1-2008
1631              * This is needed to properly handle empty env. variables
1632              */
1633 #if U_PLATFORM == U_PF_SOLARIS
1634             if ((posixID == 0) || (posixID[0] == '\0')) {
1635                 posixID = getenv(category == LC_MESSAGES ? "LC_MESSAGES" : "LC_CTYPE");
1636                 if ((posixID == 0) || (posixID[0] == '\0')) {
1637 #else
1638             if (posixID == nullptr) {
1639                 posixID = getenv(category == LC_MESSAGES ? "LC_MESSAGES" : "LC_CTYPE");
1640                 if (posixID == nullptr) {
1641 #endif
1642                     posixID = getenv("LANG");
1643                 }
1644             }
1645         }
1646     }
1647     if ((posixID == nullptr)
1648         || (uprv_strcmp("C", posixID) == 0)
1649         || (uprv_strcmp("POSIX", posixID) == 0))
1650     {
1651         /* Nothing worked.  Give it a nice POSIX default value. */
1652         posixID = "en_US_POSIX";
1653         // Note: this test will not catch 'C.UTF-8',
1654         // that will be handled in uprv_getDefaultLocaleID().
1655         // Leave this mapping here for the uprv_getPOSIXIDForDefaultCodepage()
1656         // caller which expects to see "en_US_POSIX" in many branches.
1657     }
1658     return posixID;
1659 }
1660 
1661 /* Return just the POSIX id for the default locale, whatever happens to be in
1662  * it. It gets the value from LC_MESSAGES and indirectly from LC_ALL and LANG.
1663  */
1664 static const char *uprv_getPOSIXIDForDefaultLocale()
1665 {
1666     static const char* posixID = nullptr;
1667     if (posixID == nullptr) {
1668         posixID = uprv_getPOSIXIDForCategory(LC_MESSAGES);
1669     }
1670     return posixID;
1671 }
1672 
1673 #if !U_CHARSET_IS_UTF8
1674 /* Return just the POSIX id for the default codepage, whatever happens to be in
1675  * it. It gets the value from LC_CTYPE and indirectly from LC_ALL and LANG.
1676  */
1677 static const char *uprv_getPOSIXIDForDefaultCodepage()
1678 {
1679     static const char* posixID = nullptr;
1680     if (posixID == 0) {
1681         posixID = uprv_getPOSIXIDForCategory(LC_CTYPE);
1682     }
1683     return posixID;
1684 }
1685 #endif
1686 #endif
1687 
1688 /* NOTE: The caller should handle thread safety */
1689 U_CAPI const char* U_EXPORT2
1690 uprv_getDefaultLocaleID()
1691 {
1692 #if U_POSIX_LOCALE
1693 /*
1694   Note that:  (a '!' means the ID is improper somehow)
1695      LC_ALL  ---->     default_loc          codepage
1696 --------------------------------------------------------
1697      ab.CD             ab                   CD
1698      ab@CD             ab__CD               -
1699      ab@CD.EF          ab__CD               EF
1700 
1701      ab_CD.EF@GH       ab_CD_GH             EF
1702 
1703 Some 'improper' ways to do the same as above:
1704   !  ab_CD@GH.EF       ab_CD_GH             EF
1705   !  ab_CD.EF@GH.IJ    ab_CD_GH             EF
1706   !  ab_CD@ZZ.EF@GH.IJ ab_CD_GH             EF
1707 
1708      _CD@GH            _CD_GH               -
1709      _CD.EF@GH         _CD_GH               EF
1710 
1711 The variant cannot have dots in it.
1712 The 'rightmost' variant (@xxx) wins.
1713 The leftmost codepage (.xxx) wins.
1714 */
1715     const char* posixID = uprv_getPOSIXIDForDefaultLocale();
1716 
1717     /* Format: (no spaces)
1718     ll [ _CC ] [ . MM ] [ @ VV]
1719 
1720       l = lang, C = ctry, M = charmap, V = variant
1721     */
1722 
1723     if (gCorrectedPOSIXLocale != nullptr) {
1724         return gCorrectedPOSIXLocale;
1725     }
1726 
1727     // Copy the ID into owned memory.
1728     // Over-allocate in case we replace "C" with "en_US_POSIX" (+10), + null termination
1729     char *correctedPOSIXLocale = static_cast<char *>(uprv_malloc(uprv_strlen(posixID) + 10 + 1));
1730     if (correctedPOSIXLocale == nullptr) {
1731         return nullptr;
1732     }
1733     uprv_strcpy(correctedPOSIXLocale, posixID);
1734 
1735     char *limit;
1736     if ((limit = uprv_strchr(correctedPOSIXLocale, '.')) != nullptr) {
1737         *limit = 0;
1738     }
1739     if ((limit = uprv_strchr(correctedPOSIXLocale, '@')) != nullptr) {
1740         *limit = 0;
1741     }
1742 
1743     if ((uprv_strcmp("C", correctedPOSIXLocale) == 0) // no @ variant
1744         || (uprv_strcmp("POSIX", correctedPOSIXLocale) == 0)) {
1745       // Raw input was C.* or POSIX.*, Give it a nice POSIX default value.
1746       // (The "C"/"POSIX" case is handled in uprv_getPOSIXIDForCategory())
1747       uprv_strcpy(correctedPOSIXLocale, "en_US_POSIX");
1748     }
1749 
1750     /* Note that we scan the *uncorrected* ID. */
1751     const char *p;
1752     if ((p = uprv_strrchr(posixID, '@')) != nullptr) {
1753         p++;
1754 
1755         /* Take care of any special cases here.. */
1756         if (!uprv_strcmp(p, "nynorsk")) {
1757             p = "NY";
1758             /* Don't worry about no__NY. In practice, it won't appear. */
1759         }
1760 
1761         if (uprv_strchr(correctedPOSIXLocale,'_') == nullptr) {
1762             uprv_strcat(correctedPOSIXLocale, "__"); /* aa@b -> aa__b (note this can make the new locale 1 char longer) */
1763         }
1764         else {
1765             uprv_strcat(correctedPOSIXLocale, "_"); /* aa_CC@b -> aa_CC_b */
1766         }
1767 
1768         const char *q;
1769         if ((q = uprv_strchr(p, '.')) != nullptr) {
1770             /* How big will the resulting string be? */
1771             int32_t len = (int32_t)(uprv_strlen(correctedPOSIXLocale) + (q-p));
1772             uprv_strncat(correctedPOSIXLocale, p, q-p); // do not include charset
1773             correctedPOSIXLocale[len] = 0;
1774         }
1775         else {
1776             /* Anything following the @ sign */
1777             uprv_strcat(correctedPOSIXLocale, p);
1778         }
1779 
1780         /* Should there be a map from 'no@nynorsk' -> no_NO_NY here?
1781          * How about 'russian' -> 'ru'?
1782          * Many of the other locales using ISO codes will be handled by the
1783          * canonicalization functions in uloc_getDefault.
1784          */
1785     }
1786 
1787     if (gCorrectedPOSIXLocale == nullptr) {
1788         gCorrectedPOSIXLocale = correctedPOSIXLocale;
1789         gCorrectedPOSIXLocaleHeapAllocated = true;
1790         ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
1791         correctedPOSIXLocale = nullptr;
1792     }
1793     posixID = gCorrectedPOSIXLocale;
1794 
1795     if (correctedPOSIXLocale != nullptr) {  /* Was already set - clean up. */
1796         uprv_free(correctedPOSIXLocale);
1797     }
1798 
1799     return posixID;
1800 
1801 #elif U_PLATFORM_USES_ONLY_WIN32_API
1802 #define POSIX_LOCALE_CAPACITY 64
1803     UErrorCode status = U_ZERO_ERROR;
1804     char *correctedPOSIXLocale = nullptr;
1805 
1806     // If we have already figured this out just use the cached value
1807     if (gCorrectedPOSIXLocale != nullptr) {
1808         return gCorrectedPOSIXLocale;
1809     }
1810 
1811     // No cached value, need to determine the current value
1812     static WCHAR windowsLocale[LOCALE_NAME_MAX_LENGTH] = {};
1813     int length = GetLocaleInfoEx(LOCALE_NAME_USER_DEFAULT, LOCALE_SNAME, windowsLocale, LOCALE_NAME_MAX_LENGTH);
1814 
1815     // Now we should have a Windows locale name that needs converted to the POSIX style.
1816     if (length > 0) // If length is 0, then the GetLocaleInfoEx failed.
1817     {
1818         // First we need to go from UTF-16 to char (and also convert from _ to - while we're at it.)
1819         char modifiedWindowsLocale[LOCALE_NAME_MAX_LENGTH] = {};
1820 
1821         int32_t i;
1822         for (i = 0; i < UPRV_LENGTHOF(modifiedWindowsLocale); i++)
1823         {
1824             if (windowsLocale[i] == '_')
1825             {
1826                 modifiedWindowsLocale[i] = '-';
1827             }
1828             else
1829             {
1830                 modifiedWindowsLocale[i] = static_cast<char>(windowsLocale[i]);
1831             }
1832 
1833             if (modifiedWindowsLocale[i] == '\0')
1834             {
1835                 break;
1836             }
1837         }
1838 
1839         if (i >= UPRV_LENGTHOF(modifiedWindowsLocale))
1840         {
1841             // Ran out of room, can't really happen, maybe we'll be lucky about a matching
1842             // locale when tags are dropped
1843             modifiedWindowsLocale[UPRV_LENGTHOF(modifiedWindowsLocale) - 1] = '\0';
1844         }
1845 
1846         // Now normalize the resulting name
1847         correctedPOSIXLocale = static_cast<char *>(uprv_malloc(POSIX_LOCALE_CAPACITY + 1));
1848         /* TODO: Should we just exit on memory allocation failure? */
1849         if (correctedPOSIXLocale)
1850         {
1851             int32_t posixLen = uloc_canonicalize(modifiedWindowsLocale, correctedPOSIXLocale, POSIX_LOCALE_CAPACITY, &status);
1852             if (U_SUCCESS(status))
1853             {
1854                 *(correctedPOSIXLocale + posixLen) = 0;
1855                 gCorrectedPOSIXLocale = correctedPOSIXLocale;
1856                 gCorrectedPOSIXLocaleHeapAllocated = true;
1857                 ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
1858             }
1859             else
1860             {
1861                 uprv_free(correctedPOSIXLocale);
1862             }
1863         }
1864     }
1865 
1866     // If unable to find a locale we can agree upon, use en-US by default
1867     if (gCorrectedPOSIXLocale == nullptr) {
1868         gCorrectedPOSIXLocale = "en_US";
1869     }
1870     return gCorrectedPOSIXLocale;
1871 
1872 #elif U_PLATFORM == U_PF_OS400
1873     /* locales are process scoped and are by definition thread safe */
1874     static char correctedLocale[64];
1875     const  char *localeID = getenv("LC_ALL");
1876            char *p;
1877 
1878     if (localeID == nullptr)
1879         localeID = getenv("LANG");
1880     if (localeID == nullptr)
1881         localeID = setlocale(LC_ALL, nullptr);
1882     /* Make sure we have something... */
1883     if (localeID == nullptr)
1884         return "en_US_POSIX";
1885 
1886     /* Extract the locale name from the path. */
1887     if((p = uprv_strrchr(localeID, '/')) != nullptr)
1888     {
1889         /* Increment p to start of locale name. */
1890         p++;
1891         localeID = p;
1892     }
1893 
1894     /* Copy to work location. */
1895     uprv_strcpy(correctedLocale, localeID);
1896 
1897     /* Strip off the '.locale' extension. */
1898     if((p = uprv_strchr(correctedLocale, '.')) != nullptr) {
1899         *p = 0;
1900     }
1901 
1902     /* Upper case the locale name. */
1903     T_CString_toUpperCase(correctedLocale);
1904 
1905     /* See if we are using the POSIX locale.  Any of the
1906     * following are equivalent and use the same QLGPGCMA
1907     * (POSIX) locale.
1908     * QLGPGCMA2 means UCS2
1909     * QLGPGCMA_4 means UTF-32
1910     * QLGPGCMA_8 means UTF-8
1911     */
1912     if ((uprv_strcmp("C", correctedLocale) == 0) ||
1913         (uprv_strcmp("POSIX", correctedLocale) == 0) ||
1914         (uprv_strncmp("QLGPGCMA", correctedLocale, 8) == 0))
1915     {
1916         uprv_strcpy(correctedLocale, "en_US_POSIX");
1917     }
1918     else
1919     {
1920         int16_t LocaleLen;
1921 
1922         /* Lower case the lang portion. */
1923         for(p = correctedLocale; *p != 0 && *p != '_'; p++)
1924         {
1925             *p = uprv_tolower(*p);
1926         }
1927 
1928         /* Adjust for Euro.  After '_E' add 'URO'. */
1929         LocaleLen = uprv_strlen(correctedLocale);
1930         if (correctedLocale[LocaleLen - 2] == '_' &&
1931             correctedLocale[LocaleLen - 1] == 'E')
1932         {
1933             uprv_strcat(correctedLocale, "URO");
1934         }
1935 
1936         /* If using Lotus-based locale then convert to
1937          * equivalent non Lotus.
1938          */
1939         else if (correctedLocale[LocaleLen - 2] == '_' &&
1940             correctedLocale[LocaleLen - 1] == 'L')
1941         {
1942             correctedLocale[LocaleLen - 2] = 0;
1943         }
1944 
1945         /* There are separate simplified and traditional
1946          * locales called zh_HK_S and zh_HK_T.
1947          */
1948         else if (uprv_strncmp(correctedLocale, "zh_HK", 5) == 0)
1949         {
1950             uprv_strcpy(correctedLocale, "zh_HK");
1951         }
1952 
1953         /* A special zh_CN_GBK locale...
1954         */
1955         else if (uprv_strcmp(correctedLocale, "zh_CN_GBK") == 0)
1956         {
1957             uprv_strcpy(correctedLocale, "zh_CN");
1958         }
1959 
1960     }
1961 
1962     return correctedLocale;
1963 #endif
1964 
1965 }
1966 
1967 #if !U_CHARSET_IS_UTF8
1968 #if U_POSIX_LOCALE
1969 /*
1970 Due to various platform differences, one platform may specify a charset,
1971 when they really mean a different charset. Remap the names so that they are
1972 compatible with ICU. Only conflicting/ambiguous aliases should be resolved
1973 here. Before adding anything to this function, please consider adding unique
1974 names to the ICU alias table in the data directory.
1975 */
1976 static const char*
1977 remapPlatformDependentCodepage(const char *locale, const char *name) {
1978     if (locale != nullptr && *locale == 0) {
1979         /* Make sure that an empty locale is handled the same way. */
1980         locale = nullptr;
1981     }
1982     if (name == nullptr) {
1983         return nullptr;
1984     }
1985 #if U_PLATFORM == U_PF_AIX
1986     if (uprv_strcmp(name, "IBM-943") == 0) {
1987         /* Use the ASCII compatible ibm-943 */
1988         name = "Shift-JIS";
1989     }
1990     else if (uprv_strcmp(name, "IBM-1252") == 0) {
1991         /* Use the windows-1252 that contains the Euro */
1992         name = "IBM-5348";
1993     }
1994 #elif U_PLATFORM == U_PF_SOLARIS
1995     if (locale != nullptr && uprv_strcmp(name, "EUC") == 0) {
1996         /* Solaris underspecifies the "EUC" name. */
1997         if (uprv_strcmp(locale, "zh_CN") == 0) {
1998             name = "EUC-CN";
1999         }
2000         else if (uprv_strcmp(locale, "zh_TW") == 0) {
2001             name = "EUC-TW";
2002         }
2003         else if (uprv_strcmp(locale, "ko_KR") == 0) {
2004             name = "EUC-KR";
2005         }
2006     }
2007     else if (uprv_strcmp(name, "eucJP") == 0) {
2008         /*
2009         ibm-954 is the best match.
2010         ibm-33722 is the default for eucJP (similar to Windows).
2011         */
2012         name = "eucjis";
2013     }
2014     else if (uprv_strcmp(name, "646") == 0) {
2015         /*
2016          * The default codepage given by Solaris is 646 but the C library routines treat it as if it was
2017          * ISO-8859-1 instead of US-ASCII(646).
2018          */
2019         name = "ISO-8859-1";
2020     }
2021 #elif U_PLATFORM_IS_DARWIN_BASED
2022     if (locale == nullptr && *name == 0) {
2023         /*
2024         No locale was specified, and an empty name was passed in.
2025         This usually indicates that nl_langinfo didn't return valid information.
2026         Mac OS X uses UTF-8 by default (especially the locale data and console).
2027         */
2028         name = "UTF-8";
2029     }
2030     else if (uprv_strcmp(name, "CP949") == 0) {
2031         /* Remap CP949 to a similar codepage to avoid issues with backslash and won symbol. */
2032         name = "EUC-KR";
2033     }
2034     else if (locale != nullptr && uprv_strcmp(locale, "en_US_POSIX") != 0 && uprv_strcmp(name, "US-ASCII") == 0) {
2035         /*
2036          * For non C/POSIX locale, default the code page to UTF-8 instead of US-ASCII.
2037          */
2038         name = "UTF-8";
2039     }
2040 #elif U_PLATFORM == U_PF_BSD
2041     if (uprv_strcmp(name, "CP949") == 0) {
2042         /* Remap CP949 to a similar codepage to avoid issues with backslash and won symbol. */
2043         name = "EUC-KR";
2044     }
2045 #elif U_PLATFORM == U_PF_HPUX
2046     if (locale != nullptr && uprv_strcmp(locale, "zh_HK") == 0 && uprv_strcmp(name, "big5") == 0) {
2047         /* HP decided to extend big5 as hkbig5 even though it's not compatible :-( */
2048         /* zh_TW.big5 is not the same charset as zh_HK.big5! */
2049         name = "hkbig5";
2050     }
2051     else if (uprv_strcmp(name, "eucJP") == 0) {
2052         /*
2053         ibm-1350 is the best match, but unavailable.
2054         ibm-954 is mostly a superset of ibm-1350.
2055         ibm-33722 is the default for eucJP (similar to Windows).
2056         */
2057         name = "eucjis";
2058     }
2059 #elif U_PLATFORM == U_PF_LINUX
2060     if (locale != nullptr && uprv_strcmp(name, "euc") == 0) {
2061         /* Linux underspecifies the "EUC" name. */
2062         if (uprv_strcmp(locale, "korean") == 0) {
2063             name = "EUC-KR";
2064         }
2065         else if (uprv_strcmp(locale, "japanese") == 0) {
2066             /* See comment below about eucJP */
2067             name = "eucjis";
2068         }
2069     }
2070     else if (uprv_strcmp(name, "eucjp") == 0) {
2071         /*
2072         ibm-1350 is the best match, but unavailable.
2073         ibm-954 is mostly a superset of ibm-1350.
2074         ibm-33722 is the default for eucJP (similar to Windows).
2075         */
2076         name = "eucjis";
2077     }
2078     else if (locale != nullptr && uprv_strcmp(locale, "en_US_POSIX") != 0 &&
2079             (uprv_strcmp(name, "ANSI_X3.4-1968") == 0 || uprv_strcmp(name, "US-ASCII") == 0)) {
2080         /*
2081          * For non C/POSIX locale, default the code page to UTF-8 instead of US-ASCII.
2082          */
2083         name = "UTF-8";
2084     }
2085     /*
2086      * Linux returns ANSI_X3.4-1968 for C/POSIX, but the call site takes care of
2087      * it by falling back to 'US-ASCII' when nullptr is returned from this
2088      * function. So, we don't have to worry about it here.
2089      */
2090 #endif
2091     /* return nullptr when "" is passed in */
2092     if (*name == 0) {
2093         name = nullptr;
2094     }
2095     return name;
2096 }
2097 
2098 static const char*
2099 getCodepageFromPOSIXID(const char *localeName, char * buffer, int32_t buffCapacity)
2100 {
2101     char localeBuf[100];
2102     const char *name = nullptr;
2103     char *variant = nullptr;
2104 
2105     if (localeName != nullptr && (name = (uprv_strchr(localeName, '.'))) != nullptr) {
2106         size_t localeCapacity = uprv_min(sizeof(localeBuf), (name-localeName)+1);
2107         uprv_strncpy(localeBuf, localeName, localeCapacity);
2108         localeBuf[localeCapacity-1] = 0; /* ensure NUL termination */
2109         name = uprv_strncpy(buffer, name+1, buffCapacity);
2110         buffer[buffCapacity-1] = 0; /* ensure NUL termination */
2111         if ((variant = const_cast<char *>(uprv_strchr(name, '@'))) != nullptr) {
2112             *variant = 0;
2113         }
2114         name = remapPlatformDependentCodepage(localeBuf, name);
2115     }
2116     return name;
2117 }
2118 #endif
2119 
2120 static const char*
2121 int_getDefaultCodepage()
2122 {
2123 #if U_PLATFORM == U_PF_OS400
2124     uint32_t ccsid = 37; /* Default to ibm-37 */
2125     static char codepage[64];
2126     Qwc_JOBI0400_t jobinfo;
2127     Qus_EC_t error = { sizeof(Qus_EC_t) }; /* SPI error code */
2128 
2129     EPT_CALL(QUSRJOBI)(&jobinfo, sizeof(jobinfo), "JOBI0400",
2130         "*                         ", "                ", &error);
2131 
2132     if (error.Bytes_Available == 0) {
2133         if (jobinfo.Coded_Char_Set_ID != 0xFFFF) {
2134             ccsid = (uint32_t)jobinfo.Coded_Char_Set_ID;
2135         }
2136         else if (jobinfo.Default_Coded_Char_Set_Id != 0xFFFF) {
2137             ccsid = (uint32_t)jobinfo.Default_Coded_Char_Set_Id;
2138         }
2139         /* else use the default */
2140     }
2141     snprintf(codepage, sizeof(codepage), "ibm-%d", ccsid);
2142     return codepage;
2143 
2144 #elif U_PLATFORM == U_PF_OS390
2145     static char codepage[64];
2146 
2147     strncpy(codepage, nl_langinfo(CODESET),63-strlen(UCNV_SWAP_LFNL_OPTION_STRING));
2148     strcat(codepage,UCNV_SWAP_LFNL_OPTION_STRING);
2149     codepage[63] = 0; /* NUL terminate */
2150 
2151     return codepage;
2152 
2153 #elif U_PLATFORM_USES_ONLY_WIN32_API
2154     static char codepage[64];
2155     DWORD codepageNumber = 0;
2156 
2157 #if U_PLATFORM_HAS_WINUWP_API == 1
2158     // UWP doesn't have a direct API to get the default ACP as Microsoft would rather
2159     // have folks use Unicode than a "system" code page, however this is the same
2160     // codepage as the system default locale codepage.  (FWIW, the system locale is
2161     // ONLY used for codepage, it should never be used for anything else)
2162     GetLocaleInfoEx(LOCALE_NAME_SYSTEM_DEFAULT, LOCALE_IDEFAULTANSICODEPAGE | LOCALE_RETURN_NUMBER,
2163         (LPWSTR)&codepageNumber, sizeof(codepageNumber) / sizeof(WCHAR));
2164 #else
2165     // Win32 apps can call GetACP
2166     codepageNumber = GetACP();
2167 #endif
2168     // Special case for UTF-8
2169     if (codepageNumber == 65001)
2170     {
2171         return "UTF-8";
2172     }
2173     // Windows codepages can look like windows-1252, so format the found number
2174     // the numbers are eclectic, however all valid system code pages, besides UTF-8
2175     // are between 3 and 19999
2176     if (codepageNumber > 0 && codepageNumber < 20000)
2177     {
2178         snprintf(codepage, sizeof(codepage), "windows-%ld", codepageNumber);
2179         return codepage;
2180     }
2181     // If the codepage number call failed then return UTF-8
2182     return "UTF-8";
2183 
2184 #elif U_POSIX_LOCALE
2185     static char codesetName[100];
2186     const char *localeName = nullptr;
2187     const char *name = nullptr;
2188 
2189     localeName = uprv_getPOSIXIDForDefaultCodepage();
2190     uprv_memset(codesetName, 0, sizeof(codesetName));
2191     /* On Solaris nl_langinfo returns C locale values unless setlocale
2192      * was called earlier.
2193      */
2194 #if (U_HAVE_NL_LANGINFO_CODESET && U_PLATFORM != U_PF_SOLARIS)
2195     /* When available, check nl_langinfo first because it usually gives more
2196        useful names. It depends on LC_CTYPE.
2197        nl_langinfo may use the same buffer as setlocale. */
2198     {
2199         const char *codeset = nl_langinfo(U_NL_LANGINFO_CODESET);
2200 #if U_PLATFORM_IS_DARWIN_BASED || U_PLATFORM_IS_LINUX_BASED
2201         /*
2202          * On Linux and MacOSX, ensure that default codepage for non C/POSIX locale is UTF-8
2203          * instead of ASCII.
2204          */
2205         if (uprv_strcmp(localeName, "en_US_POSIX") != 0) {
2206             codeset = remapPlatformDependentCodepage(localeName, codeset);
2207         } else
2208 #endif
2209         {
2210             codeset = remapPlatformDependentCodepage(nullptr, codeset);
2211         }
2212 
2213         if (codeset != nullptr) {
2214             uprv_strncpy(codesetName, codeset, sizeof(codesetName));
2215             codesetName[sizeof(codesetName)-1] = 0;
2216             return codesetName;
2217         }
2218     }
2219 #endif
2220 
2221     /* Use setlocale in a nice way, and then check some environment variables.
2222        Maybe the application used setlocale already.
2223     */
2224     uprv_memset(codesetName, 0, sizeof(codesetName));
2225     name = getCodepageFromPOSIXID(localeName, codesetName, sizeof(codesetName));
2226     if (name) {
2227         /* if we can find the codeset name from setlocale, return that. */
2228         return name;
2229     }
2230 
2231     if (*codesetName == 0)
2232     {
2233         /* Everything failed. Return US ASCII (ISO 646). */
2234         (void)uprv_strcpy(codesetName, "US-ASCII");
2235     }
2236     return codesetName;
2237 #else
2238     return "US-ASCII";
2239 #endif
2240 }
2241 
2242 
2243 U_CAPI const char*  U_EXPORT2
2244 uprv_getDefaultCodepage()
2245 {
2246     static char const  *name = nullptr;
2247     umtx_lock(nullptr);
2248     if (name == nullptr) {
2249         name = int_getDefaultCodepage();
2250     }
2251     umtx_unlock(nullptr);
2252     return name;
2253 }
2254 #endif  /* !U_CHARSET_IS_UTF8 */
2255 
2256 
2257 /* end of platform-specific implementation -------------- */
2258 
2259 /* version handling --------------------------------------------------------- */
2260 
2261 U_CAPI void U_EXPORT2
2262 u_versionFromString(UVersionInfo versionArray, const char *versionString) {
2263     char *end;
2264     uint16_t part=0;
2265 
2266     if(versionArray==nullptr) {
2267         return;
2268     }
2269 
2270     if(versionString!=nullptr) {
2271         for(;;) {
2272             versionArray[part]=(uint8_t)uprv_strtoul(versionString, &end, 10);
2273             if(end==versionString || ++part==U_MAX_VERSION_LENGTH || *end!=U_VERSION_DELIMITER) {
2274                 break;
2275             }
2276             versionString=end+1;
2277         }
2278     }
2279 
2280     while(part<U_MAX_VERSION_LENGTH) {
2281         versionArray[part++]=0;
2282     }
2283 }
2284 
2285 U_CAPI void U_EXPORT2
2286 u_versionFromUString(UVersionInfo versionArray, const char16_t *versionString) {
2287     if(versionArray!=nullptr && versionString!=nullptr) {
2288         char versionChars[U_MAX_VERSION_STRING_LENGTH+1];
2289         int32_t len = u_strlen(versionString);
2290         if(len>U_MAX_VERSION_STRING_LENGTH) {
2291             len = U_MAX_VERSION_STRING_LENGTH;
2292         }
2293         u_UCharsToChars(versionString, versionChars, len);
2294         versionChars[len]=0;
2295         u_versionFromString(versionArray, versionChars);
2296     }
2297 }
2298 
2299 U_CAPI void U_EXPORT2
2300 u_versionToString(const UVersionInfo versionArray, char *versionString) {
2301     uint16_t count, part;
2302     uint8_t field;
2303 
2304     if(versionString==nullptr) {
2305         return;
2306     }
2307 
2308     if(versionArray==nullptr) {
2309         versionString[0]=0;
2310         return;
2311     }
2312 
2313     /* count how many fields need to be written */
2314     for(count=4; count>0 && versionArray[count-1]==0; --count) {
2315     }
2316 
2317     if(count <= 1) {
2318         count = 2;
2319     }
2320 
2321     /* write the first part */
2322     /* write the decimal field value */
2323     field=versionArray[0];
2324     if(field>=100) {
2325         *versionString++=(char)('0'+field/100);
2326         field%=100;
2327     }
2328     if(field>=10) {
2329         *versionString++=(char)('0'+field/10);
2330         field%=10;
2331     }
2332     *versionString++=(char)('0'+field);
2333 
2334     /* write the following parts */
2335     for(part=1; part<count; ++part) {
2336         /* write a dot first */
2337         *versionString++=U_VERSION_DELIMITER;
2338 
2339         /* write the decimal field value */
2340         field=versionArray[part];
2341         if(field>=100) {
2342             *versionString++=(char)('0'+field/100);
2343             field%=100;
2344         }
2345         if(field>=10) {
2346             *versionString++=(char)('0'+field/10);
2347             field%=10;
2348         }
2349         *versionString++=(char)('0'+field);
2350     }
2351 
2352     /* NUL-terminate */
2353     *versionString=0;
2354 }
2355 
2356 U_CAPI void U_EXPORT2
2357 u_getVersion(UVersionInfo versionArray) {
2358     (void)copyright;   // Suppress unused variable warning from clang.
2359     u_versionFromString(versionArray, U_ICU_VERSION);
2360 }
2361 
2362 /**
2363  * icucfg.h dependent code
2364  */
2365 
2366 #if U_ENABLE_DYLOAD && HAVE_DLOPEN && !U_PLATFORM_USES_ONLY_WIN32_API
2367 
2368 #if HAVE_DLFCN_H
2369 #ifdef __MVS__
2370 #ifndef __SUSV3
2371 #define __SUSV3 1
2372 #endif
2373 #endif
2374 #include <dlfcn.h>
2375 #endif /* HAVE_DLFCN_H */
2376 
2377 U_CAPI void * U_EXPORT2
2378 uprv_dl_open(const char *libName, UErrorCode *status) {
2379   void *ret = nullptr;
2380   if(U_FAILURE(*status)) return ret;
2381   ret =  dlopen(libName, RTLD_NOW|RTLD_GLOBAL);
2382   if(ret==nullptr) {
2383 #ifdef U_TRACE_DYLOAD
2384     printf("dlerror on dlopen(%s): %s\n", libName, dlerror());
2385 #endif
2386     *status = U_MISSING_RESOURCE_ERROR;
2387   }
2388   return ret;
2389 }
2390 
2391 U_CAPI void U_EXPORT2
2392 uprv_dl_close(void *lib, UErrorCode *status) {
2393   if(U_FAILURE(*status)) return;
2394   dlclose(lib);
2395 }
2396 
2397 U_CAPI UVoidFunction* U_EXPORT2
2398 uprv_dlsym_func(void *lib, const char* sym, UErrorCode *status) {
2399   union {
2400       UVoidFunction *fp;
2401       void *vp;
2402   } uret;
2403   uret.fp = nullptr;
2404   if(U_FAILURE(*status)) return uret.fp;
2405   uret.vp = dlsym(lib, sym);
2406   if(uret.vp == nullptr) {
2407 #ifdef U_TRACE_DYLOAD
2408     printf("dlerror on dlsym(%p,%s): %s\n", lib,sym, dlerror());
2409 #endif
2410     *status = U_MISSING_RESOURCE_ERROR;
2411   }
2412   return uret.fp;
2413 }
2414 
2415 #elif U_ENABLE_DYLOAD && U_PLATFORM_USES_ONLY_WIN32_API && !U_PLATFORM_HAS_WINUWP_API
2416 
2417 /* Windows API implementation. */
2418 // Note: UWP does not expose/allow these APIs, so the UWP version gets the null implementation. */
2419 
2420 U_CAPI void * U_EXPORT2
2421 uprv_dl_open(const char *libName, UErrorCode *status) {
2422   HMODULE lib = nullptr;
2423 
2424   if(U_FAILURE(*status)) return nullptr;
2425 
2426   lib = LoadLibraryA(libName);
2427 
2428   if(lib==nullptr) {
2429     *status = U_MISSING_RESOURCE_ERROR;
2430   }
2431 
2432   return (void*)lib;
2433 }
2434 
2435 U_CAPI void U_EXPORT2
2436 uprv_dl_close(void *lib, UErrorCode *status) {
2437   HMODULE handle = (HMODULE)lib;
2438   if(U_FAILURE(*status)) return;
2439 
2440   FreeLibrary(handle);
2441 
2442   return;
2443 }
2444 
2445 U_CAPI UVoidFunction* U_EXPORT2
2446 uprv_dlsym_func(void *lib, const char* sym, UErrorCode *status) {
2447   HMODULE handle = (HMODULE)lib;
2448   UVoidFunction* addr = nullptr;
2449 
2450   if(U_FAILURE(*status) || lib==nullptr) return nullptr;
2451 
2452   addr = (UVoidFunction*)GetProcAddress(handle, sym);
2453 
2454   if(addr==nullptr) {
2455     DWORD lastError = GetLastError();
2456     if(lastError == ERROR_PROC_NOT_FOUND) {
2457       *status = U_MISSING_RESOURCE_ERROR;
2458     } else {
2459       *status = U_UNSUPPORTED_ERROR; /* other unknown error. */
2460     }
2461   }
2462 
2463   return addr;
2464 }
2465 
2466 #else
2467 
2468 /* No dynamic loading, null (nonexistent) implementation. */
2469 
2470 U_CAPI void * U_EXPORT2
2471 uprv_dl_open(const char *libName, UErrorCode *status) {
2472     (void)libName;
2473     if(U_FAILURE(*status)) return nullptr;
2474     *status = U_UNSUPPORTED_ERROR;
2475     return nullptr;
2476 }
2477 
2478 U_CAPI void U_EXPORT2
2479 uprv_dl_close(void *lib, UErrorCode *status) {
2480     (void)lib;
2481     if(U_FAILURE(*status)) return;
2482     *status = U_UNSUPPORTED_ERROR;
2483     return;
2484 }
2485 
2486 U_CAPI UVoidFunction* U_EXPORT2
2487 uprv_dlsym_func(void *lib, const char* sym, UErrorCode *status) {
2488   (void)lib;
2489   (void)sym;
2490   if(U_SUCCESS(*status)) {
2491     *status = U_UNSUPPORTED_ERROR;
2492   }
2493   return (UVoidFunction*)nullptr;
2494 }
2495 
2496 #endif
2497 
2498 /*
2499  * Hey, Emacs, please set the following:
2500  *
2501  * Local Variables:
2502  * indent-tabs-mode: nil
2503  * End:
2504  *
2505  */
2506