Bug #10686 Updated: Bug in "mktime()" on values out of bounds

ID: 10686 User Update by: dieter@fiebelkorn.net Status: Open Bug Type: Date/time related Operating system: MacOS X 10.0.2 (Darwin) PHP Version: 4.0.5 Description: Bug in "mktime()" on values out of bounds Yes, it is "Darwin" specific. It is enough to use MacOS X 10.0.0 to 10.0.3 (MacOS X Server is not need!) OR you can use Darwin/Intel (see: http://www.darwinfo.de), if you don't have a Apple Macintosh. I used this patch for some weeks and an more primitiv patch some month. It works well and it can test on all other platforms you have. I will made some more test with some bigger values for each field... Whats the problem? You are using an undocumented feature and all other UNIX work in the same way. But it is not safe! Dieter Previous Comments: --------------------------------------------------------------------------- [2001-05-14 14:27:59] jmoore@php.net MacOS specific. Not critical untill we can find some MacOS X servers for developers/QA Team to test on. - James --------------------------------------------------------------------------- [2001-05-07 12:53:07] derick@php.net Marking as fix before 4.0.6 --------------------------------------------------------------------------- [2001-05-05 17:19:37] dieter@fiebelkorn.net see Bug id #8828, too! On PHP documantation: mktime(hour,min,sec, year,0,mon) refers the last day of month 'mon-1'. On MacOS X this failed, there it refers the first day of month 'mon' and mktime(hour,min,sec, year,-1,mon) refers the last day of month 'mon-1'. On MacOS X 10.0.2 (Darwin 1.3): 02.03 --> 983487600 --> 02.03 01.03 --> 983401200 --> 01.03 00.03 --> 983401200 --> 01.03 -1.03 --> 983314800 --> 28.02 -2.03 --> 983228400 --> 27.02 i have check some UNIX-systems and i haven't found any man-page, where is describe what should be happend on "tm_mday" <= 0. The problem is happend on all values out of bounds, when using mktime()! IMHO this is a undocumented feature, which is used by PHP. man mktime (Linux 2.2.13) tm_mday The day of the month, in the range 1 to 31. Sun Release 4.1 int tm_mday; /* day of month (1 - 31) */ SunOS 5.5 int tm_mday; /* day of the month - [1, 31] */ HP-UX Release 10.20 int tm_mday; /* day of month - [1,31] */ HP-UX Release 11.00 int tm_mday; /* day of month - [1,31] */ AIX 4.3.2 int tm_mday; /* Day of month (1 - 31) */ Darwin 1.0.2 int tm_mday; /* day of month (1 - 31) */ I think i should by much better to check mktime() on configure and set a #define for compilation. Only on "datetime.c" must be made a patch to support MacOS X's / Darwin's mktime()-systemcall, i think! Patch to make php_mktime() more compatible [i think it is not complete for the whole PHP4-project]. It includes a time correction for values less then 1 on "mday", "hour", "min", "sec", "mon"!! *** ext/standard/datetime.c.orig Fri Dec 8 12:38:02 2000 --- ext/standard/datetime.c Sun Apr 15 17:28:46 2001 *************** *** 81,87 **** struct tm *ta, tmbuf; time_t t; int i, gmadjust, seconds, arg_count = ZEND_NUM_ARGS(); ! int is_dst = -1; if (arg_count > 7 || zend_get_parameters_array_ex(arg_count,arguments) == FAILURE) { WRONG_PARAM_COUNT; --- 81,87 ---- struct tm *ta, tmbuf; time_t t; int i, gmadjust, seconds, arg_count = ZEND_NUM_ARGS(); ! int is_dst = -1, val, chgsecs = 0; if (arg_count > 7 || zend_get_parameters_array_ex(arg_count,arguments) == FAILURE) { WRONG_PARAM_COUNT; *************** *** 148,172 **** - (((*arguments[5])->value.lval > 1000) ? 1900 : 0); /* fall-through */ case 5: ! ta->tm_mday = (*arguments[4])->value.lval; /* fall-through */ case 4: ! ta->tm_mon = (*arguments[3])->value.lval - 1; /* fall-through */ case 3: ! ta->tm_sec = (*arguments[2])->value.lval; /* fall-through */ case 2: ! ta->tm_min = (*arguments[1])->value.lval; /* fall-through */ case 1: ! ta->tm_hour = (*arguments[0])->value.lval; /* fall-through */ case 0: break; } ! seconds = mktime(ta); if (is_dst == -1) is_dst = ta->tm_isdst; --- 148,182 ---- - (((*arguments[5])->value.lval > 1000) ? 1900 : 0); /* fall-through */ case 5: ! val = (*arguments[4])->value.lval; ! if (val < 1) { chgsecs += (1-val) * 60*60*24; val = 1; } ! ta->tm_mday = val; /* fall-through */ case 4: ! val = (*arguments[3])->value.lval - 1; ! while (val < 0) { val += 12; ta->tm_year--; } ! ta->tm_mon = val; /* fall-through */ case 3: ! val = (*arguments[2])->value.lval; ! if (val < 1) { chgsecs += (1-val); val = 1; } ! ta->tm_sec = val; /* fall-through */ case 2: ! val = (*arguments[1])->value.lval; ! if (val < 1) { chgsecs += (1-val) * 60; val = 1; } ! ta->tm_min = val; /* fall-through */ case 1: ! val = (*arguments[0])->value.lval; ! if (val < 1) { chgsecs += (1-val) * 60*60; val = 1; } ! ta->tm_hour = val; /* fall-through */ case 0: break; } ! seconds = mktime(ta) - chgsecs; if (is_dst == -1) is_dst = ta->tm_isdst; --------------------------------------------------------------------------- Full Bug description available at: http://bugs.php.net/?id=10686