Bug #54050 [Opn->Dup]: microtime() returns random values

Edit report at https://bugs.php.net/bug.php?id=54050&edit=1 ID: 54050 Updated by: sixd@php.net Reported by: nospam at unclassified dot de Summary: microtime() returns random values -Status: Open +Status: Duplicate Type: Bug Package: *General Issues Operating System: Windows XP PHP Version: Irrelevant Block user comment: N Private report: N New Comment: Duplicate of bug #64370 Previous Comments: ------------------------------------------------------------------------ [2011-06-09 13:51:32] ynn at free dot fr There is a simpler test case : for ($i=0 ; $i<=100000 ; $i++) { $mt = microtime(true); if(isset($last) && $mt<$last) echo 'Last:'.$last.' - Current:'.$mt."\n"; $last = $mt; } On Windows XP, PHP 5.3.5, I get outputs that looks like : Last:1307620010.9549 - Current:1307620010.0533 Last:1307620010.9742 - Current:1307620010.0689 Last:1307620010.9742 - Current:1307620010.0689 Last:1307620010.983 - Current:1307620010.0845 Last:1307620010.9909 - Current:1307620010.0845 Last:1307620011.0142 - Current:1307620010.1158 Last:1307620011.0612 - Current:1307620010.147 Last:1307620011.0679 - Current:1307620010.1626 Last:1307620011.0865 - Current:1307620010.1783 Last:1307620011.115 - Current:1307620010.2095 ------------------------------------------------------------------------ [2011-02-18 23:31:48] nospam at unclassified dot de The seconds part of the return value seems to be my system boot time. I have now restarted the computer and the value points to that time. The fractional part is kind of oscillating around similar values. It's going up and down a bit every time but it has the tendency to decrease slowly. I have discovered a few other places in my application where it happens. But it's always the same database function. The same code. $start0 = microtime(true); $start = microtime(true); if (abs($start - $start0) > 1) { var_dump($start0); var_dump($start); } This should never print something but it does. Here are a few examples: float(1298065510.0493) float(1298068208.1106) float(1298065510.0533) float(1298068218.7475) float(1298065510.0405) float(1298068225.1549) float(1298065510.038) float(1298068231.7615) float(1298065510.0269) float(1298068237.0667) float(1298065510.0234) float(1298068244.8534) float(1298065510.0266) float(1298068250.9826) float(1298065510.0145) float(1298068257.6404) float(1298065510.8087) float(1298068265.3026) float(1298065510.809) float(1298068271.1601) float(1298065510.8087) float(1298068277.0327) float(1298065510.8029) float(1298068284.6319) ... ------------------------------------------------------------------------ [2011-02-18 22:32:24] nospam at unclassified dot de I have restarted the Apache web server and then the microtime function worked correctly in this situation a few times. Then it started to produce the same bogus data often. Now the original problem is right here again. Same place. Only the approximately 100th of like 200 times it is called. Instead of 1298064639.9228 (or similar) as everywhere else on the page it returns values around 1296908080.167 or 1296908081.2987. Quite a bit off but relatively "constant". ------------------------------------------------------------------------ [2011-02-18 22:17:00] nospam at unclassified dot de Description: ------------ I have a very weird bug here. The microtime() function with the parameter true works fine everywhere in my entire application - except in one single place. It's measuring database query durations here. It counts everything perfectly but always at a certain database query in the deepest middle of my application it returns random values. When I call it a second time just a line of code later, it returns the correct value again. It happens every time tonight. I cannot provide a testcase yet because there's another tens of thousands of lines of code hanging at this to work at all. I cannot upgrade PHP because there's no newer XAMPP package available. Version is 5.3.1. ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=54050&edit=1