|  support |  documentation |  report a bug |  advanced search |  search howto |  statistics |  random bug |  login
Doc Bug #14391 gmmktime, gmdate work incorrect
Submitted: 2001-12-09 07:29 UTC Modified: 2004-08-06 22:12 UTC
Avg. Score:4.8 ± 0.6
Reproduced:42 of 42 (100.0%)
Same Version:13 (31.0%)
Same OS:14 (33.3%)
From: pilots at farlep dot net Assigned:
Status: Closed Package: Documentation problem
PHP Version: 4CVS/5CVS OS: Windows 2000 Server
Private report: No CVE-ID:
 [2001-12-09 07:29 UTC] pilots at farlep dot net
Please see the code below. The comments are the result of running this code on my PC, (my GMT Offset = +2 hours = +7200 sec.)
You may see that the mktime() and date() functions work correct, but gmmktime() returns wrong value in winter time (when it is not a daylight saving time), and gmdate returns wrong "Z" parameter.
Note: this bug does not appear on UNIX systems.

$df="d.m.Y - H:i:s (Z), I";

echo	"<br><b>Make 02.01.1970 - 00:00:00 (Local):</b><br>",
	"Timestamp: $t<br>",			// 79200
	"GMT: ", gmdate($df,$t), "<br>",	// 01.01.1970 - 22:00:00 (7200)
	"Local: ", date($df,$t), "<br>",	// 02.01.1970 - 00:00:00 (7200)

echo	"<br><b>Make 02.01.1970 - 00:00:00 (GMT):</b><br>",
	"Timestamp: $t<br>",			// 82800 (Why !=86400 ?)
	"GMT: ", gmdate($df,$t), "<br>",	// 01.01.1970 - 23:00:00 (7200)
	"Local: ", date($df,$t), "<br>",	// 02.01.1970 - 01:00:00 (7200)

echo	"<br><b>Make 20.07.1970 - 00:00:00 (Local):</b><br>",
	"Timestamp: $t<br>",			// 17269200
	"GMT: ", gmdate($df,$t), "<br>",	// 01.07.1970 - 21:00:00 (7200)
	"Local: ", date($df,$t), "<br>",	// 02.07.1970 - 00:00:00 (10800)

echo	"<br><b>Make 20.07.1970 - 00:00:00 (GMT):</b><br>",
	"Timestamp: $t<br>",			// 17280000 (=86400*200)
	"GMT: ", gmdate($df,$t), "<br>",	// 02.07.1970 - 00:00:00 (7200)
	"Local: ", date($df,$t), "<br>",	// 02.07.1970 - 03:00:00 (10800)


Add a Patch

Pull Requests

Add a Pull Request


AllCommentsChangesGit/SVN commitsRelated reports
 [2002-01-20 10:54 UTC] max at flipstorm dot net
Here's an example:

first of all, to show that the system understands what's going on:

$ date "+%c %Z"
Tue Dec 18 22:51:16 2001 GMT
$ date -u "+%c %Z"
Tue Dec 18 22:51:21 2001 GMT

 - same result. so the machine knows what tz it's in.

Now, I'm in GMT at the moment, so 

echo date("H:i",mktime(22,20,00,12,18,01))."<p>";  //gives 22.20
echo gmdate("H:i",mktime(22,20,00,12,18,01))."<p>"; // gives 22.20

as expected. but...

echo date("H:i",gmmktime(22,20,00,12,18,01))."<p>";  //gives 21.20
echo gmdate("H:i",gmmktime(22,20,00,12,18,01))."<p>"; // gives 21.20

These should also give 22.20, so it looks to me that gmmktime() is screwed.

 [2003-08-18 19:22 UTC] lehresma at css dot tayloru dot edu
Posted this on php-dev and was advised to put it here as well.
A co-worker and I have been working on some intensive timestamp/timezone manipulation code for the past week or so, and ran across what we believed to be a bug in the way PHP handled gmmktime().  Browsing through the code, sure enough I found the source of the problem.  I fixed it, and have attached a patch.

As I was searching around, I discovered that this has been an open bug since 4.0.6 (2 years!).  I believe that this patch should fix most of the issues related to the following problem:

The Problem:
The function gmmktime() is supposed to take a date in GMT as a paramter and return a timestamp in GMT.  No manipulation of the time should occur. The only circumstance in which manipulation of the time should happen is if the seventh  parameter (is_dst) is set to 1.  This is the expected behavior.  However, in reality I noticed that when I didn't specify a seventh parameter, it was applying a daylight savings time offset when I tried to get the epoch.  See the example code.


    // Get a timestamp for the epoch (jan 1, 1970)
    echo "<pre>";
    echo gmmktime(0,0,0,1,1,1970,1)."\n";
    echo gmmktime(0,0,0,1,1,1970,0)."\n";
    echo gmmktime(0,0,0,1,1,1970)."\n";

    // Now get timestamp for June 1
    echo "\n";
    echo gmmktime(0,0,0,6,1,1970,1)."\n";
    echo gmmktime(0,0,0,6,1,1970,0)."\n";
    echo gmmktime(0,0,0,6,1,1970)."\n";


I expected the first set of outputs to look like this:


Which it does, so this is fine.  However, The second block of outputs is in June which if I had been using simply mktime(), I would have expected a DST offset (note: DST = daylight savings time) to be applied by default (on the 3rd command).  However, since I'm using gmmktime() by default no DST offset should be applied.  The odd behavior manifests  itself in the second block of output:


This is pretty meaningless until you realize what is actually going on. The last output, which is generated by a gmmktime() that has no 7th parameter, is applying an offset for daylight savings time.

The Solution
I had a moment of panic and was sincerely hoping that GMT did not actually use daylight savings time, so I did some research and found that while the timezone that contains Greenwich actually does use daylight savings time, the GMT standard time reference does not.  So right there I knew something was wrong on the PHP side.  I loaded up the source of PHP (like a good open source enthusiast), and sure enough, I discovered the source of the problem (no pun intended).

PHP is using the C function php_mktime() for both mktime() and gmmktime(). There is a flag as a parameter called "gm" which is 1 or 0 depending on whether it was called from mktime() or gmmktime().  The problem is that there is no check before applying the server's daylight savings information to the timestamp.

As a fix for this, I changed the behavior so that by default if there is no 7th parameter for gmmktime, it will not apply any DST offset.  This differs from mktime which by default will appy the server's DST offset.

I have attached patches for both PHP php-5.0.0b1 and php-4.3.2.  I would imagine the same code would work on most recent versions of PHP as it doesn't seem that this function has had much recent development.  The patch can be applied by using the following:

    $ cd /usr/local/php-4.3.2
    $ patch -p2 < ~/gmmktime-php_4.3.2.patch

Feel free to contact me if you have any questions.

PATCH for php4.3.2
diff -u clean/php-4.3.2/ext/standard/datetime.c php-4.3.2/ext/standard/datetime.c
--- clean/php-4.3.2/ext/standard/datetime.c     2003-05-04 07:22:00.000000000 -0400
+++ php-4.3.2/ext/standard/datetime.c   2003-08-18 17:18:19.000000000 -0400
@@ -116,7 +116,20 @@
        /* Let DST be unknown. mktime() should compute the right value
        ** and behave correctly. Unless the user overrides this.
-       ta->tm_isdst = -1;
+       /*
+       **  If we are using gmmktime(), do not use the local
+       **  server's setting of DST.  GMT does not use daylight
+       **  savings, so unless the user overrides us, this should
+       **  be set to 0.
+       **  - Luke Ehresman, Aug 2003, <>
+       */
+       if (gm) {
+               ta->tm_isdst = 0;
+               is_dst = 0;
+       } else {
+               ta->tm_isdst = -1;
+               is_dst = -1;
+       }
        ** Now change date values with supplied parameters.
 [2003-09-03 07:33 UTC]
From: Luke Ehresman <lehresma (at) css (dot) tayloru (ot) .edu>


Alright, after analyzing every line of the php_mktime() function, I finally figured out the problem.  There isn't actually a bug.  However, the PHP documentation is very deceptive.  The 7th parameter of the gmmktime function does not perform expected behavior.  In reality, there is no need for the functionality that it does provide.  The only reason it is there is so it is virtually the same as the mktime() function.

Anyhow..  By reading the documentation, what I gathered that it would do was to apply a daylight savings time offset to the GMT timestamp before returning.  What it really is doing is turning on or off the feature of compensating for the fact that the C "mktime" function already applied the server's daylight savings time preference.

With gmmktime, we should ALWAYS compensate for that.  I see no possible reason that you would not want to.

Anyhow, I think I am not the only one to get confused by this stuff. Perhaps it would be a good idea to either updating the documentation, or making the interfaces a little clearer in PHP, or both.

I'm kinda disappointed that it wasn't a bug though.  I was looking forward to having my PHP patch accepted.  :)

Thanks for your help,

 [2004-08-06 22:12 UTC]
This bug has been fixed in the documentation's XML sources. Since the
online and downloadable versions of the documentation need some time
to get updated, we would like to ask you to be a bit patient.

Thank you for the report, and for helping us make our documentation better.

"Parameters always represent a GMT date so is_dst doesn't influence the result."
PHP Copyright © 2001-2016 The PHP Group
All rights reserved.
Last updated: Sat Jul 30 18:01:41 2016 UTC