Diese Seite bei php.net
The PHP manual text and comments are covered by the Creative Commons Attribution 3.0 License © the PHP Documentation Group - Impressum - mail("TO:Reinhard Neidl",...)
(PHP 4, PHP 5)
session_start — Initialisiert eine Session
session_start() erzeugt eine Session oder nimmt die aktuelle wieder auf, die auf der Session-Kennung basiert, die mit einer GET- oder POST-Anfrage oder mit einem Cookie übermittelt wurde.
Um eine benannte Session zu verwenden, rufen Sie session_name() auf, bevor Sie session_start() aufrufen.
Wenn session.use_trans_sid aktiviert ist, registriert die Funktion session_start() eine interne Ausgaberoutine für das Umschreiben von URLs.
Verwendet ein Benutzer ob_gzhandler oder ähnliches mit ob_start(), dann ist die Reihenfolge der Funktionen wichtig für eine korrekte Ausgabe. Zum Beispiel muss ob_gzhandler vor Beginn der Session registriert werden.
Falls die Session erfolgreich gestartet wurde, gibt diese Funktion TRUE zurück, andernfalls FALSE.
| Version | Beschreibung |
|---|---|
| 5.3.0 | Falls der Start der Session fehlschlägt, wird FALSE zurückgegeben. Vorher wurde TRUE zurückgegeben. |
| 4.3.3 | Ab PHP 4.3.3 resultiert der Aufruf von session_start() nachdem die Session vorher gestartet wurde in einem Fehler der Stufe E_NOTICE. Der zweite Start der Session wird in diesem Fall einfach ignoriert. |
Beispiel #1 Ein Session-Beispiel: seite1.php
<?php
// seite1.php
session_start();
echo 'Willkommen auf Seite #1';
$_SESSION['farbe'] = 'grün';
$_SESSION['tier'] = 'Spinne';
$_SESSION['zeit'] = time();
// Funktioniert, falls das Session-Cookie akzeptiert wurde
echo '<br /><a href="seite2.php">Seite 2</a>';
// Oder die Session-ID mit übergeben, für den Fall, dass sie benötigt wird
echo '<br /><a href="seite2.php?' . SID . '">Seite 2</a>';
?>
Nach dem Aufruf von seite1.php enthält automatisch auch die zweite Seite, seite2.php, die Session-Daten. Lesen Sie Session Funktionen, um weitere Informationen über die Übermittlung der Session-ID zu erhalten. Dort wird zum Beispiel erklärt, worum es sich bei der Konstanten SID handelt.
Beispiel #2 Ein Session-Beispiel: seite2.php
<?php
// seite2.php
session_start();
echo 'Willkommen auf Seite #2<br />';
echo $_SESSION['farbe']; // grün
echo $_SESSION['tier']; // Spinne
echo date('d.m.Y H:i:s', $_SESSION['zeit']);
// Sie können auch hier, genauso wie bei seite1.php, SID verwenden
echo '<br /><a href="seite1.php">Seite 1</a>';
?>
Hinweis:
Um Cookie-basierte Sessions zu verwenden muss session_start() aufgerufen werden, bevor irgend etwas an den Browser geschickt wird.
Hinweis:
Die Verwendung von zlib.output_compression ist ob_gzhandler() vorzuziehen.
Hinweis:
In Abhängigkeit von der Konfiguration sendet diese Funktion verschiedene HTTP-Header. Siehe session_cache_limiter() um diese Header anzupassen,
PHP locks the session file until it is closed. If you have 2 scripts using the same session (i.e. from the same user) then the 2nd script will not finish its call to session_start() until the first script finishes execution.
If you have scripts that run for more than a second and users may be making more than 1 request at a time then it is worth calling session_write_close() as soon as you've finished writing session data.
<?php
// a lock is places on the session, so other scripts will have to wait
session_start();
// do all your writing to $_SESSION
$_SESSION['a'] = 1;
// $_SESSION can still be read, but writing will not update the session.
// the lock is removed and other scripts can now read the session
session_write_close();
do_something_slow();
?>
Found this out from http://konrness.com/php5/how-to-prevent-blocking-php-requests/
We are facing a problem and getting an error when loading a php page
Warning: session_start() [<a href='function.session-start'>function.session-start</a>]: Cannot send session cookie - headers already sent by (output started at /www/http/example/zh_CN/unsubscribe.php:1) in /www/http/example/inc/common.php on line 4
We had google and browse through some helps file and found some suggestion as below.
1. Use ob_start to cache any output before session_start function.
2. Make sure the session.save_path in php.ini is set and point to a valid path
We had tried the above but still having some the same problem.
We are using notepad ++ for the coding.
And we notice that this file is somehow being encoded to UTF-8 by one of my colleague because he though that this will help to display the result in the correct languages.
Solution:
Change the encode to use ANSI and make sure the EOL conversion is set to UNIX.
Bingo. Things are back to normal as expected.
Note:
The encode might cause this problem in certain machine but not others. So i had no clue on this. We had no problem running the php code with UTF-8 encode in a 64bits Linux Enterprise but hit the problem running in a 32 bits Linux slackware.
Hope this helps.
I recently stumbled upon a problem in my download script where a download link would be clicked but then no other links would work until that download had finished. Other instances/windows would not work either.
This is because session_start() locks the session data until the script finishes. In my case, until the script had finished sending the file data (which in a throttling script can take a while).
The solution is to use session_write_close();
This writes the current session data back to the session and closes the session so that other instances/windows can access the script using the same session.
Since adding it to my script my site can now have several files downloading at once from the same session and still serve pages.
Resolve for:
Warning: session_start() [function.session-start]: Cannot send session cookie - headers already sent by (output started at
is in changing text editor encoding to ANSI or UTF-8 without BOM.
Thanks God I found the answer!
my problem was in
session_start();
when i include this term in my first code it show me Warning Message: Cannot Modfie cache limiter
i had tryed many solves and dosn't work
finally
i write my code in other encode
and it sccussfully worked
i hope that help
Hi,
I just had this error saying that there is something wrong with session_start(); even if session_start(); is the only command in the script.
Warning: session_start() [function.session-start]: Node no longer exists in C:\xampp\htdocs\soui\index.php on line 3
Warning: session_start() [function.session-start]: Cannot send session cache limiter - headers already sent (output started at C:\xampp\htdocs\soui\index.php:3) in C:\xampp\htdocs\soui\index.php on line 3
I corrected this problem by deleting all the files in the tmp directory of the webserver, C:\xampp\tmp
A handy script that checks fot the presence of uft-8 byte order mark (BOM) in all files in all directories starting on current dir. Combined from the work of other people here...
<?php
function fopen_utf8 ($filename) {
$file = @fopen($filename, "r");
$bom = fread($file, 3);
if ($bom != b"\xEF\xBB\xBF")
{
return false;
}
else
{
return true;
}
}
function file_array($path, $exclude = ".|..|design", $recursive = true) {
$path = rtrim($path, "/") . "/";
$folder_handle = opendir($path);
$exclude_array = explode("|", $exclude);
$result = array();
while(false !== ($filename = readdir($folder_handle))) {
if(!in_array(strtolower($filename), $exclude_array)) {
if(is_dir($path . $filename . "/")) {
// Need to include full "path" or it's an infinite loop
if($recursive) $result[] = file_array($path . $filename . "/", $exclude, true);
} else {
if ( fopen_utf8($path . $filename) )
{
//$result[] = $filename;
echo ($path . $filename . "<br>");
}
}
}
}
return $result;
}
$files = file_array(".");
?>
There are some posts around about session_start() beeing terribly slow on ntfs. for me the problem occured from one day to another, system was already running fine for 2 years.
to get around it implementing a custom session_set_save_handler could be an option, although for me it worked just to put a session_write_close() somewhere near the end of my script. (maybe IIS didn't kill the php-threads fast enough?!)
For those of you running in problems with UTF-8 encoded files:
I was getting an error because of the BOM, although i set Dreamweaver to "save as" the without the BOM. It appears that DW will not change this setting in already existing files. After creating a new file withou the BOM, everything worked well.
I also recommend http://people.w3.org/rishida/utils/bomtester/index.php - a utility that remote checks for the presence of BOM.
Unfortunately, after pulling my hair out trying to figure out why my application was working fine in every browser other than IE ( Internet Explorer) (Opera, Chrome, Firefox, Safari are what I've tested this in) - when using a DNS CNAME record (like a vanity name that is different from the DNS A record, which is the hostname of the server) sessions do not work correctly.
If you store a session var while on the CNAME:
vanity.example.com and the hostname of the server is hosname.example.com
Then try to call the variable from a different page, it will not find it because of the CNAME (I guess it store the variable under the hostname, then when trying to read it it's still looking under the CNAME) the same application works fine when accessing it under the hostname directly. Keep in mind that I was testing this on an internal network.
I have found that you can use arrays with session variables in a very convenient manner. Be sure to unset the array afterwards!
The following is an example.
<?php
session_start();
$sessionvars['name'] = "John";
$sessionvars['age'] = "43";
$sessionvars['gender'] = "Male";
$sessionvars['email'] = "john.doe@example.com";
$_SESSION['arvars'] = $sessionvars;
unset($sessionvars);
echo $_SESSION['arvars']['name']."<br>";
echo $_SESSION['arvars']['age']."<br>";
echo $_SESSION['arvars']['gender']."<br>";
echo $_SESSION['arvars']['email']."<br>";
?>
Ever get an error like this:
Warning: Unknown: open(/tmp/sess_lLFJ,tk9eFs5PGtWKKf559oKFM3, O_RDWR) failed: Permission denied (13) ?
Don't ask me how, but I managed to fix it by calling
session_id() before session_start(), at the top of the page.
Try it. Hopefully it'll help someone else.
Maybe the guys here at PHP.net could explain why it worked :)
please note that at least in php 5.3.0 the setting session.use_only_cookies (in php.ini) defaults to "on". This means that using trans_id, and passing the PHPSESSID via the url does not work untill you change that setting.
Be warned that depending on end of script to close the session will effectively serialize concurrent session requests. Concurrent background "data retrieval" (e.g. applications such as AJAX or amfphp/Flex) expecting to retrieve data in parallel can fall into this trap easily.
Holding the session_write_close until after an expensive operation is likewise problematic.
To minimize effects, call session_write_close (aka session_commit) as early as practical (e.g. without introducing race conditions) or otherwise avoid the serialization bottleneck.
if you store your sessions in a database, always ensure that the type of the database column is large enough for your session values
PHP Manual specifically denotes this common mistake:
Depending on the session handler, not all characters are allowed within the session id. For example, the file session handler only allows characters in the range a-z A-Z 0-9 , (comma) and - (minus)!
See session_id() manual page for more details.
It seems like spaces in the name don't work either - got a new session id generated each time
Working with Comet, or any other parallel or simultaneous requests from a single browser (a single session) would lock the application if it uses a centralized session management.
To fix this, you should add session_commit() (Alias of session_write_close() ) after ALL session management, i.e using any session_* functions except session_start() is forbidden after that an causes an error.
When using session_start() multiple times in one script, it seems to only maintain $_SESSION variables with string keys.
<?php
session_start();
$_SESSION[0] = 'this is a test';
$_SESSION[1] = 'this is also a test';
$_SESSION['test0'] = 'this is a test';
$_SESSION['test1'] = 'this is also a test';
session_write_close();
session_start();
echo "Session var 0 = ".$_SESSION[0]."<br />";
echo "Session var 1 = ".$_SESSION[1]."<br />";
echo "Session var 'test0' = ".$_SESSION['test0']."<br />";
echo "Session var 'test1' = ".$_SESSION['test1'];
?>
Outputs:
Session var 0 =
Session var 1 =
Session var 'test0' = this is a test
Session var 'test1' = this is also a test
If two different files don't access the same session, it can only mean one thing: they aren't on the same directory level.
Examples:
a) One is under https, and one is under http. (very common and annoying error)
b) One is under /, another is under /dir1, and /dir1 was first to run. The cookie created by the session is for /dir1 and deeper only, so the other script can't read it; it sees no session so it starts a new one.
Solutions:
1) Session start should always happen at root dir. If one of your scripts discovers user is not logged in, for example, use session_destroy() and send him to the login in the root dir. Scripts that use session without requiring login needs more creative solutions, like redirecting to the root dir, setting the session, and redirecting back.
2) Use SID from page to page, instead of cookies. The upside is that this also works with http/https. The downside is very ugly URL codes and more difficult maintainability ('which pages links to gallery.php without using SID?!').
I ran into a problem that has a solution that may not be quite intuitive to people (well, atleast it wasn't for me). I had one script that called session_start() and that script would run for an indefinite amount of time. I had a second script that would also call session_start(), but would only run for a few seconds each time it is called.
Here is where the problem occurs: Once session_start is called from the long living script, no other script can call session_start until that long running script calls session_commit() (session_write_close()) or ends.
In any case, this may have seem very intuitive to others, but having an abstracted system in place, I never see the session management at the script level, so it was hard to figure out what was locking any other requests.
Hope this helps others.
I found myself in trouble when using $_SESSION['varname']. The value from varname was not shown with echo $_SESSION['varname'] in the page where I'd like to use that value.
After many tries I found that session_start() was missing in the page where I'd like to use the value from varname. When I put de session_start() in that page the problem was solved.
--
Tive problemas utilizando $_SESSION['variavel']. A variavel não aparecia na página que deveria utiliza-la.
Após muitas tentativas pude verificar que não estava usando session_start() na página que receberia o valor da variável.
Inseri session_start() nessa pagina tambem e funcionou perfeitamente!
If you want to check if a session has already been activated, you can check $_COOKIE["PHPSESSID"] for a value before reactivating that session.
Example code:
== login.php ==
<?php
session_start();
$_SESSION["data"] = "you have an active session!";
echo "<a href=\"page.php\">go</a>";
?>
== page.php ==
<?php
if(isset($_COOKIE["PHPSESSID"]))
{
session_start();
echo $_SESSION["data"];
}
else
{
echo "no cookies required!";
}
?>
This can prevent forcing a cookie on someone (very annoying in lynx) that doesnt intend to login on your site.
I kept getting:
Warning: session_start() [function.session-start]: Cannot send session cache limiter - headers already sent (output started at myfile.php:2)
Only when I had uploaded my pages from my home development server to my hosted website. the same page worked fine in development, so there must be a difference in the php setup, but since this is a site that hosts many different clients web sites, I can't get them to change thier config.
After a little bit of research and trial and error, I found the cure was to move the line:
<?php session_start(); ?>
to be the very 1st line in any .php file.
orginally I had:
<!DOCTYPE HTML PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<?php session_start(); ?>
<html>
<head>
...etc
so I changed it to :
<?php session_start(); ?>
<!DOCTYPE HTML PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html>
<head>
...etc
and all is well...
The clue is in the warning message, it ends :2) which means the session_start() call is not on line 1 of the file!
Hope this helps & good Luck
Loz
I have tried many times a lot of login scripts but sometimes when you have a lot of traffic you can note the "overloading" of the server tring to deal with all sessions specialy with the long, long scripts and classes ... checking all stuffs (ids, ips, usernames, etc) the loading page increase ...specialy when you have a "home" server with a small blog..
if you open the folder "/tmp" you will thousands of files that do nothing else that increasing the size of the folder.
Sessions are good and I just offer an alternative.
.. and for that I just created in about 20 min a simple script that can handle a simple login.
<b>(Be aware !! this is just an example .. it not make use of any check of xss atacks or other scripts !!!</b>
and make use of cookies [today all moderns browsers use cookies])
<?php
$mysqli = new mysqli('localhost', 'root', 'pass', 'nameofDB');
if (mysqli_connect_errno()) {
printf("Connect failed: %s\n", mysqli_connect_error());
exit(); }
$cookie = (isset($_COOKIE['vuid'])) ? $_COOKIE['vuid'] : '';
if(!$cookie) {
$user_level = 0;
setcookie('vuid', '0.' . md5('anonymous')); }
else {
$part = explode('.', $cookie);
$user_level = (int) $part[0];
$user_name = (string) $part[1];
if($user_level != 0) {
$is_logged = check_login($user_level, $user_name); }
else {
$is_logged = false; } }
if(isset($_GET['login'])) {
// here you MUST make a form and a function that realy check the login ...
if($YourFunctionCheckIsOk) {
setcookie('vuid', /*levels betwen 1-9*/ '9.' . md5('PutTheUserNameHere'));
echo '<meta http-equiv="refresh" content="1;URL=index.php">
Correct Login<br>
You will be redirected to the main page'; } }
elseif(isset($_GET['logout'])) {
setcookie('vuid', '0.' . md5('anonymous'));
echo '<meta http-equiv="refresh" content="1;URL=index.php">
You are out now ... see you soon<br>
You will be redirected to the main page'; }
else {
if($is_logged) {
echo 'Welcome ' . USERNAME . ' <a href="?logout">LOGOUT</a>'; }
else {
echo 'You are a Anonymous. LOGIN <a href="?login">HERE</a>'; } }
function check_login($user_level, $user_name) {
global $mysqli;
$is_logged = false;
$result = $mysqli->query('SELECT nick FROM accounts WHERE user_level = ' . $user_level);
$row = $result->fetch_array();
$result->close();
if($user_name === md5($row['nick'])) {
define('USERNAME', $row['nick']);
return $is_logged = true; }
return $is_logged = false; }
$mysqli->close();
?>
PS: you can change the nick with the ip or better use both !
like:
"level.md5(sha1(nick)).sha1(md5(ip)) etc ...
"0.4aea7c7d495676.4aea7c7d495676"
and check them separetly by <a href="http://es.php.net/explode">"exploding"</a> them ..
play arround ...
Be aware that if a browser isn't accepting cookies, it can cause the session to invisibly not save between two pages.
So for example, in Firefox 3.0.6 there is a bug where "localhost" addresses are currently not saving cookies. This will kill any attempts to use session code in development where localhost is used.
One workaround for this is to use a different synonym for the localhost address, i.e. http://127.0.0.1/, or set up a virtualhost, e.g. http://localdevelopment/
Overall, pay close attention if you see to different, unexpected session behavior in two different browsers, it may be a result of the cookies being handled differently or being dropped.
Here is some info about how to use SID with session_start().
Experiment run on WAMPserver, PHP version 5.2.6.
Using FireFox 3.0.
The definition found at http://www.php.net/manual/en/session.constants.php
describes SID is a “pre-defined constant” containing either:
1). the session name and session ID in the form of "name=ID" .
This is the same name as would be returned by session_name(), and the same id as would be returned
by session_id().
2). Or an empty string, if the session ID was set in an appropriate session cookie.
Example code:
<?php
session_start();
echo '<a href="page2.php?' . SID . '">page2</a>';
?>
This code will yield in the browser one of the following:
a). “page 2” (as a hyperlink, but mouseover shows the link is to something in the form of
http://www.example.com/page2.php?PHPSESSID=e2bd7072.... )
b). “page 2” (as a hyperlink, but mouseover shows http://www.example.com/page2.php? )
Which result is determined by whether or not the browser is sending a matching session cookie back to the script. Result “a” above always appears at least initially since the script hasn’t yet set a cookie and neither has it been done automatically yet by the php.ini setting session.use_cookies. However, after refreshing or using the back button, the form in "b" will appear because then the cookie exists (either you set it or session.use_cookies did it automatically).
If while testing locally on your Windows machine you get many warnigns like:
Warning: session_start()... failed: No such file or directory
Warning: session_start()...: Cannot send session cache limiter - headers already sent
etc.
you need to configure properly the session.save_path in your php.ini file.
Set session.save_path to an existing folder on your PC.
For example: session.save_path="C:\Temp";
I found i was getting a lot of
"Warning: session_start() [function.session-start]: Cannot send session cache limiter - headers already sent "
Warnings, on some pages, where as on other pages it worked just fine.
After some time looking around and trying things like ob_start() and session_write_close()... nothing seemed to work.
Then i realized the encoding of the page. I checked the encoding of the page and it was set to UTF-8, so I changed it to ANSI and now the Warnings went away.
Good luck any one else.
When using session_start() with cookies over an encrypted connection (like HTTPS) it is important to set php_ini('session.cookie_secure',true). This sets a flag in the cookie which will stop browsers from sending this cookie over an unsecured connection. If this is not done, a Man-In-The-Middle could easily cause the victim's browser to send his session ID in plaintext, gaining access to the account and defeating the goal of the secure connection.
http://fscked.org/blog/fully-automated-active-https-cookie-hijacking
<?php
class SESSION
{
public function __construct ()
{
self :: commence ();
}
public function commence ()
{
if ( !isset( $_SESSION ['ready'] ) )
{
session_start ();
$_SESSION ['ready'] = TRUE;
}
}
public function set ( $fld , $val )
{
self :: commence ();
$_SESSION [ $fld] = $val;
}
public function un_set ( $fld )
{
self :: commence ();
unset( $_SESSION [$fld] );
}
public function destroy ()
{
self :: commence ();
unset ( $_SESSION );
session_destroy ();
}
public function get ( $fld )
{
self :: commence ();
return $_SESSION [$fld];
}
public function is_set ( $fld ) {
self :: commence ();
return isset( $_SESSION [$fld] );
}
}
?>
This Class is handy to all who want easy session handling.
Just include () or require () this to pages where you will have to use session.
example:
<?php
include ( "/path/to/sess.inc.php" );
if( !SESSION :: is_set ( "number" ) )
{
SESSION :: set ( "number" , 1 );
}
else
{
$_SESSION ["number"] ++;
}
echo SESSION :: get ( "number" );
?>
A session created with session_start will only be available to pages within the directory tree of the page that first created it.
i.e. If the page that first creates the session is /dir1/dir2/index.php and the user then goes to any page above dir2 (e.g. /dir1/index.php), session_start will create a new session rather than use the existing one.
for resolve the problem with session_start()
Cannot send session cache limiter - headers already sent (output started at ..."
write it at first line of the page like
<?php session_start(); ?>
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<body>
<?php $_SESSION['name']='mohamed'; ?>
</body>
<html>
@ash
Whilst this may prevent session hijacking attempts, it also ruins the user experience, pressing back and forward will result in the 'security' token being out of sync. Also opening new tabs/ windows will also cause issues.
A better method would be to simply check the ip address or ip block of the user. But this is not without problems either.
To access highly sensitive information, the user should be required to enter his/her password again, even tho they are logged in. This ensures that session hijackers can not access this information.
All this goes hand-in-hand with regenerating session ids on login, and good use of https.
@My last post...
the function will generate the session "token" you just need to do a check and regenerate it on each page load...
When developing applications or just creating a general website with the use of sessions, many developers do not think about securing session hijacking attacks. For further information I recommend searching google, however I have produced a small function to be called immediately after session_start().
your_page.php:
<?php
session_start();
include_once 'session_secure.inc.php';
session_secure();
# Your content here.
?>
session_secure.inc.php :
<?php
function session_secure(){
// wrapped for the php entry....
$alph =array('A','a','B','b','C','c','D','d','E',
'e','F','f','G','g','H','h','I','i','J','K','k',
'L','l','M','m','N','n','O','o','P','p','Q','q',
'R','r','S','s','T','t','U','u','V','v','W','w',
'X','x','Y','y','Z','z');
for($i=0;$i<rand(10,20);$i++){
$tmp[] =$alph[rand(0,count($alph))];
$tmp[] =rand(0,9);
}
return implode("",shuffle($tmp));
}
?>
There are quicker ways like md5(time()*rand()), however the function above is completely random, and will render an attackers hijacking task almost impossible.
Kind of ankward it's not mentioned here, but you should also trigger the session_write_close() function as soon as possible so you'll release the lock over the session files.
Cheers,
Vladimir Ghetau
If your sessions are failing to hold the stored data ensure that session.use_only_cookies is set to 1 in your PHP.ini config.
Also using the session auto start saves on a lot of issues and removes the issue of people setting the session_start after the html begins.
session_start() generate a warning if PHPSESSID contains illegal characters
Warning: session_start() [function.session-start]: The session id contains illegal characters, valid characters are a-z, A-Z, 0-9 and '-,' in /home/para/dev/mon_site/header.php on line 17
To avoid i wrote this :
<?php
function my_session_start()
{
if (isset($_COOKIE['PHPSESSID'])) {
$sessid = $_COOKIE['PHPSESSID'];
} else if (isset($_GET['PHPSESSID'])) {
$sessid = $_GET['PHPSESSID'];
} else {
session_start();
return false;
}
if (!preg_match('/^[a-z0-9]{32}$/', $sessid)) {
return false;
}
session_start();
return true;
}
?>
That's why you shouldn't use register_globals and refer to
$_GET['id']
for the page.php?id=1
or if you don't care if it is GET or POST you can use
$_REQUEST['id']
Those will not be cleared by session_start().
To complete: you should use $_SESSION['id'] to get the content of the session.
I just wrote that session_start() will erase your querystring variable(s) once called. I want to clarify that it will only do this if a variable by the same subscript is defined in $_SESSION[].
Warning, calling session_start() will reset your querystring values.
Navigating to page.php?id=1
<?php
echo "id = $id"; // 1
session_start();
echo "id = $id"; // ""
?>
I also found using session_write_close(); after you send your custom headers sometimes helps on some servers. This protects any data from being cleared out by some of these errors mentioned.
Having a problem in IE causing your getFile.php script to be pushed to the browser as an empty text file, instead of the intended attachment?
If your php config causes session_start() to output a cookie, or anything funky, IE will not like the filename and will also try and read the file contents inline into the browser window.
Don't use session_start() or USE Output Buffering after the session is started.
<? //session_start();
function force_download($file)
{
if ((isset($file))&&(file_exists($file))) {
$fileArray = explode("/", $file);
header("Content-type: application/force-download");
header("Content-Disposition: attachment; filename=".$fileArray[sizeof($fileArray)-1]);
header("Content-Transfer-Encoding: Binary");
header("Content-length: ".filesize($file));
header('Content-Type: application/octet-stream');
header("Content-Disposition: attachment; filename=".$fileArray[sizeof($fileArray)-1]);
readfile("$file");
} else {
echo "No file selected";
} //end if
}//end function
?>
File Encoding was causing my session_start() statement to generate an unexpected warning.
Environment:
OS: Windows XP Pro (32bit)
PHP: v5.2.5
Apache: v2.0.61
I was having this warning:
Warning: session_start() [function.session-start]: Cannot send session cache limiter - headers already sent (output started at ...
I was rather pissed off because I know of this issue and I know how to fix it:
1) Just make sure the session_start() directive is processed before any other statements that modify the content headers.
2) Failing #1 above, you should also make sure that your script files do not have any beginning/trailing space or carriage return characters.
3) As a last stand, you could use output buffering to hold your stuff in memory and then spit it out at the very end. This would essentially allow you to manipulate the headers to your heart's content without getting any crap from php.
Note: with regards to #2 above, I know some devels use the interesting construct of no closing php braces e.g.:
> info.php
<?php
phpinfo();
?> (they omit this)
It seems the tag is not required by php (at least php does not puke) so the parser does not treat any trailing spaces as output to be sent to the user agent. I'm not using this loop-hole at the moment.
Anyway the above resolutions *did not* work for me (yer even #3) and I was about to give up and just gag php with good ol "@".
It was then that I thought to check the encoding of my php file. Turns out I had left my editor in UTF-8 encode mode from some previous (i18n) work I had done. I switched to ANSI and that fixed the problem. Suddenly the same file that had caused php to scream in pain was being processed without any issues.
I also noted that I could set the editor to "Encode in UTF-8 without BOM" and that also works. The BOM is the byte order mark which php probably glosses over (or does not understand). I didn't look too far into it.
I hope this helps someone else.
Cheers!
You probably know it but in case somebody did not:
session_start() must be called BEFORE any explicit operations on the $_SESSION variable; otherwise $_SESSION will be treated as an ordinary variable and not stored for future use.
If you encounter "The session id contains illegal characters" error, find out, if you get an empty session id.
If so, the problem ist fixed easily:
if(!session_id()) session_regenerate_id();
Don't *ever* use /dev/random as a session entropy file, because session_handler will hang for a *long* time after open(). This configuration will be *the* overkill:
<?php
ini_set("session.entropy_file", "/dev/random");
ini_set("session.entropy_length", "512");
?>
Use /dev/urandom instead. The reason for this is (from Wikipedia):
A counterpart to /dev/random is /dev/urandom ("unlimited" random source) which reuses the internal pool to produce more pseudo-random bits. This means that the call will not block, but the output may contain less entropy than the corresponding read from /dev/random.
We spent 2 days on figuring this out. You have been warned.
At work we had an odd warning, "Cannot send session cache limiter - headers already sent". In the following:
header('meta:1');
session_start();
the first header went across fine, but the session_start died with the warning.
What was happening was that the session contained an object which fired __autoload() when it was unserialized. The __autoload() then included the file with the class definition, but that file also output non-header text. This caused session_start() to be unable to add the session_id to the headers.
This was difficult to find because we weren't able to see the text that was output from the file that was __autoload()ed, only the warning.
Regarding the posts discussing ways of checking if a session is active, I've found that this is the best way, aslong as you dont assign any variables to the $_SESSION superglobal before calling the function.
The Function:
<?php
function session_started(){
if(isset($_SESSION)){ return true; }else{ return false; }
}
?>
An Example:
<?php
function session_started(){
if(isset($_SESSION)){ return true; }else{ return false; }
}
//Start the output buffer so we dont create any errors with headers
ob_start();
//Check to see if it has been started
if(session_started()){
echo 'The session has been started.<br />';
}else{
echo 'The session has not been started.<br />';
}
//Start the session
echo 'Starting Session...<br />';
session_start();
//Check again
if(session_started()){
echo 'The session has been started.<br />';
}else{
echo 'The session has not been started.<br />';
}
//Flush the buffer to screen
ob_end_flush();
?>
Hope this helps some people.
Beware to use $_SESSION as parameter to a function.
(here: php5 5.1.2)
Use:
$s=array(); foreach($_SESSION as $k => $v) $s[$k]=$v;
$result = VeranstaltungTermin::search($s, $quellen, $KatCount);
otherwise the garbage-collection of php5 will destroy at end of script Your $_SESSION array before writing to disk
Here's another session tip for you, relevant to IE6:
If you find that your session is unexpectedly lost when your application opens a page in a pop-up window, try closing down any other copies of IE running on your machine.
The problem is that IE doesn't always use the same IE process to host a new pop-up window - it chooses one (using some mysterious strategy) from the pool of running IE processes! Since session information (cookies) is not preserved across processes, the pop-up may loose the session (and probably redirect the user to your log-in page.)
I don't know a work-around for this, except for close all other IE windows. Microsoft state that they do not guarantee that sessions will be preserved across multiple windows: http://support.microsoft.com/default.aspx/kb/196383
TAGS: session_start headers output errors include_once require_once php tag new line
Errors with output headers related to *session_start()* being called inside include files.
If you are starting your session inside an include file you must be aware of the presence of undesired characters after php end tag.
Let's take an example:
> page.php
<?php
include_once 'i_have_php_end_tag.inc.php';
include_once 'init_session.inc.php';
echo "Damn! Why I'm having these output header errors?";
?>
> i_have_php_end_tag.inc.php
<?php
$_JUST_A_GLOBAL_VAR = 'Yes, a global var, indeed';
?>
> init_session.inc.php
<?php
session_start();
$_SESSION['blabla'] = 123;
?>
With all this stuff we will get an error, something like:
"... Cannot send session cache limiter - headers already sent (output started at ...", right?
To solve this problem we have to ignore all output sent by include files. To ensure that we need to use the couple of functions: *ob_start()* and *ob_end_clean()* to suppress the output. So, all we have to do is changing the *page.php* to this:
<?php
ob_start();
include_once 'i_have_php_end_tag.inc.php';
include_once 'init_session.inc.php';
ob_end_clean();
echo "Woo hoo! All right! Die you undesired outputs!!!";
?>
I had the same "session data lost" issue. I couldn't solve it with a single session_start(). Instead I had to include session_start() in each page.
Another X File archived...
For those encoding their pages in UTF-8 but running into problems with BOM data (Byte Order Marks) being outputted before anything else - such as $_SESSION data - but wish to use UTF-8 encoding, and who do NOT wish to use UTF-8 No BOM encoding because the latter produces weird results when using foreign language accented characters, try the following :
- Save your document with UTF-8 No BOM encoding (on a mac, using BBEdit, this is normally found on the very bottom left corner of the window). This will produce a document that will not output BOM data.
- then, in your scripts, include the following:
header('Content-Type: text/html; charset=UTF-8');
before your call to session_start()
hope that helps someone.
If you ever want to know whether a session is open or has already been closed (e.g. with session_write_close()) when you need to open one (in my case, writing a framework and, therefore, not controlling the session directly), I have been unable to find an official check. However, I did notice that session_start() has the following behaviour:
* If the session has been open before but is now closed, when you call session_start() the $_SESSION array is reset to the content of the session.
* If the session is already open, the $_SESSION array is left totally untouched.
Therefore, I came up with the following (obviously assuming you have already tested $_SESSION exists):
<?php
$_SESSION['testMe']='test';
@session_start();
if(isset($_SESSION['testMe']))
{
echo "Session was open\n";
unset($_SESSION['testMe']);
}
else
echo "Session was closed\n";
?>
Until someone can tell me the 'official' way to do it, this seems to work.
Just to save some confusion, $_SESSION, session_id() and the SID define all maintain their values after a session_write_close() so are of no use in identifying if a session is currently open or not. Hence this trick.
How many times you've seen this error when actually your code is perfect:
Warning: session_start(): The session id contains invalid characters,
valid characters are only a-z, A-Z and 0-9 in
Usually there are MANY reasons why such an error can appear into your logs, page, etc. (one way to generate it, is to send PHPSESSID='').
But it wasn't our case. We checked everything, and we even logged the entire request ($_GET, $_POST) that was generating the error.
Still no luck. Until one point, when we checked what the session cookie contains.
It seems there are few bots that are trying to buffer overflow (?) a page, using session cookies manually changed to contain something else than session id (strlen == 32 and hex number) The bot in our case was trying to mass submit spam on pages.
So we needed to check and destroy any session id that do not comply.
After adding this code:
$sn=session_name();
if(isset($_GET[$sn])) if(strlen($_GET[$sn])!=32) unset($_GET[$sn]);
if(isset($_POST[$sn])) if(strlen($_POST[$sn])!=32) unset($_POST[$sn]);
if(isset($_COOKIE[$sn])) if(strlen($_COOKIE[$sn])!=32) unset($_COOKIE[$sn]);
if(isset($PHPSESSID)) if(strlen($PHPSESSID)!=32) unset($PHPSESSID);
session_start();
we got rid of that.
( Note: The code is before session_start() )
Strange, kinda paranoid, but effective.
If you are getting headers sent error check what is the encoding of your file. In my case i had utf-8, so i changed it to ansi, and everything worked.
hope it helps!
It seems that the session file is opened exclusively.
On some occasions (Windows) I have found that the file lock is not released properly for whatever reason, therefore causing session_start() to hang infinitely on any future script executions.
My way round this problem was to use session_set_save_handler() and make sure the write function used fopen($file, 'w') instead of fopen($file, 'x')
I had a problem accessing a session which is open in another frame at the same time.
I had a frame, which is displaying a live text like a chat. It has an opend and locked session. I tried to add data to this session with another frame, but this wouldn't work.
In the first frame i had a loop which is looking for fresh data in the session and prints it out like this:
<?php
session_start();
[...]
while (true) {
echo $_SESSION["input"]."\n";
sleep(1);
}
?>
But this locks the session, and no other frame can access the session data at the same time.
So i modified this loop a bit like this:
<?php
session_start();
[...]
session_commit();
while (true) {
ob_start(); session_start(); ob_end_clean();
echo $_SESSION["input"]."\n";
session_commit();
sleep(1);
}
?>
Now the second frame is able to access the session.
The ob_start() and ob_end_clean() is just for surpressing the cookie output to the browser.
With IE and FF, session_start() appears to stop form values being preserved when the user browses 'back' to a page (I'm using SBS2K3 IIS6 PHP5.1.1).
For example if you set some hidden form values from javascript in one page, then navigate to another page and come back using the browsers back button, the form values should be preserved. But this does not happen when your your form page uses session_start().
The following seems to sort this out nicely:
session_cache_limiter('public');
session_start();
A bedder way to check wether the session is started including ip logging, so it will be a little saver. in 4 small lines.
<?php
if( !isset( $_SESSION ) ) { session_start(); }
if( isset( $_SESSION['REMOTE_ADDR'] ) && $_SESSION['REMOTE_ADDR'] != $_SERVER['REMOTE_ADDR'] )
{ session_regenerate_id(); $_SESSION['REMOTE_ADDR'] = $_SERVER['REMOTE_ADDR']; }
if( !isset( $_SESSION['REMOTE_ADDR'] ) ) { $_SESSION['REMOTE_ADDR'] = $_SERVER['REMOTE_ADDR']; }
?>
Yes, the code is a little bit self-willed ;)
James at skinsupport dot com raises a good point (warning) about additional requests from the browser. The request for favicon.ico, depending on how it is handled, can have unintended results on your sessions.
For example, suppose you have ErrorDocument 404 /signin.php, no favicon.ico file and all pages in your site where the user signs in are also redirected to /signin.php if they're not already signed in.
If signin.php does any clean up or reassigning of session_id (as all good signin.php pages should) then the additional request from the browser for favicon.ico could potentially corrupt the session as set by the actual request.
Kudos to James for pointing it out and shame on me for skimming past it and not seeing how it applied to my problem. Thanks too to the Firefox Live HTTP Headers extension for showing the additional request.
Don't waste days or even hours on this if your session cookies are not being sent or if the session data isn't what you expect it to be. At a minimum, eliminate this case and see if any additional requests could be at fault.
Another one bitten by the "a different domain name gives you a new session" in IE...
My site uses interactive PDF documents. A user saves data from them into FDF documents, and when this happens, the site also includes the URL of the PDF in the FDF data (using the FDF "/F" tag).
This URL is generated using the $_SERVER["HTTP_HOST"] value from PHP. For some as-yet-undetermined reason, when my hosting provider updated PHP (I believe to 4.4.1), for a time the $_SERVER["HTTP_HOST"] value apparently dropped the "www." from my domain name. So, some FDF files were saved with this different URL value. Needless to say, when one of these FDFs loaded the corresponding PDF, although the load and display of the PDF worked, any subsequent save operation (which relies on current session information for validating the user, etc.) initiated from within the PDF failed. The original session had been established for the "www.domain.com", but because the PDF URL lacked the "www.", PHP/IE had established a new, essentially empty, session.
I scratched my head over this problem for months and only recently happened upon this fix. In my case, manually patching the failing (binary) FDF files to include the "www." in the URL for the PDF did fix it.
Hopefully someone else finds this helpful...!
For some reason, MSIE gets confused when session_start() is called and the Content-disposition has been set to "attachment". The only solution I've found is this:
if(strpos($_SERVER['HTTP_USER_AGENT'],"MSIE")) {
session_cache_limiter('public');
}
session_start();
See also:
http://joseph.randomnetworks.com/archives/2004/10/01/
making-ie-accept-file-downloads
Recursive session usage isn't possible. By that I mean fetching (via curl or such) another page on the site which uses sessions, as part of a page which itself uses sessions.
For example:
page1.php:
<?php
session_start();
$curl = curl_init();
curl_setopt($curl, CURLOPT_URL, "http://example.com/backend.php");
curl_exec($curl);
?>
backend.php:
<?php
session_start();
print $_SESSION['foo'];
?>
In this case, requesting page1.php will hang the request indefinitely. backend.php is waiting to open the session data for reading/writing, but it's still locked by page1.php, which is waiting for backend.php to finish.
Because there's no session_end(), you can't get PHP to relinquish the lock on the session data after session_start() has occured. The solution is just to include backend.php or its functionality directly, e.g.:
require './backend.php';
If your session_start() does not set a cookie, but instead just generates a new session_id on every request, you should check if your script is echoing something before the function call. Even a line change in the code before the session_start() function prevents the cookie to be set.
In reference to the "Cannot send session cache limiter..." error message, I have tried a number of ways to supress the error. An If/Then block will work only if you store the flag in a session variable, cookie, or other method. Or, you can obviously supress error messages by the error level. However, the simplest way to do it is a simple:
@session_start();
This will hide all error messages resulting from that line of code. I would only recommend this method AFTER you have fully tested your code but it is very effective when that is your only error and you do not want to completely stop code exectution by means of a die() or exit() function.
when you're using session trans sid ENABLED, and you perform redirect such as Header( "Location..." ), redirected client will SURELY loose his session.
to avoid this, you can use a modified version of redirect function, like this:
<?php
function Redirect302( $location )
{
$sname = session_name();
$sid = session_id();
if( strlen( $sid ) < 1 )
{
Header( $location );
return;
}
if( isset( $_COOKIE[ $sname ] ) || strpos( $location, $sname."=".$sid ) !== false )
{
Header( $location );
return;
}
else
{
if( strpos( $location, "?" ) > 0 )
$separator = "&";
else
$separator = "?";
$fixed = $location . $separator . $sname."=".$sid;
Header( $fixed );
return;
}
}
?>
expamle:
Redirect302( "Location: /YourPage/index.php" );
the function performs common redirect if no session is active, and appends session "name=sid" to query string when session is already started, like PHP parser does when outputting your page.
hope, you'll find it useful.
yours:)
I ran into the problem of losing the session after i redirect the user with header( location:);
I found that the problem doesn't occur in php5. Atleast for myself. I tested it on IIS as well as Apache and both times using php5 fixed the issue.
Secondly, I found a weird solution: The website i'm building required Cyrillic type, so i encode the pages using UTF-8. The problem that happens is that UTF-8 sends information before the php tags, and therefore before the session_start(); which ultimately renders the session_start(); call ineffective.
The UTF-8 encoding sends a small chunk of information before anything. This is called BOM (byte order marks).
To ensure that they don't show up, you need to save your files with a UTF-8 encoding which bypasses BOM. BBEdit for the Macintosh with allow for this under the file encoding options.
To recap: In this instance and for my situation, UTF-8, no BOM will allow sessions to pass through the header( location:); redirect, or using PHP5.
Hope this helps someone! [i will also repeat this comment under the header() function]
To avoid the notice commited by PHP since 4.3.3 when you start a session twice, check session_id() first:
if (session_id() == "")
session_start();
Watch out for using UTF-8 encoding in your php scripts!
I don't know about other enviroments, but in Windows XP, if you edit a document and set it to be UTF-8, the editor (notepad for exapmle) inserts two invisible bytes at the beginning of the file (they read FF FE in hex for me). I suppose this happens so Windows can identify the file as UTF-8.
Since these two bytes are placed before anything else in the file, including <? ?> tags, when you execute the php script, they get outputed to the browser (even tough they won't be shown in the source of the result document, they're there!) before your php code gets a chance to run anything. This effectively cripples functions like session_start() ($_COOKIE to be exact).
The solution is to save the php file in a different encoding, or to manually remove them (I prefer the former).
I hope I helped somebody.
To avoid the notice commited by PHP since 4.3.3 when you start a session twice use this function:
<?php
function start_session(){
static $started=false;
if(!$started){
session_start();
$started = true;
}
}
?>
This should help you when your error reporting is set to E_ALL
Just a note for any of you who have the same trouble as I did. $_SESSION appears to break horribly if you attempt to assign information into integer keys.
For example, a result from a MySQL query may look like:
<?php
array(
0 => 'haravikk',
'username' => 'haravikk',
1 => 'fish',
'password' => 'fish'
);
?>
Attempting to assign that to $_SESSION either as a straight array, or by placing all keys (for example using a foreach) will FAIL and the data will be unavailable when you move to another page. My solution was to do this:
<?php
function storeSessionArray($array) {
foreach($array as $k => $v)
if (is_string($k)) $_SESSION[$k] = $v;
}
?>
This worked correctly =)
"SESSION FUNCTION IS DOMAIN SPECIFIC.
So, to savage the situation. be domain specific."
Umm. No, it isn't. It is however only domain specific when the domains are hosted on the same server.
You can still share session data between domain2.com and domain3.com just with that caveat.
As the previous note said, the session_start() function needs a directory to write to.
You can get Apache plus PHP running in a normal user account. Apache will then of course have to listen to an other port than 80 (for instance, 8080).
Be sure to do the following things:
- create a temporary directory PREFIX/tmp
- put php.ini in PREFIX/lib
- edit php.ini and set session.save_path to the directory you just created
Otherwise, your scripts will seem to 'hang' on session_start().
One thing of note that caused me three days of trouble:
It's important to note that Firefox (for one) makes two calls to the server automatically. One for the page, and one for favicon.ico.
If you are setting session variables (as I was) to certain values when a page exists, and other values when pages don't exist, the values for non-existent pages will overwrite the values for existing pages if favicon.ico doesn't exist.
I doubt many of you are doing this, but if you are, this is a consideration you need to address or you'll be bald over the course of a three day period!
Yes php does not automatically insert the SID in header redirects. You must use something like:
'<address>?SID='.SID
to manually insert the SID.
foltscane at yahoo dot com wrote about people losing session information on page redirects.
The problem is some times the redirect may kick you off to the next page before all the session variables have been saved. The true solution to lost session vars on redirect is to simply call session_write_close(); before setting the redirect header. This will insure that php finishes writing the session info before page redirect gets underway.
ie:
<?
session_start();
$_SESSION['forward'] = "This session data will not be lost!";
session_write_close();
header('Location: nextpage.php');
?>
If you are insane like me, and want to start a session from the cli so other scripts can access the same information.
I don't know how reliable this is. The most obvious use I can see is setting pids.
// temp.php
#!/usr/bin/php -q
<?php
session_id ("temp");
session_start();
if ($_SESSION) {
print_r ($_SESSION);
}
$_SESSION['test'] = "this is a test if sessions are usable inside scripts";
?>
// Temp 2
#!/usr/bin/php -q
<?php
session_id ("temp");
session_start();
print_r ($_SESSION);
?>
Here is a crazy feature -- you get one concurrent database connection per session, transparent to the programmer!
I am running php 5.0.4 on iis 6 (yeah, I know...). I noticed under stress testing that database connections were my bottleneck. I spent like 5 hours turning this on, that off, etc because I could not get php to create more than one database connection at a time, regardless of how many requests it was servicing. I tried running php as fastcgi, cgi, and isapi module. I tried using mysql and mysqli extensions. Same thing every time. Then I happened to turn off auto-start sessions and commented out my session_start() calls, and all of the sudden I get multiple connections like I expected!
Don't let this happen to you -- don't freak out! Apparently something somewhere tries to make sure your visitors don't get more than their share of your resources, so they get one db connection per session. I can't say it is what I would like to happen, but now that I know, it doesn't bother me much :) Hope this saves someone some time and headache!
For the problem of session lost after of redirect with header location...
Try with this:
<?
session_start();
$_SESSION['mySession'] = "hello";
header ("Location: xpage.php");
exit(); //This sentence do the magic
?>
If you open a popup window (please no commercial ones!) with javascript window.open it might happen IE blocks the session cookie.
A simple fix for that is opening the new window with the session ID in a GET value. Note I don't use SID for this, because it will not allways be available.
----page.php----
//you must have a session active here
window.open('popup.php?sid=<?php echo session_id(); ?>', '700x500', 'toolbar=no, status=no, scrollbars=yes, location=no, menubar=no, directories=no, width=700, height=500');
----popup.php----
<?php
session_id(strip_tags($_GET['sid']));
session_start();
//and go on with your session vars
?>
workaround when using session variables in a .php file referred by a frame (.html, or other file type) at a different server than the one serving the .php:
Under these conditions IE6 or later silently refuses the session cookie that is attempted to create (either implicitly or explicitly by invoquing session_start()).
As a consequence, your session variable will return an empty value.
According to MS kb, the workaround is to add a header that says your remote .php page will not abuse from the fact that permission has been granted.
Place this header on the .php file that will create/update the session variables you want:
<?php header('P3P: CP="CAO PSA OUR"'); ?>
Regards,
Hugo
Quick point, since this had been going round in circles for days...
IE will not accept sessions from a domain that has an non alpha-numeric character in it. My development site was running under the vhost mos_dev and it was killing me, trying to work out why IE kept dropping my sessions.
About the Session not getting saved on a header("Location:") redirect.
Make sure to call session_write_close() before doing the redirect. This will make sure that the session values get written to the disk.
[Editors Note: For more information about this
http://www.zvon.org/tmRFC/RFC882/Output/chapter5.html ]
I use name-based virtual hosting on Linux with Apache and PHP 4.3.2.
Every time when I refreshed (by pressing F5 in Internet Explorer) I noticed that I got a new session_id. Simultaneously browsing the same site with Netscape didn't give me that problem. First I thought this was some PHP issue (before I tested it with Netscape), but after searching a lot on the internet I found the problem.
Since I was using name based virtual hosting for my testserver and we have different webshops for different customers I used the syntax webshop_customername.servername.nl as the domain-name.
The _ in the domain name seemed to be the problem. Internet Explorer just denies setting the cookie on the client when there is a special character (like an _ ) in the domain name. For more information regarding this issue: http://support.microsoft.com/default.aspx?scid=kb;EN-US;316112
Stupidly enough, this information was related to asp (yuk :o)
SESSION LOST ON HEADER REDIRECT (CGI on IIS 5.0)
I realize there are numerous scattered posts on this issue, but I would like to add my 2 since it took me a whole day and a download of the LiveHTTPHeaders Mozilla plugin to figure it out.
On the **CGI** version of PHP on IIS 5.0/Windows 2000, the following code will not work as expected:
/***** sess1.php *****/
session_start();
$_SESSION["key1"] = "testvalue";
header("Location: sess2.php");
/***** sess2.php *****/
session_start();
echo "key1 = '".$_SESSION["key1"]."'";
PROBLEM:
All session data is lost after a header redirect from the first page on which the session is initialized. The problem is, the PHPSESSID cookie is not being sent to the browser (ANY browser, IE or Mozilla) on the initial session page with the header("Location: ...") redirect. This is unrelated to client cookie settings - the set-cookie: header just isn't sent.
SOLUTION:
I was able to remedy the problem by switching to the ISAPI DLL version. This seems to be an MS/IIS bug, NOT a PHP bug - go figure. I hope this saves you some headaches especially with your user authentication scripts!!
The closest matching "bug" report I found:
http://bugs.php.net/bug.php?id=14636
Just for info, session_start() blocks if another PHP using the same session is still running in background. It seems it's waiting the other PHP to finish... and sometimes it can be a problem. Create 2 different sessions by setting 2 different names : session_name() solve the problem.
Benja.
A note on Windows installations with an NTFS file structure:
Make sure you give your Internet user account (usually IUSR_MACHINENAME) "Modify" access to the session.save_path directory. Otherwise session_start() will silently - and oh, so slowly - fail.
("Modify" is a Win2k term, but the concept translates to older IIS installs).
Diese Seite bei php.net