perl-embperl mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From ___cliff rayman___ <>
Subject Re: Session Problems Due to Apache Calling Back Into Itself
Date Wed, 21 Sep 2005 05:25:04 GMT
Gerald Richter wrote:

>>Ok - I've done that.  The SES: message always comes from the 
>>parent process so it is not always easy to match the SESSION 
>>data to the child process that is actually performing the request.
>The question is, what are you seeing in case of the problem you have. Could
>you quote the SES: messages from the log for this case?
I don't see any problem with the SES messages. They show the cookies 
coming in, being modified and being sent normally.  The actual data in 
the some of the cookies are sometimes inconsistent and incorrect, and 
they often have changing  IP addresses AND changing AGENT s(although 
this can and does change often even for valid sessions).  I do have an 
example of a single session being used by both a Windows XP box and a 
Max OS/X box.   Pretty cool trick.

The problem seems directly attributable to a 500 configuration error 
occuring with a request that was using the session.  These configuration 
errors were occuring because of an error in our code which was causing 
Storable to fail.  It seems like the next connection to a process that 
has sufferered a 500 error, inherits the session data that is currently 
in udat.  Since we overwrite a lot of the session data, the session now 
contains data from the original session and whatever we overwrite from 
the current request..

Of course, now we have two users of the same session, and the 
udat/session routines do not handle this multi-user activity well.

We have fixed some issues with our code that caused 500 errors, which is 
good in any case.  Hopefully you will also modify the Embperl code to 
emit a 404 error when a file is not found instead of a 500 error.  I am 
not sure if this can cause a similar situation.

Perhaps some of the Session exceptions are not clearing udat, which 
causes the next connection to that process to inherit the existing 
session data.  If you agree, I will send a note regarding this to the 
developer of Apache::Session.


Business Consulting and Turnaround Management
[eml] cliff=  at
[phn] 888-736-3802 x701
[fax] 818-743-7404

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message