bug

[phpwiki]
VMWare Server has been acting up lately. The network connection to guests gets dropped seemingly randomly. The problem is noticed as dropped network connections are normarlly noticed: web pages are unavailable, logins fail (eg. on XP a message about AD not accepting connections or not recognizing you), and if the guest runs a service, the service is unavailable.

The problem can sometimes be fixed (or even avoided) by pinging from the guest to some external address. I've also had luck in disconnecting and reconnecting the network device for the guest.

The host is always available and the guests can be controlled via the VMware remote interface. Our VMware Server boxes use bridged networking, so that the host and guests are all on

The administrator subdirectory of mambo distribution contains 3 files named index?.php . All files required this change. administrator/includes/auth.php



inside each file, immediately following the line:
session_name( 'mosadmin' );

insert:
ini_set('session.save_handler', 'files');



must be prior to line:
session_start();



otherwise there are "failure to initiate properly" issues.

BACKGROUND INFO:

Directive session.save_handler defines session management handler. php.ini default value = 'user', which if I've read

[phpwiki]
We need to be cautious about following globus/vdt instructions to a tee.
One of the things that seems to be happening is that the PATH is getting
redifined to put all the globus software first. While this is ok for end
user accounts, it's not ok for root. We need to make sure that root does
not have it's path defined to include globus tools. Case in point, if
globus is in the path and you restart sshd you'll get the globus sshd.
This is not desirable. This path redefinition can also be problematic
when building software (even as an end user).

I did some researches for source code, such as PresenceAction.java etc and found the code has some problems. Please see the code:
-------------------------------------------------------------
...
// get the current presence list (User objects) for this page
usageSession [] sessions = service.getPresence(location);
if (sessions != null) context.put(CONTEXT_USERS, sessions);
...
---------------------------------------------------------------
It gets the "sessions" based on "location". The location is info about a specific user, for example, "/user/cyy", which seems that it tries to show the location of /lhome/portal/portal/jakarta-tomcat-4.1.18/webapps/nmi/WEB-INF/psml/user/cyy/html/default.psml. So it would not show other users on the list at all. I tried to login as "cyy" in Mozilla and in IE. Both browsers showed two "cyy" users. If I logged out and used another user name "ogce" in IE, one "cyy" disappeared in Mozilla and only the "ogce" displayed on the user list in IE. That is why the user only can see his/her own on the list.

Pravin said that the text file is somewhere under the tomcat. I did search for all of subfolders and still didn't find that. I saw one xml file, /lhome/portal/portal/jakarta-tomcat-4.1.18/webapps/nmi/db/user.xml,
which looks like it. It has user id and kind of password, but it doesn't have all of users and isn't updated when adding a new user while the new user folder I mentioned in my previous email is created.

The java class file, edu.indiana.commgrids.newswizard.NewsBean, may help find how/where to store username and password. I only saw the binary file and don't know where the source file is.

I looked at OGCE and tried to understand how the small square box works. The small box is an html file which is an iframe from the browser. It looks that PresenceAction.java and presence.vm control the information. In the presence.vm file, it would show the user names by session.User.DisplayName. The code loops based on "sessions", which info is from PresenceAction.java. As I understand each HTTP session represents one user session. I added a peice of code in the java file to get the "sessions" number in log. I got session number = 1 for myself. The number still was one after Jason logged in. I guess code might have some problems.

there was a stray = in the conditional check of user_access() that was
returning very inconsistent results for the role permissions query or
skipping it entirely. taking this out has fixed all the posting issues.
they question is: how did it sneak in?

it seems that when mailhandler rejects a post it just drops the message in
the bit bucket. it should at least have an option to put the content in
the approval queue for manual accept/reject.

and example of the rejection comes if you post a blog with some command
headers but don't include the type: blog header. while this should be
optional it's ommission causes the blog to be rejected with permission
denied error.