|
2012-01-25 19:29 AEST by Arthur Barrett - The :local: protocol is not recommended. Using a 'real' protocol, eg: pserver or sspi solves this.
In the rare case when the server is set to RunAsUser=cvsuser and the 'passwd' and or 'group' file is not
writable by cvsuser then you cannot use a 'real' protocol.
In this case the recommend workaround is to temporarily disable audit while adding users.
To resolve the file ownership issues, as a linux administrator to:
1. take a snapshot of the server vm (if the server is running on vm)
2. change the ownership on /path/to/repos/CVSROOT/group and path/to/repos/CVSROOT/passwd to
cvsuser:cvsuser
3. change the file access on /path/to/repos/CVSROOT/group and path/to/repos/CVSROOT/passwd to
to g+rwxs
4. run some tests, eg:
*regular user checkin/checkout/promote
*regular user browse using CVS Suite Studio
*cvsadmin user browse using CVS Suite Studio
* cvsadmin add a user from client PC using :pserver: protocol
We recommend you resolve this some time when the pressure is off and you've got plenty of time to run
some tests and you have a non-admin user who can also run some tests.
|
|
2012-03-03 06:36 AEST by Arthur Barrett - To clarify:
The cvsadmin currently follows this procedure using the linux server to add a new user 'mdelacru':
cvs -d :local:/db_conf/CVS_local_storage/Repos/Repo_1 passwd -r cvsuser -a mdelacru
The correct procedure is instead to use the client windows PC to add the new user:
cvs -d :sspi:njlcvs01:/estee passwd -r cvsuser -a mdelacru
The customer cannot use the 'correct' procedure, because the permissions on these files are wrong on
the server (should be cvsuser:cvsuser):
/db_conf/CVS_local_storage/Repos/Repo_1/CVSROOT/group
/db_conf/CVS_local_storage/Repos/Repo_1/CVSROOT/passwd
The procedure to change the permissions/ownership of the files has already been supplied to the
customer. However this requires several things which may take up a lot of time, including:
* conduct a security assessment on the impact of this change
* develop a test plan and rollback plan (in case of testing failure)
* perform system changes only possible from root
* test the change, rollback if required (using plan)
* update your written procedures
* re-validate/re-test the system and procedures
Because the customer was not ready to proceed with implementing the fix, we instead provided a
workaround. |