Hello,
I've been working on a CVS managed project on a lan (with a couple of other developers). We are behind a firewall.
It does not seem very reliable. CVSROOT is on a NFS share, and I can see the ",v" files are mode 666, and the directories are mode 777. This does not seem right. My predecessor did not create a developer group for access, as is mentioned here:
http://www.network-theory.co.uk/docs/cvsmanual/cvs_14.html
We often have to log into the server and fix permissions when a directory or file is added. Obviously something is brain-dead.
So I am seeking opinions on how to fix things up:
1) Is it okay to export CVSROOT over NFS? I sometimes (every several weeks) see the mount fail and the logs on the server say something about "rpc.mountd() : fhxxx() failed" (can't remember the exact message) and then I restart NFS and all is well again for several weeks.
2) I am planning on creating a "developer group" with all of our unames in it, and adjust the repo file/directory ownerships and modes to suit. Does the link above reflect a real-world scenario, or would you do something different? There will be no outside access to this respository (firewalled). The link says files should be read-only and directories should be r/w for the group...do they mean 444 for files and 775 for directories, with everything owned by agroupmember.group?
3) In the link above, the paragraph "(On some systems, you also need to set the set-group-ID-on-execution bit on the repository directories (see chmod(1)) so that newly-created files and directories get the group-ID of the parent directory rather than that of the current process.)" Does this apply on fedora?
The server and clients are all Fedora Core (FC1, server), (FC2, clients). They are all fairly up to date.
Thanks for any tips or suggestions. It's all properly backed up so I am not too worried about playing with things. The admin who does the backups and restores makes the developers manage their own trees and I have inherited this one.
--- John
