TL;DR: I will coordinate with the FSF admins to get this fixed. Nobody knew there were ACLs in use. I am sure they weren't copied.
Karl wrote: > Thérèse Godefroy wrote: > > `/webcvs/leg/leg/,index.html,': Permission denied Like Karl I look at that and it all looks "normal" to me. And except for what Karl says later I would have had no idea how this would have worked for you previously since you are not a member of the project group. > As I recall, there is some ACL or other magic allowing members of group > www (like you) to commit to other repositories? Unfortunately I'm not > able to find the information now. It is certainly plausible that > such stuff was lost in the migration (not intentionally/knowingly). Your memory is probably correct. Because I can see no other way for people not in the project group to be able to access those files. > Without something like that, I see no reason why th_g should have been > able to commit to any project repository previously. The web > directories are only writable to the group, as intended, and you are > only in a few groups: Agreed. If people in the www group were previously able to modify those files then it must have been some type of ACL. I haven't been aware of those previously. > Sure. The rationale isn't in question. The "implementation" on > Savannah of giving members of the www project access to other projects' > "webpage repositories" is the question. Whatever the implementation > was, it evidently broke with the hardware switch. I don't know the > answer and I'm afraid I cannot research it any time soon, and I don't > think Bob can either for a couple of days at least. Assaf, Michael? I know that Ruben made a copy of the file system from one place to another using rsync. There was a small problem with the uid of system uids that would not have affected non-system users. We were able to repair that problem fairy quickly and easily. With the information in these messages I think it likely that no one knew about the ACLs and therefore they were not copied at rsync time. It isn't something I would have thought of either. It is impossible for us to research this as it needs one of the FSF admins who have access to the file systems. There will have been changes made since the copy was made so copying them again isn't possible but we can add ACLs to the destination again using the archived sourced as a template. The FSF admins will have access to the original source. We can reference the original files again for fixing the ACLs. Among other things if ACLs are needed then we will need to enable them as mount options. Because the data disk is on the system disk this will require a reboot of the server so that they are enabled. I am out most of the later half of the day tomorrow (Thursday) but can get the process going with the FSF admins in the morning. I think it likely that if the issue is ACLs that we can get those restored tomorrow sometime and have things available for use again on Friday. I will file a ticket and coordinate with Ruben to have this done. I will post an update here tomorrow later in the day as to the status. Bob
