I finally got the time to try using Pier. The main goal is not to
have a rock stable wiki (I would have used SmallWiki1 for that), but
to test Pier features. Don't get me wrong, I'm not complaining, just
trying to help in what is, I hope, a constructive way.
I installed Pier from scratch, to have the latest versions. (I
followed instructions given on http://smallwiki.unibe.ch/smallwiki/ pier/installationofpier/). As I needed the authentication features, I
installed the Pier-Security too.
The versions I used are:
At the moment, I have 2 kernels running. I noted a few problems :
- Pier-Security: I added users, groups without problem. Then I
disabled the View Command for "Other" users.
The problem is, disabling view doesn't hide contents. ie: "Other"
users can still view the first page of the wiki (even if the links
are grayed), the tree contents, and worst, if they type the address
directly, they can view all the pages of the wiki. (this problem was
already reported by Cédrick Béler, and I fixed it quick&dirty by
modifying PRContentsWidget and PRTreeWidget to only display content
when context isValid)
Another thing, even after fixing the first problem, when I click on
Login, the context becomes valid, and the tree contents are displayed.
- PRFilePersistency: I subclassed PRFilePersistency to have each
kernel logging in it's own log file.
- When PRFilePersistency is instanced, it doesn't actually create
the log file. So when using the Log or Changes commands before saving
anything, I get a FileDoesNotExistException.
- When I delete a page, the object remains in the log. When I use
Changes or Log Commands, Pier is confused because it tries to display
informations about deleted pages, thus resulting in an
UndefinedObject error. If I manually delete the log, the problem is
- mutex: I got a strange problem. Suddenly, one of the 2 kernels
stopped accepting logins. The page just froze when I tried to login.
I found that something went wrong with a process in this kernel's
mutex. I tried to play with the blocked process, sending resume and
terminate messages, but I probably shouldn't have, because the image
locked, and the VM segfaulted, so I couldn't investigate it more