Emmanuel Blot wrote: > On one project, the wiki page index sometimes shows up a newly created > page, sometimes just ignores it.
I assume the page is created by normal means, i.e. edition in the wiki section? > The funny thing is that while the page did not show up in the page > index, I've still been able to access it, typing its exact URL in the > browser URL field. That's expected. We cache the list of available pages, which is not used for checking if a page to be rendered exists or not. > A colleague of mine just reported me that "one time > out of 10" a link from the WikiStart page pointing to this page > appears as a valid link, "9 time out of 10" it appears as a broken > link. It looks like some of the processes/threads know about the new > page, some do not. That seems to indicate that the invalidation mechanism through the DB did not work on that specific update, and only the process that was used to create the new page got invalidated. > I don't think if it's related or not, the page name does not follow > CamelCase name, and ends up with a '+' character. The page name should not matter, it's only contained in the cached *data*, not the key. Do you see any tracebacks in you log, possibly related to the creation of the new page? It's kind of hard to debug, as the issue is transient and resolves itself when you restart the web server. I'll add some logging statements to the cache logic, hopefully this will show what's happening in future occurrences. Would you mind opening a new ticket for this issue? Please indicate the DB backend you use, maybe the issue is DB-specific. -- Remy PS: Do you run multiple environments on your installation? I have the uneasy feeling that my usage of thread-local storage might be problematic in that case... A kind of cross-environment pollution, if you will. To be investigated...
signature.asc
Description: OpenPGP digital signature
