See https://bifrost.kolabsystems.com/T408794 for context. I verified that the RC UI continues to function (especially activesync and folder config),
so it doesn't brake anything, and that the Ping works better for https://bifrost.kolabsystems.com/T408794

With this change METADATA for every folder will be asked separately. This definitely will have performance implications. Maybe a better approach would be to add a method to reset kolab_storage::$typedata (and other cache-like data if needed). And call that method from syncroton's kolab_sync::sleep() or kolab_sync_backend_folder::hasHierarchyChanges().

In D2143#24409, @machniak wrote:

With this change METADATA for every folder will be asked separately. This definitely will have performance implications. Maybe a better approach would be to add a method to reset kolab_storage::$typedata (and other cache-like data if needed). And call that method from syncroton's kolab_sync::sleep() or kolab_sync_backend_folder::hasHierarchyChanges().

How about we call self::folders_typedata($prefix); to populate the cache, but then user folder_type() anyways to fetch missing entries (it still uses the cache if available)?

How about we call self::folders_typedata($prefix); to populate the cache, but then user folder_type() anyways to fetch missing entries (it still uses the cache if available)?

I don't like it for performance reasons, because if we know the folder has no type then why would you bother with a METADATA request? This would create many redundant requests when listing mail folders. And I'm not sure it would solve your issue.

In D2143#24421, @machniak wrote:

How about we call self::folders_typedata($prefix); to populate the cache, but then user folder_type() anyways to fetch missing entries (it still uses the cache if available)?

I don't like it for performance reasons, because if we know the folder has no type then why would you bother with a METADATA request? This would create many redundant requests when listing mail folders. And I'm not sure it would solve your issue.

I suppose there are multiple parts to it:

• Line 873: Removing the shortcut for groupware folders has the effect that we list folders from the roundcube cache consistently, also for groupware folders, thus allowing new folders being discovered (at the cost of the folder listing)
• Line 898: Using folder_type instead of directly accessing $folderdata just has the effect that we first attempt to get the data from the cache, but then also fetch it if it's not existing (thus solving my problem).

IMO it would be better to make sure that the one cache in roundcube is good enough for performance and avoid extra caching layers that need to be invalidated separately, but if you prefer something non-intrusive we can also do something like: https://git.kolab.org/D2155

If you think that this internal cache is not really needed then go this way and remove use of self::$typedata. In rare cases when we list folders for all folder types separately in a single request, we'll end up with several requests to the Roundcube imap cache, instead of one. I'm afraid that we might call folder_typedata() or folder_type() methods quite a lot, but I didn't check that, so I'm not sure.