It can be that another thread already filled the zone into the cache,
thus deal with it properly, because that other zone can be used by that
other thread.

As any other operations, thus there won't be any issue with the main
contexts (an assertion runtime message from GSimpleAsyncResult).

Debarshi Ray authored Jan 05, 2015 and Milan Crha committed Jan 05, 2015

Printing the error message to stdout after we have redirected it to
the parent process causes the write system call to get stuck.

The sync request are done asynchronously, in a dedicated thread,
which means that the database file can be closed meanwhile. The
database has no reference counting, thus it's required to track
whether there are any pending sync requests, to not free the
structure too early.

The IDLE state wasn't OFF, because there was an ongoing wait for DONE.

The namespace override didn't work at all, thus made it work first,
then added two new options to ignore other users namespace and
shared folders namespace. All these options modify server-supplied
list of namespaces.

This used to save store's summary after each single folder removal,
which includes also an fsync() call. Just remove unknown folders and
save changes once significantly speeds-up the initial folder hierarchy
sync with many folder removals.