The goal here is to reduce the amount of tick() polling that we do by default.
For things like pools, networks, and interfaces, the constant polling is
not very helpful and causes CPU churn and slowness for remote connections.
Switch to a more on demand style. Pages that want new information for
these objects now request a priority tick that only refreshes the info
we want.
This isn't perfect, but neither was the previous solution in the face of
things like XML updates behind our back. The real solution here is libvirt
event support across the board.
This base connection object will be used to simplify the API in various
places, reduce libvirt API calls, and better share code between virtinst
and virt-manager. For now it just centralizes connection opening.
This also exposed various places where our handling for older libvirt
was busted, so raise our minimum host version to 0.6.0, the first
version that supports threaded client requests.
Now that virt-manager requires very modern versions of gtk, it's
quite unlikely that we will even run on a system with an old
libguestfs version, so don't bother checking versions.
When a VM is undefined, whether explicitly through virt-manager or behind
our back, we close and cleanup any associated vmmDetails windows. However
there's a race here: if the details window is the last remaining top
level window, cleanup() is called twice which tracebacks.
Fix this by dispatching the exit_app routine in an idle callback, to
avoid racing with any window delete routines.
Reported-by: Leonardo Garcia <lagarcia@br.ibm.com>
And drop a whole bunch of legacy dbus fallback stuff. Now that we
are bumping the host dependencies to very recent gtk, none of the
back compat should matter.
Despite being a known quantity, autotools is so overkill for our needs,
so let's drop it and replace it with a much simpler and easy to customize
system.
Cole Robinson