On Tue, Oct 15, 2019 at 09:42:45AM -0300, Daniel Henrique Barboza wrote:
I was hoping to quickly re-send the qemu_driver cleanups I've
sent some time ago, now using Glib. I started by attempting to
change the first VIR_AUTOUNREF() call in qemu_driver.c to
g_autoptr(), which happens to be a virStorageSourcePtr type,
then I realized that it wasn't that simple.
It should be that simple with this commit:
commit 667ff797e8eb8d82f30ab430216a8d2eef6b915a
Author: Daniel P. Berrangé <berrange(a)redhat.com>
Date: Fri Oct 4 17:14:10 2019 +0100
src: add support for g_autoptr with virObject instances
we should be able to use g_autoptr for any virObject, without
having to lock-step convert to GObject.
What actual problem did you find ?
Following up the instructions found on commit 16121a88a7, I started
the conversion. Then 'make check' started to fail because some
calls to virObjectRef/virObjectUnref were still remaining
in the code, messing up stuff related with mirrorChain in
qemu_blockjob.c. Turns out it was easier to burn through all the
instances and change them to use GLib.
Yes, if you convert from virObject to GObject, you *must*
convert all virObjectRef/Unref calls at that time.
This is being sent as RFC because x-I am aware that
docs/hacking.html
mentions that we shouldn't mix up certain GLib macros with Libvirt
ones, thus I am uncertain of whether I have messed up or not.
'make check' works, did a few sanity checks with libvirtd as
well.
Yes, the need to not mix g_auto* with VIR_AUTO*, is why I did commit
667ff797e8eb8d82f30ab430216a8d2eef6b915a to let you use g_autoptr
with virObject, without first converting to GObject.
Regards,
Daniel
--
|:
https://berrange.com -o-
https://www.flickr.com/photos/dberrange :|
|:
https://libvirt.org -o-
https://fstop138.berrange.com :|
|:
https://entangle-photo.org -o-
https://www.instagram.com/dberrange :|