[libvirt] [PATCH 06/11] util: use glib base64 encoding/decoding APIs

[Peter Krempa]

On Mon, Sep 30, 2019 at 13:19:41 +0100, Daniel Berrange wrote:
> On Mon, Sep 30, 2019 at 02:03:53PM +0200, Peter Krempa wrote:
> > On Mon, Sep 30, 2019 at 13:56:06 +0200, Pavel Hrdina wrote:
> > > On Mon, Sep 30, 2019 at 12:45:56PM +0100, Daniel P. Berrangé wrote:
> > > > On Mon, Sep 30, 2019 at 01:41:52PM +0200, Pavel Hrdina wrote:
> > > > > On Mon, Sep 30, 2019 at 09:02:36AM +0200, Peter Krempa wrote:
> > > > > > On Fri, Sep 27, 2019 at 18:17:28 +0100, Daniel Berrange wrote:
> > > > > > > Replace use of the gnulib base64 module with glib's own base64 API family.
> > > > > > > 
> > > > > > > Signed-off-by: Daniel P. Berrangé <berrange at redhat.com>
> > > > > > > ---
> > 
> > [..]
> > 
> > > > > 
> > > > > Here I agree with Peter, for this series I would use VIR_FREE() where
> > > > > it's possible and only for glib objects we can use g_autoptr.
> > > > > 
> > > > > But eventually I would like to switch to g_autofree and friends in order
> > > > > to eliminate our specific helpers in favor of glib helpers.
> > > > > 
> > > > > This also brings a question if we should keep our wrappers for glib or
> > > > > use it directly.  For example the string functions that we have.
> > > > 
> > > > Where any libvirt code just duplicates something that alrady exists, then
> > > > I think there's no compelling reason to keep it, the best code is code
> > > > that doesn't exist.
> > > > 
> > > > I don't want todo too many big bang replacements though, so I think best
> > > > to deprecate existing libvirt code and phase it out incrementally in many
> > > > cases.
> > 
> > I agree in case of the other infrastructure for automatic pointers as
> > that will require more changes.
> > 
> > In this case I don't see why we shouldn't just replace all use of
> > VIR_AUTOFREE with g_autofree if the idea is to use g_autofree from now
> > on.
> 
> Well that's 1500 uses, across 150 files, so quite a big bang conversion.
> It would need to be split up quite alot otherwise it will be a backport
> conflict magnet. Certainly we want to clean this at some point, its just
> a question of timing.
> 
> My preference is to focus on things with functional benefit as the higher
> priority.

Then please stick with VIR_AUTOFREE for any code you plan to introduce.