[Spice-devel] [PATCH spice-gtk v2 0/4] Public include cleanups

[Jonathon Jongsma]

On Tue, 2015-10-06 at 23:12 +0200, Pavel Grunt wrote:
> On Tue, 2015-10-06 at 15:23 -0500, Jonathon Jongsma wrote:
> > On Fri, 2015-09-25 at 08:53 -0500, Jonathon Jongsma wrote:
> > > On Fri, 2015-09-25 at 10:51 +0200, Christophe Fergeau wrote:
> > > > On Thu, Sep 24, 2015 at 05:20:25PM -0500, Jonathon Jongsma wrote:
> > > > > This is a second approach suggested by Marc-Andre to fix up some of the
> > > > > issues
> > > > > with circular includes in the public headers. It will break the build of
> > > > > applications using spice-gtk, but is very easy to fix since we print out
> > > > > an
> > > > > explicit compiler error in that case. If accepted, I intend to push this
> > > > > after
> > > > > the current release comes out.
> > > > 
> > > > I'd make it only warn for a release or 2 before enforcing this with an
> > > > error so that applications get a chance to be updated.
> > > 
> > > 
> > > I thought about that. It would definitely be a gentler approach. But if
> > > we only make this an warning, builds will probably still break due to
> > > patch 4/4. I suppose we could delay patch 4 for a couple releases too. 
> > > 
> > > Actually, I could push the first 3 patches (with errors changed to
> > > warnings) before the upcoming release. Then I could push the 4th patch
> > > patch after the release is done. Would that give enough time, or would
> > > you prefer an additional release between?
> > > 
> > 
> > So, we still need a decision on this. I don't think it makes sense to
> > use #warning unless we drop patch #4 for a couple of releases. If we
> > keep patch 4, we're going to break builds for apps anyway, so we should
> > use #error to make it clear *why* the build is breaking. I think I'd
> > prefer to simply bite the bullet and push everything in now with an
> > #error. 
> > 
> > Jonathon
> > 
> I agree with you, it is easy to just ignore the #warning, so the #error seems
> better.
> 
> Pavel
> 
> 

I've pushed the first 3 commits (with #warnings instead of #errors) for
now. After having to bisect virt-viewer with these include changes, I'm
a more hesitant about pushing immediately with an #error. I'm going to
wait and think on it for a little while.

Jonathon