[Spice-devel] [RFC spice-common] ci: Automatically test integration with spice-common users

[Pavel Grunt]

Thanks for the comments

On Fri, 2016-12-09 at 09:46 +0100, Christophe Fergeau wrote:
> On Thu, Dec 08, 2016 at 01:44:14PM -0500, Frediano Ziglio wrote:
> > > +client:
> > > +  stage: test
> > > +  script:
> > > +  - git clone $GIT_DIR_URL/spice-gtk
> > > +  - cd spice-gtk
> > > +  - git submodule update --init --recursive
> > > +  - cd spice-common
> > > +  - git checkout $LATEST_HASH
> > > +  - cd ..
> > > +  - git commit -a -m "Update submodule"
> > > +  - ./autogen.sh
> > > +  - make
> > > +
> > > +server:
> > > +  stage: test
> > > +  script:
> > > +  - git clone $GIT_DIR_URL/spice
> > > +  - cd spice
> > > +  - git submodule update --init --recursive
> > > +  - cd spice-common
> > > +  - git checkout $LATEST_HASH
> > > +  - cd ..
> > > +  - git commit -a -m "Update submodule"
> > > +  - ./autogen.sh
> > > +  - make
> > 
> > This looks like an attempt to have a more stable ABI/API or push
> > developers to update spice-server/spice-gtk if ABI/API breaks 

And you've done that ;)

> > which
> > is opposite of submodule idea.
I am not sure if it is the idea behind the spice-common submodule :) I
think it is about shipping common stuff.


> > Probably is up to how many people looks at CI but usually the idea
> > is that should break only on regressions (or system/network build
> > failures).
> > From the link you sent seems that the build is broken for spice-
> > common
> > while is from the user prospective completely alright.
> > Not again the patch, just a consideration. 
> 
> My reaction is a bit similar. It's useful to know whether spice-
> gtk.git
> / spice-server.git are working together with spice-common.git,
> however
> I'd see this as purely informative.

me too. It is not ideal when you need to fix a few surprises on a
submodule update.

>  As this is a submodule, it can be
> intentional/known that one of the module does not build (yet) with
> spice-common master. But if it's not known, it's good to be informed
> of
> this ;)
> 

Pavel