[OE-core] [PATCH 2/2] scripts: Add yocto-compat-layer-wrapper

[Joshua Lock]

On Thu, 2017-03-30 at 10:43 -0600, Aníbal Limón wrote:
> 
> On 03/30/2017 10:13 AM, Patrick Ohly wrote:
> > On Thu, 2017-03-30 at 10:03 -0600, Aníbal Limón wrote:
> > > 
> > > On 03/30/2017 12:02 AM, Patrick Ohly wrote:
> > > > On Wed, 2017-03-29 at 15:44 -0600, Aníbal Limón wrote:
> > > > ...
> > > > > +show_help() {
> > > > > +	printf "Usage: %s [-o output_log] [-h] LAYER_DIR
> > > > > ...\n" $0
> > > > > +}
> > > > > +
> > > > 
> > > > ...
> > > > > +env_dir=$(mktemp -d -t yocto-compat-XXXX)
> > > > > +echo "The environment will be setup at $env_dir"
> > > > > +echo ""
> > > > 
> > > > The directory gets created, but not removed.
> > > 
> > > I didn't remove the temp directory because may be the user wants
> > > to
> > > access the dir after the check.
> > 
> > I think that this should be something that the user explicitly
> > needs to
> > request. As it is now, accumulating directories in /tmp when
> > invoking
> > the script is just unexpected and not the normal behavior of tools
> > creating something in /tmp.
> 
> Ok, i will add an option to don't delete and change the default
> behavior
> to delete after check.
> 
> 
> > 
> > > > > +echo "Cloning oe-core..."
> > > > > +git clone $oe_core_repo $env_dir
> > > > > +if [ $? -ne 0 ]; then
> > > > > +	echo "Failed to clone oe-core repository"
> > > > > +	exit 1
> > > > > +fi
> > > > > +
> > > > > +echo "Cloning bitbake..."
> > > > > +git clone $bitbake_repo $env_dir/bitbake
> > > > > +if [ $? -ne 0 ]; then
> > > > > +	echo "Failed to clone bitbake repository"
> > > > > +	exit 1
> > > > > +fi
> > > > 
> > > > Cloning bitbake and OE-core each time the script runs will be
> > > > fairly
> > > > slow. There's also a chicken-and-egg problem: if you don't have
> > > > bitbake,
> > > > where's the script?
> > > > 
> > > > I'd prefer to use an existing checkout of both, just as for the
> > > > layers
> > > > which are to be tested.
> > > 
> > > I choose to clone the oe-core/bitbake to ensure there are a clean
> > > environment, without any previous layer added.
> > 
> > I don't quite get that argument. When using existing bitbake and
> > OE-core
> > directories, how can they have layers added to them? I understand
> > that
> > the existing checkout might contain additional modifications and/or
> > might not match current master, but I consider that a feature. As
> > it
> > stands now, the caller of the script cannot test against Yocto 2.3
> > once
> > it is released, because the script will always check out master.
> > 
> > Control over that can of course also be added to the script, but I
> > just
> > don't see the need for all this additional complexity. Just my 2
> > cents.
> 
> That's the original issue to create this script, if a user tries to
> validate a layer and has modifications in configuration (bblayers,
> local, auto), the check will be contaminated in some how.

Those are all build-directory state, not items which are tracked in the
git repository.

Rather than cloning repositories to ensure clean state your script
could ensure a unique/new build directory by sourcing oe-init-build-env 
with a non-standard build directory, perhaps checking whether the
directory exists first and adding random characters until you get
something new?

Joshua