On 10/09/2013 03:32 AM, Daniel P. Berrange wrote:
On Tue, Oct 08, 2013 at 08:45:23PM -0600, Eric Blake wrote:
> Maintenance branches cherry-pick some, but not all patches, and
> sometimes in different order, which means that 'make syntax-check'
> is likely to fail for hard-to-predict reasons. Yet having a
> common workflow makes it easier to switch between branches. This
> patch sets up a filter so that 'make syntax-check' is a no-op
> other than to print a warning if it detects that the user is
> running in a git checkout that branches from some place earlier
> than origin; 'make -k syntax-check force-syntax-check=1' can be
> used to override the heuristics.
>
> Tested on master (no change in behavior) and v1.1.2-maint (where
> the syntax check was indeed suppressed). Based on a report by
> Cole Robinson.
>
> * cfg.mk (syntax-check): Add some conditional filtering.
>
> Signed-off-by: Eric Blake <eblake(a)redhat.com>
> ---
>
> If approved, I'll backport this to all the v*-maint branches.
I'm not sure I really like this. Quite a few of the checks we do
are quite critical to code quality and IMHO should continue to
be validated on maint branches to ensure that we don't screw up
when resolving conflicts.
I would be ok with skipping checks which are merely style related,
but not skipping checks which are code correctness issues. Obviously
skipping the copyright date check is reasonable.
But the point is right now, you CAN'T successfully run 'make
syntax-check' on a maint branch; you already HAVE to change workflow to
test branches differently than master. This patch doesn't change the
fact that a different workflow is needed to test a maint branch, but at
least makes it so that using the same workflow as on master will alert
you to what steps to actually take, rather than the current situation of
flat out failing without telling you how to work around the failure).
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org