On Tue, Jun 5, 2012 at 9:29 AM, Weißhardt, Florian
<Florian.Weisshardt@ipa.fraunhofer.de> wrote:
> Hi Jack and Tully,
>
> just to add to that discussion: we figured out the same thing while
> releasing the same code in both electric and fuerte using the same revision
> number. A workaround for us was to always increase the version number and
> triggering new releases for electric and fuerte directly after each other.
> Because a later update for just one ROS release didn't work.

Thanks for that information, Florian.

Just to confuse matters further, yesterday I released a new
art_vehicle stack using the latest rosrelease script to Fuerte only,
and it exhibited the same failure.

Today, I released a new version of the velodyne stack to both Electric
and Fuerte. So far, that one looks OK. At least the
velodyne-0.9.1.yaml file seems fine.

My guess is that the rosrelease bug Tully fixed was valid and
necessary, but others remain. Or, maybe the art_vehicle stack really
is broken in some manner I have not yet been able to determine.

We need better verification tools for stack maintainers and release
managers. Releasing a package that already passes unit and pre-release
tests should ideally be a purely mechanical process.

> I think the use case of having the same code released for two (or even more)
> ROS releases under the same version number is not so rare.

For people maintaining stacks outside Willow Garage, I suspect it is
quite common.

Our University installs the current ROS distro on machines in the CS
labs at the beginning of each semester. Their admins are
understandably reluctant to install a new version during classes. Even
if they did, we prefer not to put our students through a migration
while they are busy working on their class projects. Given the ROS
distro schedules for the last four ROS distributions, we have *always*
needed to support our code on both the latest and previous versions.
Forking a separate branch for each distro is *not* an attractive
option. Thanks to tick-tock, we have so far been successful in
supporting two successive ROS distros with a single source tree
(sometimes with a few #ifdef's).

I feel sure that many others have similar needs.
-- 
 joq