[ros-users] hydro on saucy vs. raring
iluetkeb at gmail.com
Wed Jan 29 10:47:13 UTC 2014
wow, I didn't realize it was that much. Well, I will try compiling
from source. btw, the docs said something about unspecified
"optimizations" being done to the packages. Is there information
available on what that would be?
btw, while I have a better picture of the effort involved now, this
seems to mean that ROS will increasingly spend effort on Ubuntu
versions which are no longer supported by Canonical, particularly if
the ROS release cycle is lengthened as planned. Non-LTS releases are
now supported for only 9 months after release.
Maybe it would be an idea to have a "core" set of packages, which are
likely much less than 800, which could be made available on a larger
set of distribution versions. Something like the base messages,
middleware, and the like. Let me know if I could help with that.
On Wed, Jan 29, 2014 at 10:41 AM, Tully Foote <tfoote at osrfoundation.org> wrote:
> Hi Ingo,
> Although in the abstract it does seem like a relatively small upgrade. It
> would actually be a lot of work for a lot of people throughout the
> community. We've explicitly picked the distribution cycles and target
> platforms to reach a balance between coverage and the effort required. We've
> also received strong feedback from the community that they would like fewer
> slower releases. This is both because it's more work as a maintainer, as
> well as wanting more stability as a user just to keep up with upstream
> changes. There was a lot of discussion last spring on this such as this
> in the past we've added new Ubuntu distros in the middle of a ROS Distro
> cycle and based on that experience we plan never to do it again. Doing this
> sort of release into a new Ubuntu Distro is effectively creating another ROS
> Distro. As all the 3rdparty dependencies must again be reviewed and all
> source code patched for potential variations, such as the boost upgrade.
> This requires a coordinated effort to rerelease every package in the
> ecosystem. (~800 packages by ~ 200 maintainers) For reference this usually
> takes a few months to get to coverage in a normal release cycle. So doing it
> again mid cycle is very hard to justify.
> As you mention generally recompiling from source is fairly easy. I believe
> that we have coverage on the rosdep rules for 13.10, thus the toolchain for
> compiling from source will work fine. The first users of packages on the new
> platform will run into issues like the Boost version has changed, and
> maintainers are usually receptive to patches being submitted back. We
> relatively regularly test building all of desktop-full on i7 machines where
> it builds in a few hours from scratch. As an optimization you can pick the
> subset of the packages that you need saving both build time and the number
> of packages which need to be verified.
> At OSRF we have already been working hard on getting Indigo ready for many
> weeks(working on things like getting python3 compatiblity as well as making
> sure that upstream packages are released and available) , and if we were to
> spend the next several weeks coordinating a new Hydro release we would have
> to delay the Indigo release.
> So with the above reasons, our recommendation for using ROS on 13.10 is to
> compile Hydro from source patching where necessary. Or wait until Indigo
> Igloo is available in binary format.
> On Wed, Jan 29, 2014 at 12:45 AM, Ingo Lütkebohle <iluetkeb at gmail.com>
>> given that support for Ubuntu 13.04 is just ending, would it be an
>> option to move ROS's support target for Hydro to 13.10 instead?
>> At the moment, the packages for Hydro do not even install on 13.10,
>> largely because of the boost upgrade it seems, which should be a
>> fairly trivial issue when recompiling.
>> Ingo Lütkebohle, Dr.-Ing.
>> Machine Learning and Robotics Lab, IPVS, Universität Stuttgart
>> PGP Fingerprint 3187 4DEC 47E6 1B1E 6F4F 57D4 CD90 C164 34AD CE5B
>> ros-users mailing list
>> ros-users at lists.ros.org
> ros-users mailing list
> ros-users at lists.ros.org
Ingo Lütkebohle, Dr.-Ing.
Machine Learning and Robotics Lab, IPVS, Universität Stuttgart
PGP Fingerprint 3187 4DEC 47E6 1B1E 6F4F 57D4 CD90 C164 34AD CE5B
More information about the ros-users