[ros-users] Yet another fix for vcstools + git

Ken Conley kwc at willowgarage.com
Wed Oct 26 00:08:20 UTC 2011


Hi Peter,

I appreciate the description and patch -- it's best if you can just
attach these to the existing tickets, or open new tickets.  That way
you are setup to be notified by updates in the ticket, and it saves
the back-and-forth communication between e-mail lists and ticket
comments.

BTW: vcstools now has a new tracker now that it's been migrated to be
a standalone library, so if you want to file a new ticket (instead of
reopen 3251), the tracker is linked from the wiki page:
http://www.ros.org/wiki/vcstools

thanks,
Ken

On Tue, Oct 25, 2011 at 3:56 PM, Peter Soetens <peter at thesourceworks.com> wrote:
> Ticket # 3251 attempted to fix a git submodules bug, but the proposed
> patch was a hack which assumed submodules wouldn't change across
> branches. In our case, they do and the patch breaks our build on
> oneiric of the orocos_toolchain_ros package. This could easily break
> other stacks as well as more git stacks with submodules are created.
>
> The correct fix is to delay the checkout and then update the submodules.
> Also, there was a bug in the current version that if version was empty,
> the checkout would not necessarily use HEAD.
>
> I fixed all these issues and tested on a rosinstall file containing:
>
> - git: {local-name: ros-git-test, uri:
> 'http://git.mech.kuleuven.be/robotics/orocos_toolchain_ros.git' }
>
> as above, with a version: master and with a version: oneiric item
> The oneiric version is without submodules, the master is with submodules.
>
> I can't retest the case of ticket #3251 since that repos no longer exists,
> but I think I got all cases covered.
>
> Please consider applying since this bug breaks our oneiric build:
> http://build.willowgarage.com/job/prerelease_electric_orocos_toolchain_ros_oneiric_amd64/2/console
>
> The alternative for us is to define orocos-toolchain-ros as a variant
> in the rosdistro file, but I could not find documentation on how to
> test or release a variant.
>
> Peter
>
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
>



More information about the ros-users mailing list