[Ros-release] Build failures for camera1394 on Groovy

[Tully Foote]

The issue is that we only have 4 machines in the old style build farm, so
if you kick off 6 copies of a prerelease which runs for 1.5 hours that
leads to nothing else happening for at least 1.5 hours.  And if a few
people do this, suddenly the whole day is gone.

Tully


On Fri, Mar 1, 2013 at 1:41 PM, Jack O'Quin <jack.oquin at gmail.com> wrote:

> Makes sense for devel builds, maybe not so much for pre-release tests.
> On Mar 1, 2013 12:33 PM, "Tully Foote" <tfoote at willowgarage.com> wrote:
>
>> Yes, we're only running one arch at the moment to keep devel builds from
>> bogging down the buildfarm.  We rarely saw devel builds failing on one arch
>> but not another, and we also wanted to keep the email volume down.
>>
>> Tully
>>
>>
>> On Fri, Mar 1, 2013 at 9:48 AM, Jack O'Quin <jack.oquin at gmail.com> wrote:
>>
>>> Is Hudson only running Precise, amd64 pre-release tests these days?
>>>
>>> I have not gotten any other responses, and don't see anything in the
>>> queue.
>>>
>>> On Fri, Mar 1, 2013 at 10:57 AM, Tully Foote <tfoote at willowgarage.com>
>>> wrote:
>>> > Ahh that makes sense.  Sorry I looked at master which was catkinized.
>>>  Yes,
>>> > that error will occur if you are setting a package dependency outside
>>> of the
>>> > declared stack dependencies.  rostest is in the ros_comm stack not the
>>> ros
>>> > stack.  If you depend on packages in the ros "stack/metapackage" you
>>> need to
>>> > have it in your stack.xml, with the packages in your manifest.  The ros
>>> > dependency brings in most of the build system so it's basically
>>> required,
>>> > though you can improperly depend on dependencies like ros_comm to
>>> bring it
>>> > in for you.
>>> >
>>> > Tully
>>> >
>>> >
>>> > On Fri, Mar 1, 2013 at 8:34 AM, Jack O'Quin <jack.oquin at gmail.com>
>>> wrote:
>>> >>
>>> >> On Fri, Mar 1, 2013 at 8:40 AM, Jack O'Quin <jack.oquin at gmail.com>
>>> wrote:
>>> >> > On Fri, Mar 1, 2013 at 3:39 AM, Tully Foote <
>>> tfoote at willowgarage.com>
>>> >> > wrote:
>>> >> >>
>>> >> >> I'm not sure what's going on here.  In groovy you will get that
>>> error
>>> >> >> if a dependency which is supposed to be a ros package doesn't
>>> exist on your
>>> >> >> path.  The tools will then assume it must be a rosdep if it's not
>>> an
>>> >> >> existing package.  I'll look into it tomorrow. I think it might be
>>> that the
>>> >> >> prerelease is installing the union of build and run depends,
>>> skipping the
>>> >> >> test_depends.  If that's true, then rostest won't be installed, if
>>> that's
>>> >> >> how it should be will take a trip to the white board with some
>>> sketching to
>>> >> >> answer.
>>> >> >
>>> >> > That groovy-devel branch of camera1394 on github now has a <depend
>>> >> > package="rostest" />. I would not expect that to be interpreted as a
>>> >> > test_depends.
>>> >> >
>>> >> >
>>> >> >
>>> https://github.com/ros-drivers/camera1394/blob/groovy-devel/manifest.xml
>>> >> >
>>> >> > Is there some <depend> other than "ros" needed in the stack.xml for
>>> >> > rostest to be accessible?
>>> >>
>>> >> Looks like the "ros" stack dependency no longer works. I added <depend
>>> >> stack="ros_comm" />, and the pre-release tests seem to work now. That
>>> >> must be why rostest was not getting installed.
>>> >>
>>> >> If the tests all succeed, I will release camera1394-1.8.1 to Groovy
>>> soon.
>>> >>
>>> >> Is the stack dependency on "ros" even needed these days? Without
>>> >> rosstack, it's hard to figure out the correct dependencies for dry
>>> >> stacks.
>>> >> --
>>> >>  joq
>>> >
>>> >
>>> >
>>> >
>>> > --
>>> > Tully Foote
>>> > tfoote at willowgarage.com
>>> > (650) 475-2827
>>>
>>>
>>>
>>> --
>>>  joq
>>>
>>
>>
>>
>> --
>> Tully Foote
>> tfoote at willowgarage.com
>> (650) 475-2827
>>
>


-- 
Tully Foote
tfoote at willowgarage.com
(650) 475-2827
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-release/attachments/20130301/0a5ab72f/attachment-0009.html>