[Debian-med-packaging] [MoM] Another CamITK training issue [Was: Bug#701250: camitk: ftbfs with GCC-4.8]

[Andreas Tille]

Hi Emmanuel,

On Mon, Feb 25, 2013 at 08:29:52AM +0100, Emmanuel Promayon wrote:
> 
> I will do my best to solve bug #701250 upstream, which seems to be
> the best (the next CamiTK release is planned for the end of this
> week, so hopefully, that will be part of the update as well as the
> nonfree plugin).

Fine.
 
> In order to achieve the recommended:
> apt-get -t experimental install g++ g++-4.7 g++-4.8 libc6-dev
> inside the pbuilder environement, it seems that what is given in the
> wiki pbuilder tricks page [1] is the best thing to do.
> 
> My first question is: in order to force the use of g++ v4.8 should I
> then temporarily (locally, and without forgetting to erase it before
> committing!) add the following "build-depends" in the d/control
> file:
>  g++ (>=4.8)
> Or is there a better way?

I admit I never tried / had to try this - but I do not see any problem
in your approach and so I'd say you might go for it.  Otherwise it might
be a good idea to ask on debian-mentors list.

> I also saw a ftbfs of CamiTK on mipsel [2], which is a bit puzzling.
> It terminates with this error:
> c++: internal compiler error: Killed (program cc1plus)
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <file:///usr/share/doc/gcc-4.6/README.Bugs> for instructions.
> make[3]: *** [actions/image/itkfilters/CMakeFiles/action-itkfilters.dir/ConnectedComponents.cpp.o]
> Error 4
> 
> The only problem I can think of is a memory problem.
> ConnectedComponents is based on itk which extensively use C++
> templates. It therefore needs a lot of memory to compile. Is it
> possible the memory usage went too high for the mipsel machine?

We had a similar problem in other package (just do not remember which
one for the moment.)  I'd suggest to ask on the mips porters mailing
list to ask for hints.  (You can write to any list @lists.debian.org
without beeing subscribed - just mention that you prefer CC.)

> My second question: is there a way to investigate this error a
> little bit further?

If you do not get any help and else fails we need to try one of the
porter boxes (DDs have access.)  (I actually own a video recorder
running on MIPS where I planed to create a Debian chroot on - this
plan is not yet realised.)

> And last questions (for now!):
> There is two override disparities [3] for libqtpropertybrowser3 and
> libcamitk3-data and one that was filed as a bug for libcamitk3 and
> then solved [4].
> I am not sure I fully understand what is the mechanism/problem
> behind this override disparities.

I guess in this cases the override which is maintained by ftpmaster
is correct and the d/control file should be adapted.  (Sometimes it
is the other way around and ftpmaster will adjust.)

> Should I modify the d/control files to solve the two disparities?

The d/control file should simply make sense.  If you consider that the
override is correct and the package wrong you should fix the package.
Looking onto the libs I think it should be really changed to section
libs in the d/control file.  I'm not fully sure about
   libcamitk3-data
whether it fits better into doc or science.  If you have good reasons
to insist in doc you might ask ftpmaster to fix the override file.
Otherwise just change d/control accordingly.

> Should I add a different section: for the target package that have
> disparities? What should I do for [4]? (how can the bug be marked
> "solved" when it still says "science" in the svn control file?)

The status of svn is not relevant for the bug.  Only *uploaded* packages
can fix a bug.  Please fix d/control and close the bug in d/changelog.
The next upload will then close the bug.

> Thank you in advance for your light!

Hope this helps

      Andreas.
 
> [1] http://wiki.debian.org/PbuilderTricks#How_to_satisfy_experimental_dependencies
> [2] https://buildd.debian.org/status/fetch.php?pkg=camitk&arch=mipsel&ver=3.0.7-1&stamp=1361136371
> [3] http://packages.qa.debian.org/c/camitk.html
> [4] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=700644#11
> 
> _______________________________________________
> Debian-med-packaging mailing list
> Debian-med-packaging at lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/debian-med-packaging
> 

-- 
http://fam-tille.de