Re: RFS: viennacl
On 07/30/2011 01:28 PM, Michael Tautschnig wrote:
> Hi Michael,
> Thanks a lot for the quick response!
>>> - auxiliary/converter is shipped as binary, although converter.cpp probably is
>>> its source!?
>> Yes. What do you propose? Should I remove auxiliary/converter from the
>> source package? If so, how? Via a patch in debian/patches?
>>> - all of viennacl/ could probably be generated at build time, using the above
>> Would you prefer it if it was done that way? I can give it a try.
>>> - doxygen could easily be run at build time.
>> Yes, but the output is already contained in the original tar-ball. What
>> should I do with it?
> I think for all of the above you should speak to upstream about having them
> removed. I don't quite know about their responsiveness (or willingness to do
> so), hence for the moment you might want to start out with a repacked tar ball,
> adding +dfsg to the package version (note: you will also need to change the
> watch file in this case: add a line opts=dversionmangle=s/\+dfsg//), and ideally
> a get-orig-source target in debian/rules. Thereby you'd end up with a much much
> cleaner source package, which really is the *source*, and not some intermediate
> state. It would be nice if upstream would follow that reasoning...
> Please let me know if you need any further information on this!
I will contact upstream about this. I fear they will be reluctant to
remove the auxiliary/convert binary, because that one links against a
boost library, and they might be reluctant to require their users to
install boost as a pure build-dependency.
I'll give it a try nonetheless.