PyViennaCL next upload

[Toby St Clere Smithe]

Hi Anton, Andreas,

Anton Gladky <gladk at debian.org> writes:
> as far as I know Andreas Tille proposed a nice and really necessary tool
> to exclude and repack source tarballs [1], [2] (CC-ing him). But, it seems,
> it is not yet in uscan (sorry, I may be wrong).

I have updated the source package as Thorsten suggested, having also
made a new bug-fix release upstream. Thanks to Andreas' work, it was
much easier than I had anticipated, needing only minor changes to
debian/copyright, debian/watch and debian/rules (the latter just to turn
off a warning produced by a build system check for the excluded files).

I apologise for the time wasted by my not realising this sooner!

Could you sponsor the new package? It is available on mentors [1], and
in git[2].

[1] http://mentors.debian.net/debian/pool/main/p/pyviennacl/pyviennacl_1.0.2+dfsg-1.dsc
[2] http://anonscm.debian.org/gitweb/?p=debian-science/packages/pyviennacl.git;a=shortlog;h=refs/tags/debian/1.0.2%2Bdfsg-1

> I would also strongly recommend you to implement autotests using DEP-8 [3].
> An example for python packages you can find here [4], [5]. Then your package
> will regularly be tested.

I agree that this would be nice. But, as things stand today, I don't
think it can work: PyViennaCL currently depends on OpenCL, and so any
tests would not run unless an OpenCL implementation was available, which
is (I think) not the case on the buildds. The only plausible free
implementations right now are Mesa, which is new and unstable (as well
as only really supported on r600g hardware), and beignet, which has
similar problems, and only supports Intel HD4000+ hardware (I believe).

Either when the OpenCL situation settles down or when I modify
PyViennaCL to support CPU-only computation, I will definitely implement
autopkgtests. Of course, in the CPU-only case, then the OpenCL code
paths (the raison-d'être of PyViennaCL!) will not be tested. In any
case, many thanks for the useful links.


Best regards,

-- 
Toby St Clere Smithe
http://tsmithe.net