Hi, my message seems to got lost; re-posting:
Sébastien Villemot <sebast...@debian.org> writes: > On Wed, Oct 04, 2017 at 03:43:00PM +0200, Anton Gladky wrote: >> one more > option is to drop pandas on those archs (filing RM bug) >> and fill the list of supported archs explicitly in d/control. > > If the pandas binaries are known broken on those archs, then it is > the only acceptable solution (I guess the debate is then about how > broken they are). They are often not really totally broken, since packages (with their own tests!) can still be built on top of Pandas. So, while there are important bugs, I would not mark them as (RC) "makes the package unsuitable for release", but as (important) "major effect on the usability of a package, without rendering it completely unusable to everyone". Disabling the tests and lowering severity would IMO here the sensible solution. I would prefer to keep the failing archs unless we know that they are really useless. That helps for portability of the dependent packages: One of the "selling points" to why packagers benefit from Debian packages is that they get the packages tested on a variety of platforms. And, especially the ARM processor is quite important, at least for astronomy (Raspberry is used for telescope control and first data reduction by amateurs). Please, rethink the idea of removing Pandas on non-x86. A partly buggy package still works well with other packages. Cheers Ole
