Re: Would an (l10n) NMU of atlas interfere with the gfortran trnasition?
On Mon, Mar 24, 2008 at 11:58:33PM +0100, Matthias Klose wrote:
> > No comment on the transition issues, but I hope you're fixing these l10n
> > issues by removing all of the idiotic debconf templates *completely* from
> > the package. No one needs a library package popping up a high-priority
> > debconf note to read the contents of /proc/cpuinfo to them, and my gorge
> > rises every time I see one of these debconf notes from the atlas family of
> > packages.
> I disagree somewhat ... please make the package fail to build in these
> cases. afaiu the blas libraries are optimized libs which can be
> replaced by refblas and lapack. Use these, and only include blas on
> architectures where these libs do work.
This isn't the package that has the gratuitous debconf notes informing users
that there were testsuite failures (which are also evil). This package
instead outputs gratuitous debconf notes if you install a package variant
which is optimized for a subarch on a system that doesn't have a compatible
CPU. Yuck.
As for the debconf notes about testsuite failures, as near as I can tell the
only remaining failures on x86 are bugs in either the testsuite, or in the
toolchain used to build the testsuite. Neither is a very good reason to
accost end users with this information, nor is it a very good reason to make
the package fail to build if the real problem is that the testsuite itself
is unreliable.
--
Steve Langasek Give me a lever long enough and a Free OS
Debian Developer to set it on, and I can move the world.
Ubuntu Developer http://www.debian.org/
slangasek@ubuntu.com vorlon@debian.org
Reply to: