On Mon, May 01, 2017 at 01:12:19PM +0200, Julien Cristau wrote:
> On Wed, Apr 19, 2017 at 13:42:39 -0400, Aaron M. Ucko wrote:
>
> > Package: gem2deb
> > Version: 0.33.1
> > Severity: normal
> >
> > Please have compiled Ruby extensions depend on compatible versions of
> > the ruby metapackage; for instance, an extension built only for 2.3
> > would depend on ruby (>= 2.3), ruby (<< 2.4), and an extension built
> > for both 2.3 and 2.4 would depend on ruby (>= 2.3), ruby (<< 2.5).
> > (Python packaging settled on an analogous approach some time ago.)
> > Ideally, this dependency would go to a dedicated ${ruby:Depends}
> > substvar, but adding it to ${misc:Depends} would allow for much faster
> > adoption.
> >
> > As it stands, we can get situations like #860512, in which the m68k
> > build of telegram-desktop failed because ruby 2.3 wound up paired with
> > a ruby-fast-xs installation that only covered 2.2. (The 0.8.0-3+b2
> > m68k binNMU was intended to add 2.3 support, but accidentally picked
> > up old metapackages.) With an explicit versioned dependency in place,
> > generic automated tools would have caught this problem.
> >
> As this allows broken package combinations, it seems to me the severity
> should be serious, not normal.
fair enough. Can it algo be ignored for stretch? I reallt don't think we
should change that this late.
Attachment:
signature.asc
Description: PGP signature