Holger Levsen wrote:
> maybe this can be solved by having two (or three) diffoscope packages:
One thing I would like to ensure that we do not cause a reduction in
functionality for some extremely prevalent use-cases.
I'm not just thinking of end-users here: in fact, I am thinking of
various automated systems such as (a) tests.reproducible-builds.org,
(b) try.diffoscope.org, as well as (c) the reproducible testing that
is performed via salsa.debian.org's own CI.
These are, AIUI, all installing "diffoscope" with recommends enabled,
or at least they work on the expectation that installing the
recommends of this package will result in a fully-featured Diffoscope.
Of course, all of these could be changed to install a -full package or
to install Suggests, etc. etc. but I think we should retain the
current expectations for now. After all, there may be far more than
the three use-cases I listed above that we have even less control
over.
In technical terms, this would seem to imply a diffoscope-core (or a
diffoscope-minimal or a diffoscope-bikeshed), that would (at least)
not Recommend the largest packages Ximin highlighted.
Regards,
--
,''`.
: :' : Chris Lamb
`. `'` la...@debian.org 🍥 chris-lamb.co.uk
`-
