On building everything from source.
Le Fri, Dec 27, 2013 at 09:36:33AM +0800, Paul Wise a écrit :
> On Thu, Dec 26, 2013 at 7:34 AM, Paul Wise wrote:
>
> > Building as much as possible from source at package build time should
> > be the norm not the exception, otherwise the build process for that
> > source will bitrot and we will never detect that.
>
> Having said that, I expect there will be exceptions to this (examples
> I can think of below). We currently don't have any standard way to
> check that these exceptions are still buildable from source.
Thanks for the clarification, Paul.
On my side, I feel an increased social and sometime formal pressure to build
things from source with no clear limit, and this is chilling effects that make
me abandon more and more packages and focus on the most trivial works to
package. More becomes less, for instance when Upstream adds pictures to their
documentation, etc.
I would welcome clearer rules to decide what is strictly required to be built
from source. In particular, I would welcome something like "The maintainer
decides, and can be overruled by the Technical Comittee if needed". What
follows is not intended as a gratuitous rant but is my experience of a
packager: the long delays at the NEW queue, the variability of the decisions
(which depend on who reviews the package), the unpredictability of when a
source package goes to NEW (that is, when the list of binary packages changes),
and the impossibility of using our archive as a guide of what is acceptable
(since it contains things that would not be accepted anymore) are making it
too difficult to support packages over years.
Needless to say, since we have a nice upstream guide, reminding Upstream about
the importance of being as rebuildable as possible is something that is easy to
be done. But not if the emails finish by things like "in the meantime, we have
deleted your documentation from our packages".
Have a nice day,
--
Charles Plessy
Illkirch-Graffenstaden, France
Reply to: