[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Review of fairsim, libjtransforms-java and libjlargearray-java



Am 02.10.2017 um 18:35 schrieb Emmanuel Bourg:
> Le 2/10/2017 à 17:51, Markus Koschany a écrit :
> 
>> There is at least one major downside of the current behavior. You
>> completely loose the flexibility to choose that your package is able to
>> run with an older version of libfoo. There is always a strict dependency
>> on the latest version which can be a real hassle if you try to backport
>> packages. Unfortunately we don't have a C/C++ mechanism like symbols
>> files in Java but the answer should not be to enforce a dependency on
>> the latest version.
> 
> Ideally the version used for the resolved dependencies should be the
> ones defined in the pom, such that the packages express the same
> requirements than the Maven artifacts they contain. That would be a less
> strict middle ground.

In my opinion this should have been a new option where users can choose
to opt-in. I find it less than optimal that the behavior of an existing
option was changed without further notice and documentation and now
everyone is expected to cope with the situation.

> Regarding the backports, what the current maven-debian-helper behavior
> really limits is the ability to download a .deb built for sid and
> install it directly in stable.

That would be an interesting feature but I think it is unrelated to the
problem at hand.

 Rebuilding the package for stable is
> still possible with no extra modification, and the resolved dependencies
> will then refer to the versions in stable.

Ok, this is true. But the dependency will be >= the latest version in
backports or stable. This makes it more complicated to use customized
packages. In the end the user is forced to override the strict
dependency in debian/control if he prefers to use an earlier package
version.

Markus

Attachment: signature.asc
Description: OpenPGP digital signature


Reply to: