On Thu, Nov 22, 2012 at 13:45:57 +0100, Niels Thykier wrote: > Unfortunately, it turns out that libitext-java declares no > dependnecies relations (except libitext-java-gcj) in testing at all, > which is obvously wrong. > In sid, I have taken the liberty of properly fixing this by using > javahelper to compute the proper dependency relations automatically > (as that version cannot migrate to testing anyway being built against > the bouncycastle in sid). > > The question is then, should I backport the changes for > auto-generating the dependencies and insert the upper bound on the > bouncycastle dependency -OR- should I simply manually hardcode all the > missing dependencies in the tpu upload? > The former seems safer, so I think I'd go with that.
> I have attached the debdiff between libitext-java in sid and testing > to give you an idea of the changes involved. In case you are > wondering about the explicit version'ed dependencies in the diff, it > is because the tools involved are not able to properly give version > bounds on the dependencies[0]. > > ~Niels > > [0] Sadly, Java packages do not have a shlibs/symbols system (nor the > "rename package on ABI breakage"-proceedure). They really should start doing at least the latter. Cheers, Julien
signature.asc
Description: Digital signature
