On Tue, 5 May 2015 16:16:41 +0200 Sven Bartscher <sven.bartscher@weltraumschlangen.de> wrote: > > > Am 05.05.2015 um 16:10 schrieb Joachim Breitner <nomeata@debian.org>: > > Am Dienstag, den 05.05.2015, 16:02 +0200 schrieb Sven Bartscher: > >>> Am 05.05.2015 um 10:38 schrieb Joachim Breitner <nomeata@debian.org>: > >>> your fix #778987 might have been too tight: Since you have a strict "=" > >>> dependency, whenever we have to binNMU libghc-cabal-dev, we also have to > >>> do the same to cabal-install, and in the meantime, cabal-install is > >>> uninstallable in sid. I think we should try to avoid that. > >>> > >>> Wouldn’t it be sufficient to have a ">=" dependency on the version it > >>> was built with? Possibly even ">= <upstream-part-of-the-version>“? > >> > >> I guess it doesn’t need to depend on the exact version it was built against. > >> I would suggest > >>> = <upstream-part-of-the-version> && << <upstream-part-of-the-version>.1 (this .1 is intentional) > >> This way cabal-install needs to be rebuilt whenever there is a new upstream > >> version of libghc-cabal-dev. > >> If we don’t set that upper bound, we risk that libraries (like hgettext) are > >> built against newer upstream versions of libghc-cabal-dev than cabal-install, > >> which could cause a mismatch between the CLI of cabal-install and that of > >> the library. > > > > sounds ok. Care to prepare a patch? > > I can do that next weekend. I made the changes, but I couldn't test to build the package due to my broken sbuild. (see [1] for reason if you're interested) If someone would like to test the package for me, I could push the untested changes (which I would like to avoid under normal circumstances). Regards Sven
Attachment:
pgpBdPwXkGvEE.pgp
Description: Digitale Signatur von OpenPGP