[Haskell-cafe] Re: Why cabal forces reinstallation of happstack-utils?

[Henning Thielemann]

Ivan Lazar Miljenovic schrieb:
> Henning Thielemann <schlepptop at henning-thielemann.de> writes:
>> This makes sense. It would be certainly better if cabal-install would
>> alert about the found inconsistency instead of trying to fix it. Then
>> there might an additional cabal-install flag, that makes cabal-install
>> prefer the output of ghc-pkg over the hackage database 00-index.tar in
>> case of inconsistencies.
> 
> What do you mean by "prefer the output of ghc-pkg over the hackage
> database"?  If there is a broken/invalid dependency chain, then
> dependencies _have_ to be rebuilt to fix it.  ghc-pkg, however, doesn't
> say _how_ those packages where built or from where.
> 
>> Could the problem also be solved by changing the version of packages
>> with locally modified cabal files from a.b.c.d to a.b.c.d+1 ?
> 
> No, since it will always prefer the highest version, and if its the one
> that's broken then it will just complain that that package doens't exist
> and will refuse to build what you want.

I meant it the other way round: Given that package X-a.b.c.d does not
compile due to dependencies that turn out to be too restricted, then I
could relax the dependency version bounds in X.cabal. If I also increase
X's version to a.b.c.d+1, then cabal-install should be happy.