
Hi Duncan, The major problem that I see is with # cabal update # cabal upgrade sequence of operations (that is, upgrading all installed packages). I use this regularly to keep packages that I use (or are going to use) up to date. This is the functionality that, I suppose, everybody expects from the system like hackage/cabal. And, so far as I can see, the major problem is that 'cabal upgrade' will happily upgrade packages, that the core package depends on, and eventually will create configuration with circular dependencies. Can we do something simple - like, for example, forbidding updates of the packages that the core depends on (directly and indirectly)? -- Valentyn. On Feb 2, 2009, at 5:34 PM, Duncan Coutts wrote:
On Tue, 2009-02-03 at 02:11 +0100, Achim Schneider wrote:
Duncan Coutts
wrote: That is probably how people are getting into this mess. Using upgrade is not necessarily such a good idea. It does not distinguish between the interesting packages you might want to upgrade and the core packages that your probably do not want to touch.
There's no need to use cabal upgrade to get into this kind of mess, I successfully pulled conflicting cabal versions just by using cabal install.
Indeed, it's just easier to shoot yourself in the foot using upgrade.
Would anyone object if I disable the upgrade command in the next version and not re-enable it until it works?
Duncan
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe