[Haskell-cafe] Hackage inconsistent? (pandoc.cabal in 1.13.1)

[Michael Snoyman]

Sorry to jump into this thread at a random point.

There's another aspect to this discussion that hasn't been brought up.
Imagine that I release a package bar-1.0.0.0, that depends on:

    build-depends: base, foo >= 1.1

Then, suppose foo-1.2 is released and bar-1.0.0.0 is incompatible. If you
simply release a new version of bar with a stricter upper bound, cabal may
decide to install the original bar-1.0.0.0 together with foo-1.2, which
will lead to a broken build. Blackling the package is the theoretical
solution to this problem, but may not work correctly[1]. However, by
overwriting the original cabal file with a tweaked one, cabal will be able
to make a more sensible decision.

I'm not trying to advocate anything in particular here, just point out an
aspect of the discussion that I think has been missed.

[1] https://github.com/haskell/cabal/issues/1792

On Fri, Oct 31, 2014 at 11:56 AM, Peter Simons <simons at cryp.to> wrote:

> Hi Lennart,
>
>  > Bumping [the (n+1)th digit] is not safe, as the maintainer might
>  > decide to publish a new version that adds a new component.)
>
> Suppose I publish foo-1.0. Then a sequence of edits would produce
> versions 1.0.1, 1.0.1.1, 1.0.1.1.1, and so on.
>
> Why would there be any problem?
>
> Best regards,
> Peter
>
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.haskell.org/pipermail/haskell-cafe/attachments/20141031/dcf1075e/attachment.html>